Breast

Imaging and
Interventional
 BR001-EB-X

Male Breast Disease: Pictorial Review of Multimodality Images

All Day Room: NA Hardcopy Backboard

Participants
Jinah Kim, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Yun Woo Chang, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eunji Lee, MD , Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa jin Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jiyoung Hwang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Seong Sook Hong, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ywchang@schmc.ac.kr
TEACHING POINTS
1. Review of uncommon male breast disease that prone to many of the same pathological processes as female breasts. 2. Review of
various male breast disease by multimodality breast images including mammography, ultrasonography, breast MRI or chest CT scan.
TABLE OF CONTENTS/OUTLINE
1. Characteristic imaging features of gynecomastia that the most common abnormality of male breast 2. Benign male breast
diseases including lipoma, myofibroblastoma, fibroadenomatous change, fat necrosis, abscess, cellulitis, intramammary lymph nodes
3. Malignant male breast disease including infiltrative ductal carcinoma, liposarcoma, metastatic carcinoma of axilla
 BR002-EB-X

Synthetic Mammography: Advantages, Disadvantages, Artifacts, and Pitfalls

All Day Room: NA Hardcopy Backboard

Participants
Erin Alencherry, MD, Cleveland, OH (Presenter) Nothing to Disclose
Shiraz Rahim, MD, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Zubair Syed, MD, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Donna M. Plecha, MD, Strongsville, OH (Abstract Co-Author) Research Grant, Hologic, Inc
Holly N. Marshall, MD, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

erinalencherry@gmail.com
TEACHING POINTS
To discuss synthetic mammography (SM) in terms of its acquisition technique from digitial breast tomosynthesis (DBT) data set To
discuss the differences and advantages of DBT/SM compared to DBT/ full field digital mammography (FFDM) To discuss the
appearance of benign and pathologic findings on SM and how the technique is different from FFDM To highlight artifacts and
potential pitfalls of DBT and SM
TABLE OF CONTENTS/OUTLINE
1. Introduction to DBT/SM a. Discuss DBT b. Discuss SM 2. Advantages/differences of DBT/SM compared to DBT/FFDM a.
Decreased radiation dose b. Shorter acquisition time, decreasing motion c. SM algorithm enhances certain features, examples of the
following will be provided i. calcifications ii. margins iii. distortion 3. Artifacts and potential pitfalls of DBT and SM a. High density
artifacts i. Shadowing ii. Zipper or slinky artifact b. Motion i. Bounce ii. Zig-zag slinky artifact iii. Blur c. Pseudocalcifications i.
Appearance of pseudocalcifications on SM ii. Discussion of looking at DBT data set and other synthetic image d. Pseudoasymmetry
e. Decreased axillary contrast resolution i. Appearance of axilla on SM ii. Discussion of evaluating the axilla on DBT images
 BR003-EB-X

Is It Really a Simple Intramammary Lymph Node?

All Day Room: NA Hardcopy Backboard

Participants
Eduardo V. Ferreira, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Vitor A. Sperandio, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
David C. Bastos, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Almir Bitencourt, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana A. Souza, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Elvira F. Marques, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Maria Luiza D. Albuquerque, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Camila Guatelli, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciana Graziano, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
eduardovlemesf@hotmail.com
TEACHING POINTS
The purpose of this exhibit is to discuss the evaluation of simple intramammary lymph nodes trough a multimodality pictorial essay,
utilizing mammography, ultrasound and MRI. The increase of breast cancer screening exams nowadays led to the rising of masses
findings. The differentiation of other breast masses with habitual lymph nodes have extremely important relevance, since the last
are benign findings, so any conduct other than routine screening is not adequate, raising the cost to health system with additional
exams, raising the x-ray dosage in the patient and could lead to unnecessary biopsies. They are usually less than 1 cm in greatest
diameter. Being seen on mammography as a circumscribed oval or reniform non-calcified mass with a central or peripheral lucency
that represents fat within the hilum. Without exception the nodes are smoothly circumscribed and well defined and have low
density. If those features are not present it is not a simple lymph node. Thus, the radiologists have to be aware how to diagnostic
simple intramammary lymph nodes to lead the patient in the proper assessment.
TABLE OF CONTENTS/OUTLINE
Introduction Typical characteristics Atypical characteristics. Conclusion
 BR004-EB-X

Radiation Treatment Planning: What the Breast Imager Should Know

All Day Room: NA Hardcopy Backboard

Participants
Tyler Litton, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Jarred P. Tanksley, MD, Durham, NC (Abstract Co-Author) Nothing to Disclose
Debbie L. Bennett, MD, Saint Louis, MO (Abstract Co-Author) Reviewer, Biomedical Systems
For information about this presentation, contact:
tyler.litton@duke.edu
TEACHING POINTS
-Know the indications and rationale for radiation therapy in lumpectomy and mastectomy patients -Understand how breast cancer
stage affects treatment options -Appreciate how axillary and internal mammary lymph node involvement affects the radiotherapy
treatment field -Become familiar with the radiation treatment process (treatment planning, number of visits, timing of radiation after
surgery)
TABLE OF CONTENTS/OUTLINE
-Review the role of radiation therapy in breast cancer treatment -Discuss which patients require radiation therapy as part of
treatment plan -Discuss margin status in breast conservation therapy and radiation field considerations as informed by the American
Society for Radiation Oncology (ASTRO) Clinical Practice Statements -Discuss implications of lymph node involvement on the
radiation therapy field and how this may contribute to late effects (e.g. lymphedema, coronary artery disease, etc.) -Illustrate
experience of radiation therapy from patient perspective (timing, number of visits, duration of visits) -Present 2-4 multi-modality
cases which highlight pertinent information provided by the breast radiologist that influences size and type of radiation field (e.g.
axillary and/or internal mammary lymphadenopathy, multifocal/multicentric disease, inflammatory breast cancer, local recurrence)
 BR005-EB-X

Complex Cystic and Solid Breast Masses - What Should I Know?

All Day Room: NA Hardcopy Backboard

Participants
Vinicius C. Felipe SR, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Marilia M. Azevedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Almir Bitencourt, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana A. Souza, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Maria Luiza D. Albuquerque, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciana Graziano, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Elvira F. Marques, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Mirian R. Poli, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Camila Guatelli, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rubens Chojniak, MD, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
The purpose of this exhibit was to evaluate complex cystic and solid breast lesions by mammography, ultrasonography (US) and
advanced magnetic resonance (MR) techniques, correlating imaging with histologic findings. Complex cystic and solid masses have a
substancial chances of being malignant (23 - 31%). For that reason, percutaneous or surgical biopsy is usually indicated and US is
the modality of choice for characterizing and guiding biopsy. Common benign causes of complex cystic and solid masses include
fibrocystic changes, intraductal or intracystic papilloma without atypia, fibroadenoma and phyloides tumor. Atypical findings include
atypical ductal hyperplasia, lobular neoplasia, and atypical papilloma. Malignant findings include ductal carcinoma in situ (DCIS),
intracystic papillary? carcinoma and invasive carcinomas. Advanced MR techniques like dynamic imaging provide a high level of
diagnostic confidence in the characterization of complex cystic breast lesion, thus allowing early diagnosis and significantly reducing
patient morbidity and mortality.
TABLE OF CONTENTS/OUTLINE
Introduction Complex cystic and solid mass Enhancement breast lesions Benign lesions Malignant lesions Breast MRI Conclusion
 BR006-EB-X

Malignant Breast Lesions: Can We Predict the Tumor Subtype?

All Day Room: NA Hardcopy Backboard

Participants
Pamela Garcia Suarez, MD, Pilar, Argentina (Presenter) Nothing to Disclose
Javier M. Martinez Martinez, MD, Pilar, Argentina (Abstract Co-Author) Nothing to Disclose
Karen M. Palacio, MD, Pilar, Argentina (Abstract Co-Author) Nothing to Disclose
Eunice A. Spengler, Pilar, Argentina (Abstract Co-Author) Nothing to Disclose
Dario M. Schejtman, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Ignacio Mc Lean, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
PGARCIAS@cas.austral.edu.ar
TEACHING POINTS
1- To recognize the morphological findings in oncological breast lesions according to tumor subtype. 2- To compare our imaging
findings with the published bibliography. 3- To describe the importance correlating radiological and histopathological findings.
TABLE OF CONTENTS/OUTLINE
Introduction. Background. Radiological findings related to tumor subtype. Luminal A. Luminal B. Triple Negative. HER2/Neu positive.
Discussion. Conclusion.
 BR012-EB-X

Incidental Extra-Mammary Findings on Breast MRI: A Pictorial Illustration

All Day Room: NA Hardcopy Backboard

Participants
Ali A. Al-Saraf, MD, Chicago, IL (Presenter) Nothing to Disclose
Deepa Sheth, MD, Chicago, IL (Abstract Co-Author) Research Grant, Guerbet SA
For information about this presentation, contact:
ali.al-saraf@uchospitals.edu
TEACHING POINTS
-Utilize a systematic approach for interpreting extra-mammary findings seen on breast MRI. -Become familiar with the frequency of
extra-mammary findings seen on breast MRI based on organ system. -Recognize the utility of pursuing additional imaging/diagnostic
work-up of the findings identified.
TABLE OF CONTENTS/OUTLINE
-A pictorial illustration of the cases encountered at our institution will be presented with emphasis on pertinent findings and workup.
-A total of 20 cases will be presented with findings grouped into the following categories: osseous, hepatic, biliary,
cardiac/vascular, pulmonary, neurology, and endocrine. -A systematic approach will be discussed including the utility of additional
workup of the pathology presented.
 BR100-ED-X

Imaging of Lesions of Uncertain Malignant Potential on Core Biopsy

All Day Room: NA Digital Education Exhibit

Participants
Rosa M. Lorente-Ramos, MD, PhD, Madrid, Spain (Presenter) Nothing to Disclose
Javier Azpeitia Arman, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Teresa Rivera Garcia, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Miguel Angel Lara Alvarez, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Jose Maria Lopez-Arcas Calleja, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Carlos Oliva Fonte Sr, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rosa.lorenteramos@salud.madrid.org
TEACHING POINTS
To review the pathological classification of lesions, with emphasis on lesions of uncertain malignant potential (B3). To illustrate
clinical presentation, imaging findings (mammogram, US, MR) of cases from our series of breast lesions with pathologic correlation.
To emphasize pitfalls, diagnostic difficulties, differential diagnosis and management of these lesions.
TABLE OF CONTENTS/OUTLINE
We present: 1.Histopathological classification of breast lesions. 2. Definition of B3 lesions (uncertain malignant potential). 3.
Imaging (Mammograms, US, MR) of the different entities included in this group: atypical epithelial proliferation of ductal type, lobular
intraepithelial neoplasia, complex sclerosing lesions/radial scar, papillary lesions and columnar cell change. 4. Differential diagnosis
and management.Tips- Discordant diagnoses must be reviewed, sometimes repeat core biopsy or surgery are mandatory. -
Remember entities which may cause an underestimation on core biopsy to choose the best approach in patient management
 BR101-ED-X

BRCA Mash-Up: Pictorial Review of BRCA-Associated Breast Cancer Phenotypes

All Day Room: NA Digital Education Exhibit

Participants
Lindsey Storer, MD, Los Angeles, CA (Presenter) Nothing to Disclose
Cheryce P. Fischer, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
Stephanie A. Lee-Felker, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Melissa M. Joines, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
Women with BRCA1 or BRCA2 mutations face up to a 65% lifetime risk of developing breast cancer. BRCA1 and BRCA2-associated
breast cancers have been shown to have distinct imaging features and radiologists must recognize the characteristics of these
tumors in order to increase the likelihood of earlier detection. Additionally, BRCA1 cancers tend to be more aggressive than BRCA2
cancers, with more mammographically occult cancers in the BRCA1 population. After reviewing this exhibit, participants will
understand key clinical differences resulting from BRCA1 and BRCA2 mutations and describe the associated characteristic imaging
features.
TABLE OF CONTENTS/OUTLINE
• Epidemiology of BRCA mutations • Key clinical differences resulting from BRCA1 and BRCA2 mutations • Comprehensive imaging
review highlighting the characteristic findings of BRCA1 and BRCA2-associated breast cancers on 2D and 3D mammography,
ultrasound, and MRI
 BR102-ED-X

Technique and Benefits for Same Day Dual-Energy Contrast-Enhanced Mammography and Breast Biopsy
All Day Room: NA Digital Education Exhibit

Participants
Noelle E. Hoven, MD, Minneapolis, MN (Presenter) Nothing to Disclose
Mark C. Wickre, MD, Minneapolis, MN (Abstract Co-Author) Nothing to Disclose
Tim H. Emory, MD, Saint Paul, MN (Abstract Co-Author) Nothing to Disclose
Michael T. Nelson, MD, Minneapolis, MN (Abstract Co-Author) Nothing to Disclose
An L. Church, MD, Minneapolis, MN (Abstract Co-Author) Nothing to Disclose
Jessica E. Kuehn-Hajder, MD, Minneapolis, MN (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ekwoc002@umn.edu
TEACHING POINTS
Dual-energy contrast enhanced mammography (DECM) is a technique using intravenous contrast injection with dual energy
radiography to detect any additional enhancing breast lesions. In our experience of over 500 cases, DECM has been more effective
than standard mammography in identifying occult lesions in the breast tissue of patients presenting for biopsy for a known lesion.
Contrast mammography has the potential to be complimentary in a center with access to gold standard contrast enhanced MRI,
and a reasonable substitute when MRI is not possible for any number of reasons. Done on the biopsy day, any new lesion(s) can
undergo workup at this time; bringing a more complete assessment while streamlining patient care. Describe the dual energy
contrast enhanced mammography technique Detail the benefits and possible drawbacks in using this technique Provide a
comparison of DECM with contrast enhanced MRI
TABLE OF CONTENTS/OUTLINE
Dual energy contrast enhanced mammography technique Our breast center protocol Advantages and disadvantages of DECM
Comparison of contrast enhanced MRI and contrast mammography as reported in the literature Future directions - We are in the
process of reviewing our 500+ DECM exams to quantify extent and pathology of occult enhancing breast lesions
 BR103-ED-X

Breast Cancer Recurrence in the Post-Operative Breast: Catch Me If You Can!

All Day Room: NA Digital Education Exhibit

Participants
Rute S. Rothwell, Worcester, MA (Presenter) Nothing to Disclose
Jade Watkins, MD, Worcester, MA (Abstract Co-Author) Nothing to Disclose
Carolynn M. DeBenedectis, MD, Worcester, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Rute.Rothwell@umassmed.edu
TEACHING POINTS
1. Discuss common mammographic findings suggestive of breast cancer recurrence: in Post Lumpectomy Breast TRAM Flap
Mastectomy 2. Discuss current recommendations for post-operative surveillance mammography
TABLE OF CONTENTS/OUTLINE
1. The following imaging findings will be reviewed: Features of recurrence in Post Lumpectomy Breast Features of recurrence in
Tram Flap Features of recurrence in Mastectomy Available screening guidelines
 BR104-ED-X

A Radiologist's Practical Guide to use of DWI during Breast MRI

All Day Room: NA Digital Education Exhibit

Participants
Frederick Kelcz, MD, PhD, Madison, WI (Presenter) Shareholder, Elucent Ringers, LLC
For information about this presentation, contact:
fkelcz@uwhealth.org
TEACHING POINTS

1. Diffusion weighted imaging (DWI) can aid MRI interpretation by improving specificity, often the bane of this modality. 2. Apparent
diffusion coefficient (ADC) values of fibroadenomas are typically in the range of 1.5 - 2.0 x 10-3 mm2/sec, quite different from
invasive ductal cancer, which is typically below 1.2 x 10-3 mm2/sec. Actual ADC values associated with any tissue will depend on
details of the diffusion sequence. One role for DWI may be in confirming fibroadenoma for lesions with some atypical BI-RAD
features. 3. Fat in the breast, intermixed with glandular tissue imposes limitations on appropriate use of diffusion data 4. Breast MRI
is often associated with image distortion and displacement, challenging interpretation of small lesions. Strategies for coping with
this phenomenon will be discussed.
TABLE OF CONTENTS/OUTLINE
1. Need to know physics of DWI; meaning of ADC values. 2. Examples of displacement and distortion; how to compensate for these
limitations 3. Examples of mass and non-mass benign and malignant lesions using conventional BI-RADS and application of diffusion
data - when to use it, when not to use it. 4. ADC 'double threshold' so as not to confuse fibroadenoma and mucinous breast
cancer. 5. Review of recent literature on advanced diffusion methods such as Kurtosis imaging
 BR105-ED-X

Multimodality Approach to the Nipple-Areolar Complex: A Pictorial Review and Diagnostic Algorithm
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Mireia Pitarch, MD, Sabadell, Spain (Presenter) Nothing to Disclose
Javier H. Del Riego, MD, Sabadell, Spain (Abstract Co-Author) Nothing to Disclose
Clara C. Aroztegui, Sabadell, Spain (Abstract Co-Author) Nothing to Disclose
Laura Nebot, Sabadell, Spain (Abstract Co-Author) Nothing to Disclose
Sandra Medina, Sabadell, Spain (Abstract Co-Author) Nothing to Disclose
Oscar Aparicio, MD, Sabadell, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
mireiapitarch@gmail.com
TEACHING POINTS
* We describe and illustrate the normal anatomy and anatomic variants of the nipple-areolar complex (NAC). * We show the
imaging techniques used to study the NAC. * We review benign and malignant conditions that can affect the NAC, and propose a
diagnostic algorithm for the imaging workup.
TABLE OF CONTENTS/OUTLINE
Normal anatomy and anatomic variants of the NAC. Imaging techniques used to evaluate the NAC (mammography, ultrasound, MRI,
galactography). Benign and malignant processes that involve the NAC and their appearance on imaging modalities, with clinical-
pathological correlations. Diagnostic imaging algorithm.
 BR106-ED-X

Breast Asymmetry. What Does It Mean? What Should I Do?

All Day Room: NA Digital Education Exhibit

Participants
Rosa M. Lorente-Ramos, MD, PhD, Madrid, Spain (Presenter) Nothing to Disclose
Javier Azpeitia Arman, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Carmen Estrada, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Jose Maria Lopez-Arcas Calleja, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Jose Manuel Garcia Gomez, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Maria Angeles Martinez Izquierdo, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rosa.lorenteramos@salud.madrid.org
TEACHING POINTS
To review the definition and classification of asymmetry in breast mammograms. To illustrate imaging features (mammogram, US,
MR) of cases from our series with pathologic correlation. To analyze and discuss the specific management of those lesions,
including diagnostic difficulties and imaging work-up (additional mammogram projections, US, tomosynthesis, MR) To emphasize
pitfalls and clues to differential diagnosis.
TABLE OF CONTENTS/OUTLINE
- BIRADS Fifth edition classification and definition of asymmetries: asymmetry, global asymmetry, focal asymmetry , developing
asymmetry. -Different possible etiologies of each type. -Difficulties in imaging and diagnostic work-up: additional mammogram
projections, US, tomosynthesis, MR, comparison with previous studies and follow-up. - A case-based review with pathological
correlates will illustrate imaging findings and pathological correlation of challenging cases including, ductal and lobar breast cancer,
pseudoangiomatous stromal hyperplasia (PASH), hormonal changes.
 BR107-ED-X

Tumor Biology and Imaging Manifestations: Can You Match the Imaging Findings to the Correct Tumor Type?
All Day Room: NA Digital Education Exhibit

Participants
Linda DeMello, MD, Providence, RI (Presenter) Nothing to Disclose
Ana P. Lourenco, MD, Providence, RI (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ldemello2@lifespan.org
TEACHING POINTS
1. To discuss the genetic biomarkers and biologic behavior of breast cancers a. ER, PR, Her2/Neu or ERRB2 receptor, Ki-67 protein
b. Luminal A, Luminal B, Basal, Triple Negative, Her-2 Enriched subtypes 2. To present a series of cases from our tertiary breast
center in an interactive format to illustrate common radiologic manifestations of breast cancer subtypes a. Can we predict the
subtype based on imaging? 3. To recognize the importance of these imaging findings and how they correlate with treatment and
prognosis 4. Estimate future implications of tailored personalized treatment in breast cancer
TABLE OF CONTENTS/OUTLINE
1. Introduction to breast cancer biology 2. Compare and contrast breast cancer subtypes and their pathophysiology a. Luminal A,
Luminal B, Basal, Triple Negative, Her-2 Enriched 3. Interactive case review of the imaging findings of each subtype a. Multiple
pathology-proven cases from our institution 4. Treatment and prognosis of each subtype 5. Future directions in personalized
treatment of breast cancer
 BR108-ED-X

Basic Concepts and Implications of Molecular Classification of Breast Cancer

All Day Room: NA Digital Education Exhibit

Participants
Rosa M. Lorente-Ramos, MD, PhD, Madrid, Spain (Presenter) Nothing to Disclose
Javier Azpeitia Arman, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Miguel Angel Lara Alvarez, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Jose Maria Lopez-Arcas Calleja, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Berta Obispo Portero, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Jose Manuel Garcia Gomez, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rosa.lorenteramos@salud.madrid.org
TEACHING POINTS
Review the classification of subtypes in breast cancer. Analyse implications of breast cancer subtyping in clinical settings: different
treatment , survival, follow-up. Discuss the imaging findings of each molecular subtype .
TABLE OF CONTENTS/OUTLINE
Introduction to molecular classification of breast cancer: intrinsic subtypes (gene expression), surrogate intrinsic subtypes (defined
by routine histology and IHC data). Definitions of the major molecular subtypes of invasive breast cancer. Clinical implications:
different molecular profiles with different degrees of treatment sensitivity , patterns of spread (lymph nodes, metastases) and
different outcomes. Discussion of different imaging findings in the different subtypes. A case-based review with pathological
correlates will illustrate imaging findings and limitations in diagnosis, posttreatment evaluation and follow-up of the different
subtypes of breast cancer.
 BR109-ED-X

You Found What in My Breast?!

All Day Room: NA Digital Education Exhibit

Participants
Jennifer E. Caero, MD, Dallas, TX (Presenter) Nothing to Disclose
Joseph Spigel, MD, Coppell, TX (Abstract Co-Author) Nothing to Disclose
Zeeshan A. Shah, MD, Indianapolis, IN (Abstract Co-Author) Nothing to Disclose
Raynal R. Hamilton, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
David Parish, MD, Jonesboro, AR (Abstract Co-Author) Nothing to Disclose
Sean D. Raj, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
sean.raj@gmail.com
TEACHING POINTS
Review benign pathologies and findings that can simulate malignant disease in the breast, including infectious etiologies, foreign
bodies, dermal, and iatrogenic processes. Identify these entities based on characteristic multi-modality imaging appearances.
Review tips, tricks, potential pitfalls, mimics and the role of clinical history in the diagnosis of these entities.
TABLE OF CONTENTS/OUTLINE
We will present a case-based pictorial essay of our experience from our academic and community-based breast imaging network.
Cases include: Foreign bodies/Iatrogenic Post-surgical gossypiboma including surgical mesh, towel, and staples Bullet fragments
mimicking calcification Retained catheter fragments Post-laparotomy emphysema Sewing and acupuncture needles Infection The
'fillarial dance' (active, mobile parasite on US cine) Calcified chronic fillarial infection Calcified trichinosis Cysticercosis Dermal Tattoo
pigment-nodal/cutaneous Neurofibromatosis Hidradenitis Fat necrosis following Methylene blue injection prior to sentinel node Other
Amyloidosis from Sjogren Syndrome Poland Syndrome Peri-implant effusion Gold therapy sequelae
 BR110-ED-X

Radioactive Seed Localization-Speed Bumps of Early Adoption: How to Pave the Way to Successful Use of
Non-Wire Localizations
All Day Room: NA Digital Education Exhibit

Participants
Geoffrey H. Miller, MD, Charlottesville, VA (Presenter) Nothing to Disclose
Jason W. DeBerry, MD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Shayna Showalter, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Jennifer A. Harvey, MD, Charlottesville, VA (Abstract Co-Author) Stockholder, Hologic, Inc; Research Grant, Volpara Health
Technologies Limited; Stockholder, Volpara Health Technologies Limited;
David Brenin, MD, Charlottesville, VA (Abstract Co-Author) Research funded, Theraclion
Carrie M. Rochman, MD, Charlottesville, VA (Abstract Co-Author) Research Consultant, Theraclion
Jonathan Nguyen, MD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Matthew M. Miller, MD, PhD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Anneke Schroen, MD,MPH, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Brandi T. Nicholson, MD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
GHM7Y@hscmail.mcc.virginia.edu
TEACHING POINTS
Understand the advantages and disadvantages of using radioactive seed localization versus needle-wire localization. Identify
potential pitfalls and unique situations which are specific to non-needle-wire localizations, specifically radioactive seeds. Describe
strategies that can be used to overcome these pitfalls.
TABLE OF CONTENTS/OUTLINE
Overview: a brief history of radioactive seed and other non-needle-wire localization techniques Brief description of the
implementation of a radioactive seed localization program Case examples of potential problems and unique situations involving
radioactive seed localization Hematoma/seroma at/near localization site Flanking localization Localization of axillary lymph nodes
More than one ipsilateral finding Malpositioned radioseed Summary Review when the use of radioseeds may be contraindicated
Suggestions for new users
 BR111-ED-X

Optimizing Mammographic Positioning: One Technologist at a Time

All Day Room: NA Digital Education Exhibit

Participants
Lena F. Gowharji, MD, Burlington, MA (Presenter) Nothing to Disclose
Jennifer C. Broder, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Christopher B. Martel, MS, Andover, MA (Abstract Co-Author) Nothing to Disclose
Cathleen M. Kim, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Meera Sekar, MD, Lexington, MA (Abstract Co-Author) Nothing to Disclose
Jeanette Y. Chun, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Audrey L. Hartman, MD, MS, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Meaghan Mackesy, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Kristine Mallinson, Lexington, MA (Abstract Co-Author) Nothing to Disclose
Michelle R. McSweeney, DO, Burlington, MA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1. Describe the importance of breast positioning in mammography and the new MQSA requirements for ongoing assessment and
improvement of image quality 2. Demonstrate how large batch analysis of positioning metrics can provide insight into trends of
deficiencies by technologist and the opportunity for directed training 3. Exhibit how a one-time teaching intervention is not
sufficient
TABLE OF CONTENTS/OUTLINE

1. Importance of positioning in mammography a. Cases b. Description of MQSA EQUIP Initiative c. Educational review of seven
positioning quality metrics computed from images 2. A program for assessment & improvement of breast positioning a. Description of
program Software for analysis of large batches of images for positioning Results analyzed by technologists for individualized
improvement b. Intervention: Individualized training provided c. Results Initial analysis Results after re-education by trainer 3.
Challenges and future goals a. Lessons learned from initial experience Initial re-training in one area led to increased deficiencies in
other areas Improvements not sustained, need for continuous monitoring/interventions b. Future goals Potential for frequent/real-
time feedback and ultimately improved patient care
 BR112-ED-X

Preparation of Digital Mammograms for the Application of Deep Learning Algorithms

All Day Room: NA Digital Education Exhibit

Participants
Alexandra (Ali) Silver, BSC, Kelowna, BC (Presenter) Nothing to Disclose
Yuhao Huang, Kelowna, BC (Abstract Co-Author) Nothing to Disclose
Carson McKay, BSC, Kelowna, BC (Abstract Co-Author) Nothing to Disclose
Rasika Rajapakshe, PhD, Kelowna, BC (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ali.silver@alumni.ubc.ca
TEACHING POINTS
This educational presentation will illustrate the following: 1. Annotation of Breast abnormalities recorded as DICOM Grayscale
Softcopy Presentation State (GSPS) (DICOM PR Modality); 2. De-identification of DICOM files; 3. Description of DICOM PR Graphic
Objects including Coordinates, Area, Mean, Standard Deviation and Shape of Object; 4. Utilization of PR Graphic Objects to create
a Binary Mask image, referencing to the corresponding mammogram (Image Pair); 5. Generation and display of an Image Pair for
each PR file and its referenced mammogram, and; 6. Generation of PNG files from Image Pairs for application in Machine Learning
Algorithm.
TABLE OF CONTENTS/OUTLINE
1. Introduction and background on Machine learning. | 2. How Annotation of Breast abnormalities are saved as a DICOM Grayscale
Softcopy Presentation State (GSPS) (DICOM PR Modality). | 3. Explanation of PR file format. | 4. Description of the creation of a
"PR Viewer" program to create and display the Binary Mask image and its corresponding mammogram. | 5. Utilization of Image Pairs
for Machine Learning Algorithm.
 BR113-ED-X

Embracing EQUIP in Our Daily Mammography Environment: Best Practices and Lessons Learned (The EQUIP
Way or the Highway)
All Day Room: NA Digital Education Exhibit

Participants
Emily Marshall, PhD, Chicago, IL (Presenter) Nothing to Disclose
Yue Zhang, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Zheng Feng Lu, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
David V. Schacht, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Ingrid Reiser, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Kirti M. Kulkarni, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Emily.Marshall2@uchospitals.edu
TEACHING POINTS
1. Understanding the rationale for implementing Enhancing Quality Using the Inspection Program (EQUIP). 2. Initiatives taken at our
institution to successfully implement EQUIP. 3. Demonstrate how a strong, interdisciplinary team can create a manageable
workflow. 4. Understand how the EQUIP workflow supports medical outcomes audit preparation.
TABLE OF CONTENTS/OUTLINE
In our institution, we have successfully implemented EQUIP since July of 2017. The team comprises of lead radiologist,
mammography manager, QC technologist and imaging physicists. A dedicated tool for monthly retrospective review was developed
by in-house IT support, which tracks and quantifies quality scores and comments per individual technologist. A competency check-
list has been developed and is reviewed quarterly by the team to: (1) evaluate equipment performance (2) identify technical
problems with our imaging systems (3) review technologists' and physicists' mandated QC outcomes (4) track reject rates. This
comprehensive team approach has enabled the lead technologist and radiologist to (1) oversee mammography QC process in a
timely manner (2) review technologists' performance and offer positive and constructive feedback (3) encourage technologists to
take ownership of their patients (4) to address issues identified through these meetings sooner than later.
 BR114-ED-X

Cone Beam CT: A Promising Approach to Non-Oncologic Imaging of the Surgically-Altered Breast
All Day Room: NA Digital Education Exhibit

Participants
Alison J. Matich, MD, Rochester, NY (Presenter) Nothing to Disclose
Avice M. O'Connell, MD, Rochester, NY (Abstract Co-Author) Nothing to Disclose
Paige Myers, Rochester, NY (Abstract Co-Author) Nothing to Disclose
Megan Pencek, MD, Rochester, NY (Abstract Co-Author) Nothing to Disclose
Howard Langstein, Rochester, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

alison_matich@urmc.rochester.edu
TEACHING POINTS
To understand the indications, advantages, and limitations of Cone Beam CT in post-operative imaging evaluation of the non-
oncologic post-operative breast.
TABLE OF CONTENTS/OUTLINE
A. Introduction to Cone Beam CT technologyB. Indications for imaging in the non-oncologic, post-operative setting * Reduction
mammoplasty * Breast augmentationC. Advantages & limitations from the patient's perspectiveD. Advantages & limitations from the
plastic surgeon's perspectiveE. Possibilities for the future * Fat grafting
 BR115-ED-X

Diagnostic Breast Imaging and Interventions in the Age of Digital Breast Tomosynthesis: How We Do It
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Amy Maduram, MD , San Diego, CA (Abstract Co-Author) Nothing to Disclose
Mohammad Eghtedari, MD, PhD, La Jolla, CA (Abstract Co-Author) Nothing to Disclose
Rebecca Rakow-Penner, MD, PhD, San Diego, CA (Abstract Co-Author) Research Grant, General Electric Company
Vivian Lim, MD, La Jolla, CA (Abstract Co-Author) Consultant, CureMetrix, Inc
Haydee Ojeda-Fournier, MD, La Jolla, CA (Presenter) Research Consultant, View Point
For information about this presentation, contact:
hojeda@ucsd.edu
TEACHING POINTS
A quasi-three dimensional technology, digital breast tomosynthesis (DBT) has become widely available because of multiple studies
showing increased cancer detection rates and lower call back rates in the screening setting. While most practices start their DBT
program with screening, soon the benefits of DBT in diagnostic evaluation and guiding interventions become evident. At the end of
this educational exhibit the learner will: 1. Understand the importance of DBT as a unique imaging modality, 2. Learn to localize
breast lesions with DBT, 3. Become familiar with the imaging appearance of malignancy by BDT, and 4. Present an algorithm for
diagnostic evaluation based on DBT.
TABLE OF CONTENTS/OUTLINE
Introduction; Historical perspective; DBT basics; Localization: Skin, Quadrants, O'clock positions; DBT evaluation of: masses,
asymmetries, architectural distortions, and calcifications; DBT as supplemental imaging; DBT assessment of extent of disease;
Workflow simplification by addition of DBT; Algorithm for diagnostic evaluation based on DBT; Conclusion; Test yourself with image
and multiple choice questions.
 BR116-ED-X

Breast Cancer in Younger Women: Screening and Diagnostic Imaging Examinations

All Day Room: NA Digital Education Exhibit

FDA Discussions may include off-label uses.

Participants
Takayoshi Uematsu, MD, PhD, Nagaizumi, Japan (Presenter) Nothing to Disclose
For information about this presentation, contact:
t.uematsu@scchr.jp
TEACHING POINTS
Breast cancer in younger women aged < 40 years represents a challenge to public health due to the considerable number of years
of life lost, difficult early detection, and often late diagnosis. We will: 1. Propose screening algorism using multimodality breast
imaging based on breast cancer risk stratification including breast cancer genetics and mammographic breast density. 2. Discuss
most effective usage of multimodality breast imaging to faster and accurate diagnosis of breast cancer in younger women with less
costly care.
TABLE OF CONTENTS/OUTLINE
1. Introduction 2. Epidemiology of breast cancer in younger women 3. Risk factors of breast cancer in younger women 4. Screening
algorism for breast cancer in younger women 5. Multimodality breast imaging for breast cancer in younger women 6. Summary
 BR117-ED-X

Breast Diseases Arising in Patient of Chronic Renal Failure or Kidney Transplantation: What Radiologists
Should Know
All Day Room: NA Digital Education Exhibit

Participants
Eunji Lee, MD , Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Yun Woo Chang, MD, PhD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Jinah Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa jin Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jiyoung Hwang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Seong Sook Hong, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ywchang@schmc.ac.kr
TEACHING POINTS
1. To review the various breast diseases arising in patients with chronic renal failure (CRF)/ kidney transplantation (KT) and to
explain the affecting factors of the breast lesions in these patients. 2. Although breast screening is not routinely required in CRF/KT
patients, knowlegde of related disease spectrum for CRF/KT patients would be helpful for differential diagnosis and decision to
undergo biopsy.
TABLE OF CONTENTS/OUTLINE
1. Cyclosporine related breast disease in patients with KT; benign and malignant 2. Breast disease related with CRF with
hemodialysis; collateral vessel engorgement, diffuse breast enlargement, post-traumatic hematoma or fat necrosis. 3. Diabetic
mellitus related mastopathy in CRF patients. 4. Screening imaging evaluation for patients with CRF/KT.
 BR118-ED-X

Encouraging Mammography Screening: Strategies for Designing a Culturally and Linguistically Targeted Breast
Cancer Educational Program for a Multicultural Population
All Day Room: NA Digital Education Exhibit

Participants
Jenny Fung, BA, New York, NY (Abstract Co-Author) Nothing to Disclose
Alicia Li, BA, New York, NY (Abstract Co-Author) Nothing to Disclose
Lina Jandorf, New York, NY (Abstract Co-Author) Nothing to Disclose
Laurie R. Margolies, MD, New York, NY (Presenter) Research Consultant, FUJIFILM Holdings Corporation
For information about this presentation, contact:
laurie.margolies@mountsinai.org
TEACHING POINTS

Though advancements have been made in decreasing breast cancer incidence and mortality rates, minority groups still experience
significant disparities in breast cancer-related health outcomes. This has been partly attributed to low mammography screening
rates. These populations face group-specific barriers to screening related to low health literacy, limited English proficiency, and
cultural beliefs. Our intervention involved designing culturally and linguistically targeted breast cancer educational presentations for
African American women, African-Born women, Chinese women, Latina women, LGBTQ individuals, and Muslim women. Preliminary
findings demonstrate high levels of participant engagement and positive response across all target cultural groups. We believe that
this intervention has the potential to become a model for effective educational outreach to minority patients.
TABLE OF CONTENTS/OUTLINE
1. Review breast cancer statistics and screening rates among different minority groups 2. Identify barriers to screening for different
minority groups 3. Describe systematic approach to creating a culturally and linguistically targeted presentation 4. Illustrate
examples of addressing common misconceptions regarding breast cancer and screening through a narrative-based presentation
 BR119-ED-X

How to Identify Women Who Are Above Average Risk for Developing Breast Cancer: Retrospective Review of
our Institution's First Year Experience After Implementing a Breast Cancer Risk Assessment Program in Breast
Imaging
All Day Room: NA Digital Education Exhibit

Participants
Jessica L. Fournier, MD, Burlington, MA (Presenter) Nothing to Disclose
Cathleen M. Kim, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Jeanette Y. Chun, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Audrey L. Hartman, MD, MS, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Meaghan Mackesy, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Meera Sekar, MD, Lexington, MA (Abstract Co-Author) Nothing to Disclose
Michelle R. McSweeney, DO, Burlington, MA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
Review who is considered high risk for developing breast cancer and recommended screening guidelines. Review high risk screening
breast MRI data. Review our implementation and workflow of a breast cancer risk assessment program performed in our breast
imaging centers. Review our institution's first year results.
TABLE OF CONTENTS/OUTLINE
Women at above average risk for developing breast cancer benefit from earlier mammographic screening in addition to more robust
supplemental screening with MRI. However, most high risk women and their providers are unaware of their elevated risk. In 2017,
our institution implemented a breast cancer risk assessment program starting in breast imaging. All breast imaging patients were
offered breast cancer risk assessment using a risk assessment software, CRA Health, which used Tyrer-Cuzick and BRCAPRO
models. Over a 1 year period, 88% of patients agreed to have risk assessment. 14% of patients were calculated to be above
average risk. 9% of above average risk patients underwent MRI. 3% of these patients were recommended for biopsy. All biopsied
pathology revealed no cancers. 1. Above Average Risk Review. 1a. Current ACR Above Average Risk Screening Guidelines. 2. Breast
Cancer Risk Assessment Program 2a. CRA Health Risk Assessment Software Tool. 2b. Workflow 3. Single Institutional Retrospective
Review. 3a. Results
 BR120-ED-X

Diagnostic Imaging for Accelerated Partial Breast Irradiation

All Day Room: NA Digital Education Exhibit

Participants
Kazunori Kubota, MD, PhD, Bunkyo-Ku, Japan (Presenter) Nothing to Disclose
Tomoyuki Fujioka, MD, PhD, Bunkyo, Japan (Abstract Co-Author) Nothing to Disclose
Ukihide Tateishi, MD,PhD, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Mio Mori, Nagasaki, Japan (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kbtmrad@tmd.ac.jp
TEACHING POINTS
Accelerated Partial Breast Irradiation (APBI) is becoming a treatment for breast conservation therapy in early breast cancer, and
radiologists also need to know about the indication and the procedure with diagnostic imaging. The objective of this exhibit is 1. To
understand the therapeutic indications and to determine MRI diagnosis of breast cancer suitable for APBI. 2. To know the
postoperative condition of the breast and the method to insert the applicator under the ultrasound image guide.
TABLE OF CONTENTS/OUTLINE
1. Preoperative diagnostic imaging and indications for APBI - Early stage breast cancer, negative surgical margin, negative lymph
node metastasis - Preoperative MRI diagnosis - Preoperative ultrasonography for mapping of breast cancer 2. APBI insertion method
and early postoperative diagnostic imaging - The state of breast soon after the surgery with ultrasound - Where is the true cavity?
- Where to puncture and which way to insert the applicator (from inframammary fold via retromammary space) Preoperative
diagnosis, interventional technique for inserting the applicator, and radiation therapy can be implemented at the radiology
department. In this educational exhibit, We introduce diagnostic images and procedures of the istrut-adjusted volume implant
(SAVI), which is APBI using a sealed brachytherapy source.
 BR121-ED-X

Extent of Breast Cancer: Assessment with Sonography (US), Realtime Elastography (EG), and Magnetic
Resonance Imaging (MRI)
All Day Room: NA Digital Education Exhibit

Participants
Orlando Catalano, MD, Napoli, Italy (Abstract Co-Author) Nothing to Disclose
Antonio Nunziata, MD, Ercolano, Italy (Presenter) Nothing to Disclose
Roberta Fusco, Naples, Italy (Abstract Co-Author) Nothing to Disclose
Salvatore Filice, Naples, Italy (Abstract Co-Author) Nothing to Disclose
Antonella Petrillo, MD, Naples, Italy (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
orlandcat@tin.it
TEACHING POINTS
US has a tendency to underestimate breast cancer size. In this singlecenter, prospective study we evaluate the potential
additional value of EG in measuring lesions size.
TABLE OF CONTENTS/OUTLINE
We performed US, EG, and MRI within 30 days before surgery. The maximal diameter was measured by two blinded radiologists for
each modality as well as on the fresh specimen. Inclusion criteria were patients candidate to surgery, BIRADS 5 and 6 lesions,
absence of contraindication to MRI, insite surgery. Exclusion criteria were previous neoadjuvant therapy, postsurgical recurrence,
lesions >5 cm, T4 lesions, multicentric lesions, insufficient images quality. We enrolled 19 patients with 23 malignant lesions (15
infiltrating ductal carcinomas, 7 infiltrating lobular carcinomas, 1 insitu ductal carcinoma). MRI made the lowest error in terms of
mean and standard deviation. Pearson coefficient was 0.893 for US, 0,886 for EG, 0.998 for MRI. MannWitney test ranked 0.3674
between US and MRI, 0.5309 between US and EG, 0.8432 between EG and MRI. Using Kruskall Wallis test to assess the mismatch
between the three modalities the p value resulted <0.0001. The Intraclass Correlation Coefficient between the three modalities was
0.89. Both US and EG underestimate breast lesions. In selected cases, merging the two modalities may improve the final
measurement.
 BR122-ED-X

Male Breast Masses: A Multimodality Pictorial Review with Radiopathologic Correlation

All Day Room: NA Digital Education Exhibit

Participants
Cristina J. Quintero, MD , Bryn Mawr, PA (Presenter) Nothing to Disclose
Genghis Sanchez, MD, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Mohammad S. Hussain, DO, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Kristin Alman, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Maria Montano, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Jiao Wang, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Ronald A. Pitt, MD,MSc, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Anahid Pahlawanian, DO, Bryn Mawr, PA (Abstract Co-Author) Nothing to Disclose
Alicia L. Picard, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
quinteroc@mlhs.org
TEACHING POINTS
The male breast is affected by the same variety of pathologies affecting female breasts. Gynecomastia is the most common male
breast disorder, however, malignant breast tumors can be encountered and these usually present at a more advanced stage of
disease than women due to a delay in diagnosis. In this educational exhibit the learner will review: Anatomy and physiology of the
male breast with a focus on the differences compared to female breastEpidemiology and risk factors of male breast masses ACR
Appropriateness Criteria for imaging the male breast Imaging features of male breast masses on a case-based format with
pathologic correlationManagement of male breast pathologyA proposed diagnostic algorithm for evaluating breast male masses
TABLE OF CONTENTS/OUTLINE
Review of male breast anatomy Epidemiology, demographics and risk factors of male breast masses with a focus on male breast
cancer ACR Appropriateness Criteria for imaging male breast Case-based radiologic and pathologic correlation of benign and
malignant etiologies such as but not limited to gynecomastia (and its different types), fat necrosis, granulomatous inflammation,
myofibroblastoma, primary breast cancer, lymphoma, metastatic disease and axillary lymphadenopathyReview of work-up,
treatment, and follow-up
 BR123-ED-X

Granulomas in the Breast: Do I Need to Worry?

All Day Room: NA Digital Education Exhibit

Participants
Eileen Delaney, MD, Worcester, MA (Presenter) Nothing to Disclose
Farhana Sharmeen, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Vincent G. Champion, MD, Westwood, MA (Abstract Co-Author) Nothing to Disclose
Jordana Phillips, MD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Consultant, General Electric
Company
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
fsharmee@bidmc.harvard.edu
TEACHING POINTS
1. Granulomatous lesions of the breast are rare and represent a broad array of benign diseases, including sarcoidosis, mammary
duct ectasia, periductal mastitis, infection, granulomatous mastitis, and foreign body reaction. 2. Histopathology is key in
diagnosing granulomatous lesions of the breast as many of these entities mimic malignancy on imaging. 3. The treatment and
management of granulomatous lesions of the breast vary depending on the etiology.
TABLE OF CONTENTS/OUTLINE
1. Overview of granulomatous disease and key histologic features. 2. Granulomatous Mastitis a. Types of Granulomatous Mastitis i.
Idiopathic ii. Cystic Neutrophilic Mastitis b. Clinical and imaging findings c. Management 3. Reaction to Foreign Bodies/Trauma a.
Types of Foreign Bodies i. Suture Granuloma ii. Silicone Granuloma iii. Fat Necrosis b. Clinical and Imaging Findings c. Management 4.
Mammary Duct Ectasia and Chronic PeriductalMastitis a. Clinical and Imaging Findings b. Management 5. Infectious etiologies a.
Types of Infectious Pathogens i. Mycobacterial ii. Fungal iii. Parasites b. Clinical and Imaging Findings c. Management 6. Sarcoidosis
a. Clinical and Imaging Findings b. Management
 BR124-ED-X

MRI-Detected Breast Lesions: Comparison with Second-look US Using Real-time Virtual Sonography and
Histopathological Findings
All Day Room: NA Digital Education Exhibit

Awards
Cum Laude
Participants
Kazuaki Nakashima, MD, Nagaizumi, Japan (Presenter) Nothing to Disclose
Takayoshi Uematsu, MD, PhD, Nagaizumi, Japan (Abstract Co-Author) Nothing to Disclose
Kaoru Takahashi, MD, Nagaizumi, Japan (Abstract Co-Author) Nothing to Disclose
Seiichirou Nishimura, MD, Nagaizumi, Japan (Abstract Co-Author) Nothing to Disclose
Takashi Sugino, MD, Nagaizumi, Japan (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
MRI-detected breast lesions are difficult to diagnose because other modalities are usually not helpful. Real-time virtual sonography
improves lesion detectability and enables comparison of the MRI and ultrasonographic findings. The aims of this exhibit are: To
understand ultrasonographic and histopathological findings of MRI-detected breast lesions. To discuss appropriate management of
MRI-detected lesions depending on each institution and patient background to reduce unnecessary biopsy.
TABLE OF CONTENTS/OUTLINE
1. Introduction 2. Frequency and histopathology of MRI-detected lesions 3. Further examinations Second-look ultrasonography
with/without real-time virtual sonography Biopsy Follow-up examinations 4. Illustrative cases with histopathologic findings 5. How
do we manage MRI-detected lesions? 6. Summary
 BR125-ED-X

Tomosynthesis and Its Benefits in the Diagnosis of Architectural Distortions: A Way to Increase Our
Diagnostic Capacity
All Day Room: NA Digital Education Exhibit

Participants
Flavia B. Sarquis, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Lucia I. Beccar Varela, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Soledad Nocetti, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Belen Pulido, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Paola Pucci, MD, Vicente Lopez, Argentina (Presenter) Nothing to Disclose
Rita Polo, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Bernardo O. Blejman, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Florencia Melendez, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
sarquisflavia@gmail.com
TEACHING POINTS
To define Architectural distortion (AD) of the breast according to the description of the American College of Radiology (ACR) in the
Breast Imaging Reporting and Data System (BI-RADS). To identify the most common benign and malignant etiologies that can be
seen with clinical presentation of AD. To examine the diagnostic utility of tomosynthesis in the identification of architectural
distortions on mammographic views. To discuss and compare the role of other imaging modalities To recommend practical tools for
appropriate management of tomosynthesis detected architectural distortions.
TABLE OF CONTENTS/OUTLINE
Introduction. Definition of AD. What is Tomosynthesis and how does it work? Brief review of the principles of DBT advantages of
Digital breast tomosynthesis in improving architectural distortion visibility and characterization. Understanding the addition of 3D
Mammography to current practice. Implementation of tomosynthesis-guided breast biopsy. Case-based illustration. Conclusion and
take-home points.
 BR126-ED-X

A Multi-Modality, Case-Based Review of the 8th Edition AJCC Breast Cancer Staging and Prognosis Algorithm:
The Why and How It Impacts Staging
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit
Participants
Katrina Korhonen, MD , Philadelphia, PA (Presenter) Nothing to Disclose
Austin Pantel, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. McDonald, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Amy Clark, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
David A. Mankoff, MD, PhD, Philadelphia, PA (Abstract Co-Author) Speaker, Koninklijke Philips NV; Consultant, General Electric
Company; Advisory Board, RefleXion Medical Inc; Consultant, Blue Earth Diagnostics Ltd; Research Funded, Siemens AG; Advisory
Board, ImaginAb, Inc; Spouse, Owner, Trevarx
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
For information about this presentation, contact:
katrina.korhonen@uphs.upenn.edu
TEACHING POINTS
The 8th edition AJCC breast cancer staging guidelines modify the TNM anatomic staging model with the incorporation of prognostic
biomarkers such as tumor grade, hormone and HER2-receptor status, and gene expression panels. These new biomarkers provide
prognostic information, guide patient treatment selection, and predict patient outcomes to therapy. Newly incorporated biomarkers
such as Oncotype Dx may result in some patients with estrogen-receptor positive, lymph-node-negative cancers being downstaged
to stage IA. 18F-FDG-PET/CT plays a valuable role in patients with clinical stage IIB or higher cancer by identifying extra-axillary
lymph node metastases and unsuspected distant metastatic disease.
TABLE OF CONTENTS/OUTLINE
1. Summary of salient changes to TNM categories 2. Overview of new prognostic biomarkers incorporated into the updated
guidelines, including tumor grade, hormone and HER2 expression status, and Oncotype Dx recurrence score 3. Multi-modality case-
based review underscoring the importance of radiology in the new staging guidelines a. Case examples highlighting the role of
imaging, including both conventional breast imaging modalities and FDG-PET/CT, in changing management b. Cases highlighting how
the inclusion of new biomarkers may result in a prognostic stage that differs from the anatomic stage
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ David A. Mankoff, MD, PhD - 2013 Honored EducatorDavid A. Mankoff, MD, PhD -
2018 Honored Educator
 BR127-ED-X

Developing Asymmetry: Learning Algorithm

All Day Room: NA Digital Education Exhibit

Participants
Yesenia Bermudez Cano, MD, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Veronica Gonzalez Bascon, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Carolina Hadad, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Silvina Cadullo, Bueno Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Virginia Secco, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
yesebeca@hotmail.com
TEACHING POINTS
1. Review and incorporate the concept of developing asymmetry in mammographic reports. 2. In the search for carcinoma hidden in
a developing asymmetry, highlight the importance of comparing with previous studies. 3- Very useful learning algorithm in case of
developing asymmetry, as a guide to know the next step to follow when this finding is presented
TABLE OF CONTENTS/OUTLINE
Table of Contents: 1. Introduction 2. Objectives 3. Definition of developing Asymmetry 4. Learning algorithm: Additional projections
- Ultrasound as a complementary diagnostic study - Biopsy 5. Causes of developing asymmetry (benign or malignant): radio-
pathologic correlation. 6. Recommendations 8. Conclusion.
 BR128-ED-X

What Can It Be Except Gynecomastia?: Imaging Findings of the Common and Uncommon Causes of the
Breast Lump in Male Patients
All Day Room: NA Digital Education Exhibit

FDA Discussions may include off-label uses.

Participants
Gamze Durhan, MD, Ankara, Turkey (Abstract Co-Author) Nothing to Disclose
Omer Onder, Ankara, Turkey (Presenter) Nothing to Disclose
Aynur Azizova, Ankara, Turkey (Abstract Co-Author) Nothing to Disclose
Funda Elibol, Mu?la, Turkey (Abstract Co-Author) Nothing to Disclose
Meltem G. Akp?nar, Ankara, Turkey (Abstract Co-Author) Nothing to Disclose
Figen B. Demirkazik, MD, Ankara, Turkey (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
gamzedurhan@gmail.com
TEACHING POINTS
1. Familiarity with the imaging findings of various benign and malignant male breast lesions allows correct imaging interpretation
avoiding unnecessary biopsies and treatment. 2. Majority of breast lumps in male patients are benign and the most common reason
is gynecomastia. Three mammographic patterns of gynecomastia are nodular, dendritic and diffuse glandular forms. Except
gynecomastia, other common and uncommon benign conditions of the male breast are pseudogynecomastia, lipoma, ductal
hyperplasia, cyst, chronic inflammation, ductal ectasia, intraductal papilloma, myofibroblastoma and pilomatricoma. 3. The incidence
of breast cancer is very low in male patients. The most common histological type of primary breast cancer is ductal carcinoma.
They are usually detected as spiculated, lobulated, microlobulated masses or ill-defined hypoechoic lesions. Microcalcifications
unlike in female patients are rarely seen. Other malignant causes of breast lump in male patients include metastases to the breast
and breast involvement by the hematologic malignancies.
TABLE OF CONTENTS/OUTLINE
Epidemiology and overview of the male breast tissue and diseases. Radiologic investigation of the common and uncommon benign
breast lesions in male patients. Imaging features of the primary breast cancer and metastatic tumors to the breast in male patients.
Summary.
 BR129-ED-X

A Standardized Approach and Lexicon for Effective Reporting of Breast Cancer MRIs Pre and Post-
Neoadjuvant Chemotherapy
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit
Participants
Jenny Qian, MD, New York, NY (Presenter) Nothing to Disclose
Shreena Shah, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Natsuko Onishi, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth J. Sutton, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
shb9167@nyp.org
TEACHING POINTS
1. Understand the importance and clinical impact of findings on breast magnetic resonance imaging (MRI) pre and post-neoadjuvant
chemotherapy 2. Know that variations in MRI interpretation exist due to the lack of a standardized reporting approach and lexicon
3. Recognize important imaging findings that should be included in the report that can impact a patient's treatment plan 4. Provide
terminology that effectively communicates treatment response
TABLE OF CONTENTS/OUTLINE
- Sample cases illustrating typical MRI findings pre- and post-neoadjuvant therapy - Standardized approach and lexicon to use as a
template when interpreting MRI in the neoadjuvant setting, including: a. Reproducible measurement of index tumor and extent in
two dimensions b. Methods of describing multifocal/multicentric disease c. Presence or absence of pertinent findings that impact
management (i.e. involvement of the nipple areolar complex) d. Reporting of axillary and internal mammary lymph nodes e. Lexicon
for change in tumor and lymph nodes post-treatment f. Standardized terms to communicate residual enhancement in the tumor bed
g. Concise impression that accurately reflects body of the report without repeating it, including a statement indicating presence or
absence of a complete imaging response post-neoadjuvant chemotherapy h. Appropriate BI-RADS assessment post-treatment
 BR130-ED-X

Breast Cancer Screening in Transgender Patients

All Day Room: NA Digital Education Exhibit

Participants
Divya N. Chowdhry, MD, Rochester, NY (Presenter) Nothing to Disclose
Avice M. O'Connell, MD, Rochester, NY (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
In the United States, 1.4 million adults identify as transgender. Despite growing national awareness, the transgender population
faces barriers to healthcare. Also, due to the infrequency of transgender breast cancer cases, there is a lack of evidence-based
screening guidelines. Several gender-affirming hormonal and surgical interventions are available which impact appearance on imaging
and risk of breast cancer. Familiarity with expected imaging findings in the transgender population can help avoid medically-induced
physiologic changes being mistaken for pathology. Breast cancer risk is thought to increase with estrogen, family history, BRCA
mutation, radiation, and BMI >35. Mastectomy significantly reduces but does not eliminate risk of breast cancer.
TABLE OF CONTENTS/OUTLINE
(1) Clarification of transgender terminology. (2) Discussion of challenges to delivering sensitive and quality healthcare to
transgender patients. (3) Description of hormonal therapies and surgical options for gender-affirmation. (4) Review of cases
demonstrating common mammographic and ultrasound findings in this population. (5) Summarization of breast cancer incidence and
risk factors in transgender patients. (6) Presentation of existing screening guidelines while highlighting areas for further research.
 BR131-ED-X

Contrast-Enhanced Spectral Mammography (CESM): Lessons Learned from Early Clinical Implementation
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Robyn G. Roth, MD, Cherry Hill, NJ (Presenter) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Allison S. Gittens, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Pauline Germaine, DO, Camden, NJ (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
roth-robyn@cooperhealth.edu
TEACHING POINTS
Contrast-enhanced Spectral Mammography (CESM) has been in clinical use at our institution since November 2012. We have
performed over 2,500 CESM cases, with over 250 cases of biopsy-proven malignancy. This educational exhibit will illustrate various
clinical applications of CESM in both the screening and diagnostic settings. The lessons learned in the early clinical implementation
of CESM will be highlighted with multimodality imaging and pathologic correlation.
TABLE OF CONTENTS/OUTLINE
Review technique, indications, and clinical applications of CESM Demonstrate how we use CESM in clinical practice in the screening
and diagnostic setting Show normal and abnormal cases, with MRI and pathology correlation Discuss limitations and CESM-specific
artifacts
 BR132-ED-X

CESM as a Tool for Staging and Post-Neoadjuvant Chemotherapy Treatment Response

All Day Room: NA Digital Education Exhibit

Participants
Maria C. Sciotto, MD, Camden, NJ (Presenter) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Robyn G. Roth, MD, Cherry Hill, NJ (Abstract Co-Author) Nothing to Disclose
Miriam Enriquez, MD, Camden, NJ (Abstract Co-Author) Nothing to Disclose
Pauline Germaine, DO, Camden, NJ (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

sciotto-maria@cooperhealth.edu
TEACHING POINTS
1. CESM provides improved lesion detection and characterization by combining digital mammography with the added assessment of
lesion vascularity following contrast administration, thus introducing a physiologic component into diagnostic imaging. 2. With these
benefits in mind, we have performed CESM for staging of known breast cancer and to assess response to neoadjuvant
chemotherapy based on residual tumor enhancement in a small subset of patients unable to undergo breast MRI. 3. The goal of
neoadjuvant chemotherapy is to achieve complete pathologic response. CESM is a quick and effective way of guiding these clinical
decisions.
TABLE OF CONTENTS/OUTLINE
1. Introduction to CESM What is CESM? CESM technique and protocol 2. Indications for CESM High risk screening, diagnostic work
up, staging, restaging post neoadjuvant chemotherapy 3. Advantages of CESM 4. Limitations of CESM 5. When CESM can be used
for staging over MRI 6. Pictorial depiction of CESM use for assessing response to neoadjuvant chemotherapy CESM initial staging,
CESM post neoadjuvant chemotherapy staging and final pathology at excision
 BR133-ED-X

The Art of Diagnosing

All Day Room: NA Digital Education Exhibit

Participants
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
For information about this presentation, contact:
drakarina.pesce@gmail.com
TEACHING POINTS
1. Develop skills of observation and interpretation 2. Describe how artists portray manifestation of pathology breast 3. Transfer the
skills learnt from appreciation of artworks to patient care
TABLE OF CONTENTS/OUTLINE
In this review of masterpieces, we will propose hypothetical diagnoses according to the findings observed in the anatomy of the
breasts and armpits, with a speculative approach on how the mammographic study of the muses would be observed had they been
assisted by today's technology. Combining pictorial art with the mammographic diagnosis of breast pathology can be used as an
attractive tool for teaching the specialty.
 BR134-ED-X

Breast Cancer In Patients Under The Age Of Forty: Imaging, Assessment, and Management
All Day Room: NA Digital Education Exhibit

Participants
Claire H. Meriwether, La Jolla, CA (Abstract Co-Author) Nothing to Disclose
Rebecca Rakow-Penner, MD, PhD, San Diego, CA (Abstract Co-Author) Research Grant, General Electric Company
Mohammad Eghtedari, MD, PhD, La Jolla, CA (Abstract Co-Author) Nothing to Disclose
Vivian Lim, MD, La Jolla, CA (Abstract Co-Author) Consultant, CureMetrix, Inc
Alaa Alshamrani, MD, Dhahran, Saudi Arabia (Abstract Co-Author) Nothing to Disclose
Haydee Ojeda-Fournier, MD, La Jolla, CA (Presenter) Research Consultant, View Point
For information about this presentation, contact:
hojeda@ucsd.edu
TEACHING POINTS
Women under the age of 40 constitute 3% of in situ breast cancer, 4% of invasive breast cancer and 2% of breast cancer related
deaths. While in women over the age of 40 cancers are mostly detected by mammographic screening, in younger women breast
cancers present with clinical symptoms more often than are detected by high-risk screening. The radiologist must have an
appropriate level of suspicion while evaluating young women so as not to miss breast cancer. At the end of this educational exhibit
the learner will: 1. Know the incidence and types of breast cancers seen in women under age 40, 2. Learn the optimal imaging
modalities for evaluating young women with breast cancer, 3. Recognize the imaging appearance of breast cancers seen in young
women, and 4. Review the updated protocols for high risk screening in young women.
TABLE OF CONTENTS/OUTLINE
Introduction; Facts and figures of breast cancer in the patient under age 40; Breast cancer profile in the young women; The triple
negative breast cancer; Pregnancy associated breast cancer; High-risk screening recommendations for young women; Multimodality
breast imaging in women under 20, 20-29, and 30-40 year of age; The clinical management of breast cancer in young women;
Reproductive considerations; Conclusion.
 BR135-ED-X

Incidental FDG Uptake in the Breast: A "Hot" Topic

All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Ronald W. Mercer, MD, Boston, MA (Presenter) Nothing to Disclose
Kevin J. Donohoe, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Valerie J. Fein-Zachary, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Parisa Lotfi, MD, Newton, MA (Abstract Co-Author) Nothing to Disclose
Vandana M. Dialani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Evguenia J. Karimova, MD, Memphis, TN (Abstract Co-Author) Nothing to Disclose
Michael D. Fishman, MD, Boston, MA (Abstract Co-Author) Consultant, Zebra Medical Vision Ltd
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rmercer@bidmc.harvard.edu
TEACHING POINTS
1. 18F-fluorodeoxyglucose (FDG ) PET/CT continues to increase in frequency to stage and monitor most types of malignancy. 2.
Incidental foci of 18F-fluorodeoxyglucose uptake in the breasts on PET/CT occurs in approximately 0.8-6.3% of cases, with 37.5-
88.3% of these representing malignancy. 3. Broad differential diagnosis exists for PET positive findings, including malignancy
(primary breast carcinoma, metastasis, and lymphoma) as well as benign entities (fibroadenoma, post-partum lactation, fibrocystic
change, papilloma, infection, fat necrosis, gynecomastia, or normal breast tissue). 4. Given relatively high rate of malignancy in
patients with incidental FDG-avid breast lesions, diagnostic evaluation and, if appropriate, tissue diagnosis is warranted. More
controversial factors to consider include SUVmax, patient age, and lesion size.
TABLE OF CONTENTS/OUTLINE
1. Discuss the incidence, significance, and work-up of incidental breast lesions identified on PET-CT performed for reasons other
than known breast malignancy. 2. Develop a differential for these incidental findings incorporating clinical data into the management
decision. 3. Understand the limitations of current imaging modalities around management of incidental foci of FDG uptake in the
breast.
 BR136-ED-X

The Value of Adding Contrast Enhanced Spectral Mammography to the Diagnostic Breast Imaging Practice:
How to Implement a Program and Examples of the Utility of CESM as a Problem Solving Tool
All Day Room: NA Digital Education Exhibit

Participants
Caitlin Ward, MD, MSc, London, ON (Presenter) Nothing to Disclose
Anat Kornecki, MD, London, ON (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Caitlin1.ward@lhsc.on.ca
TEACHING POINTS
1. To discuss role of contrast enhanced mammography in current day breast imaging and scenarios where contrast enhanced
mammography can be offered2. To discuss the benefits to patient care seen with the use of contrast enhanced mammography in
routine practice3. To identify the steps required to implement a contrast enhanced mammography program
TABLE OF CONTENTS/OUTLINE
1. Background regarding development of contrast enhanced mammography2. Review of image acquisition and physics including
radiation dose to patients3. Discussion of indications for contrast enhanced mammography4. Discussion of value of contrast
enhanced mammography to improvement in patient care5. Steps to starting a contrast enhanced mammography program, including:
a. Equipment/ software requirements b. Personnel training c. Scheduling/protocol development d. Patient safety6. Review of cases
to illustrate early examples from our institution and the variety of findings and uses including: a. Mammographically occult cancer b.
Multicentric cancer c. Phylloides d. Incidental contralateral breast cancer e. Mastitis f. Benign masses and causes of asymmetry
 BR137-ED-X

What Radiologists Should Know About Artifacts in Breast MRI

All Day Room: NA Digital Education Exhibit

Participants
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
maria.chico@hospitalitaliano.org.ar
TEACHING POINTS
1. Analyze the types of breast magnetic resonance imaging artifacts 2. Discuss the origin and appearance of the most common
artifacts encountered in breast MRI along with possible correcting methods to avoid or reduce them
TABLE OF CONTENTS/OUTLINE
1. Introduction: A wide variety of artifacts can be seen in clinical breast MR imaging. Although the elimination of some artifacts may
require a service engineer, the radiologist and MR technologist have the responsibility to recognize MR imaging problems and the
way to solve them in daily practice. 2. Techniques in breast MRI. 3. Types of artifacts A)Machine hardware- and software- related
artifacts B)Patient-related artifacts 4. Illustration of different types of artifacts. 5. Tips and tricks to avoid them 6. Conclusion
 BR138-ED-X

Out With the Old... Let Go with the New BI-RADS MRI Guide for the Resident
All Day Room: NA Digital Education Exhibit

Participants
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Maria B. Orruma, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Carolina Hadad, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Maria P. Swiecicki, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

maria.chico@hospitalitaliano.org.ar
TEACHING POINTS
1. Describe breast lesions by using proper BI-RADS descriptors. 2. Determine the appropriate BI-RADS category for the lesion of
interest. 3. Learn through pictures, diagrams, with interactive cases of normal and abnormal commonly encountered findings in
everyday practice.
TABLE OF CONTENTS/OUTLINE
1. Introduction: The ACR BI-RADS structure provides a systematic and efficient method to evaluate MRI findings and facilitate
resident and breast imaging fellowship training. Both the ACR and the Society of Breast Imaging recommend that breast imaging
education within residency and fellowship training should be designed to require the use of BI-RADS terminology, assessment
categories, and management recommendations. Educating radiologists on the use of BI-RADS is an effective way to improve their
interpretive skills in MRI breast exams. 2. History of BI-RADS 3. Lexico 4. Description of the variables studied in breast MRI. 5.
Interpretation of MRI findings 6. Assessment Categories 7. Self-testing section in multiple choice form at 8. Conclusion: In this
educational exibit of the changes and new descriptors in the MRI section of the fifth edition of BI-RADS with pictorial examples, the
resident would be able to achieve improved understanding of the MRI BI-RADS lexicon and its appropriate applications.
 BR139-ED-X

False Negative Breast Lesions - What They Can Teach

All Day Room: NA Digital Education Exhibit

Participants
Eduardo Kaiser Ururahy Nunes Fonseca, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Matheus Horie, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Milena R. Peixoto, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Flavio I. Shitara, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Larissa M. Yano, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Erica Federicci, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Renato L. Ribeiro, MD, San Paolo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
edukaiser.unf@gmail.com
TEACHING POINTS
The purpose of this exhibit is: -To present the findings of false negatives exams from our institution in both mammograms and
breast MRI. -To discuss the possible causes of mistakes using these cases as examples of a teachable moment and how to learn
from them to prevent similar errors in the future. -To briefly discuss the importance of medical audit in a Breast Imaging Center and
how to learn from these cases;
TABLE OF CONTENTS/OUTLINE
We will present selected different cases of false negative mammograms and breast MRI from our teaching file, from 2011-2018. Key
points and possible causes for these false negative results will be discussed in each case, as well as how to prevent error
reoccurrence; The list of cases includes: 1. False negative Mammogram (the lesion was there and was not visible even in
retrospect) 2. Mammogram missed lesions (the lesion was there and was not detected but was visible in retrospect) 3.
Misinterpreted Mammogram lesions (the lesions was detected, but misinterpreted) 4. False negative MRI (the lesion was there and
was not visible even in retrospect) 5. MRI missed lesions (the lesion was there and was not detected but was visible in retrospect)
6. Misinterpreted MRI lesions (the lesions was detected, but misinterpreted)
 BR140-ED-X

Looking Beyond: Extramammary Findings in Breast MRI

All Day Room: NA Digital Education Exhibit

Participants
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Carolina Hadad, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Roy Lopez Grove, MD, Temperley, Argentina (Abstract Co-Author) Nothing to Disclose
Maria B. Orruma, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

maria.chico@hospitalitaliano.org.ar
TEACHING POINTS
1. Analyze the clinical relevance of these unexpected findings. 2. Describe the sites and the nature of the extramammary findings in
magnetic resonance imaging. 3. Illustrate with examples the findings found, and if there were the methods used for correct
characterization.
TABLE OF CONTENTS/OUTLINE
1. Introduction: Breast MRI is typically confined to a small field of view (FOV) and it is mainly focused on the assessment of the
mammary gland, pectoralis muscles, skin and axillary lymph nodes. However, other structures are also partially visible, such as the
lungs, mediastinum, chest wall, spine, and upper abdomen. Some of these findings can sometimes be clinically significant and could
modify the patient's management, requiring more complementary examinations. 2. General description of breast MIR protocol 3.
Interpretation of MRI 4. Tips to optimize the handling of the relevant extramammary findings. 5-Conclusion
 BR141-ED-X

Breast Density Issues and Impacts on Mammography Screening

All Day Room: NA Digital Education Exhibit

FDA Discussions may include off-label uses.

Participants
Takayoshi Uematsu, MD, PhD, Nagaizumi, Japan (Presenter) Nothing to Disclose
For information about this presentation, contact:
t.uematsu@scchr.jp
TEACHING POINTS
Mammographic breast density issues are rapidly becoming a hot topic in the medical community and the media worldwide. It is
important for radiologists to understand the context, scientific evidence, and controversies. We will: 1. Review current practice
guidelines and new clinical implications improving radiologists' understanding of the relevant subject of breast density. 2. Enable
them to respond appropriately to questions from patients, clinicians, and the media.
TABLE OF CONTENTS/OUTLINE
1. Introduction 2. Clinical implications of breast density 3. How to assess mammographic breast density 4. Possible supplemental
breast cancer screening modality for women with dense breasts 5. How to deal with breast density issues 6. Summary
 BR142-ED-X

The Role of Diffusion Weighted MR Imaging in Breast Lesions

All Day Room: NA Digital Education Exhibit

FDA Discussions may include off-label uses.

Participants
Aitana Palomares, MD, Toledo, Spain (Presenter) Nothing to Disclose
Lina M. Cruz Hernandez, ARRT, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Pilar Sanchez Camacho, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Paul M. Aguilar Angulo, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Cristina Romero, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Celia Astor Rodriguez, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
aitana.yz@gmail.com
TEACHING POINTS
• To review the usefulness of Diffusion Weighted Magnetic Resonance Imaging (DW-MRI) in the assesment of breast lesions • To
illustrate benign and malignant breast lesions with different techniques: mammography, ecography, and MRI; focusing on the value
of Diffusion-WI to discriminate between both types of lesions • To describe potential uses of this technique and its limitations
TABLE OF CONTENTS/OUTLINE
1.- DW-MRI: Technique, concept of Brownian motion, concept of restricted diffusion and ADC value 2.- Imaging findings of breast
lesions in mammography, ecography and MRI: • Benign lesions: - Fibroadenoma - Tubular Adenoma - Radial Scar - Intraductal
Papilloma - Stromal Fibrosis - Benign Phyllodes Tumour - Fibroadenosis • Malignant lesions: - Ductal invasive carcinoma - Colloid
carcinoma - Lobular carcinoma - Ductal invasive carcinoma with fibrosis - We analyze the behaviour of these lesions in DW and ADC
maps and how this correlates with its nature (benign vs. malignant) 3.- Potential clinical applications: Evaluation of treatment
response and locoregional staging (lymph nodes)
 BR143-ED-X

Common and Uncommon Breast Lesions in Women Under 30: Clinical Presentation, Diagnosis, and
Management
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit
Participants
Haley R. Clark, MD, Dallas, TX (Presenter) Nothing to Disclose
Jody C. Hayes, MD, Southlake, TX (Abstract Co-Author) Nothing to Disclose
Kanwal A. Merchant, MD, Dallas, TX (Abstract Co-Author) Stockholder, Sensogram Technologies; Spouse, Stockholder, Sensogram
Technologies
Lena A. Omar, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Basak E. Dogan, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1. Women younger than 30 frequently present with symptoms of palpable abnormality, breast pain, or nipple discharge. 2. According
to ACR appropriateness criteria, first line imaging should be ultrasound (US), and if biopsy is warranted, US guided needle biopsy. If
biopsy reveals malignancy, mammography is warranted for presurgical planning. 3. Young women can present during pregnancy or
lactation period. Common findings are galactocele, enlarging fibroepithelial lesions, and less likely malignancy. Appropriate diagnosis
of abscess associated with puerperal mastitis is critical in order to appropriately route the patient to drainage. 4. Rare lesions
include granular cell tumor, sarcoma, and malignancy. 5. While malignancy is rare, triple-negative breast cancer can mimic a benign
lesion. All newly palpable solid lesions should therefore be evaluated with care.
TABLE OF CONTENTS/OUTLINE

I. Review of ACR appropriateness criteria for imaging women <30 II. Approach to solid, benign-appearing lesions: when to biopsy,
follow up or do nothing III. Infectious mastitis and abscess IV. Idiopathic granulomatous mastitis: clinical presentation and
constellation of pathognomonic findings. V. Breast cancer, and rare neoplasms
 BR144-ED-X

The Medical Student Experience: From Diagnosis to Treatment of Breast Cancer Improving the Medical
Student Rotation in the Breast Care Center
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit
Participants
Sarah B. Thomas, DO, Pittsburgh, PA (Presenter) Nothing to Disclose
Betty E. Shindel, MD, Mars, PA (Abstract Co-Author) Nothing to Disclose
Matthew S. Hartman, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Hamanpreet Bandesha, Blue Point, NY (Abstract Co-Author) Nothing to Disclose
Robin N. Sobolewski, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Sarah.Thomas@AHN.org
TEACHING POINTS
1. Improve the medical student mammography rotation experience through interactive targeted learning experiences.2. Focus of the
medical student rotation on the patient experience in the breast care center through the diagnosis and treatment process. 3. The
medical student follows various patients through a screening mammography, diagnostic mammography, biopsy and surgery.
TABLE OF CONTENTS/OUTLINE
1. Traditional medical student rotation.2. Purposes of improving the medical student rotation.3. Ways to make the experience in
women's imaging and the breast care center more interactive.- Structured experiences to distinguish between screening
mammography and diagnostic mammography.- Opportunity for Radiology-Pathology correlation.i. Observe ultrasound and
stereotactic core biopsy.ii. Observe needle localization for surgical planning.iii. Follow the patient from needle localization to the
Operating Room.iv. Review pathology with the pathologistv. Review the pathology and imaging and determine concordance.- Attend
breast surgery clinic and discuss surgical and treatment options with the Breast Surgeon.- Attend breast tumor board with the
Radiologist, Pathologist, Oncologist and Breast Surgeon to understand the multi-disciplinary approach.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Matthew S. Hartman, MD - 2016 Honored Educator
 BR145-ED-X

Utilization of Molecular Breast Imaging in Women with Breast Foreign Bodies

All Day Room: NA Digital Education Exhibit

Participants
Cameron Leitch, MD, Rochester, MN (Presenter) Nothing to Disclose
Katie N. Hunt, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Carrie B. Hruska, PhD, Rochester, MN (Abstract Co-Author) Institutional license agreement, CMR Naviscan Corporation
Amy Lynn Conners, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Dana H. Whaley, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

leitch.cameron@mayo.edu
TEACHING POINTS
Molecular breast imaging (MBI) is a promising tool for adjunctive screening to mammography in women with dense breast tissue.
However, questions arise regarding its utility in women with foreign bodies, such as silicone/saline implants, silicone/paraffin
injections, and pacemakers. The purpose of this educational exhibit is to demonstrate effectiveness of MBI in evaluation of breasts
with foreign bodies, including normal imaging features, variations in imagining appearance, as well as advantages and limitation of
MBI compared to mammography in this patient population. Knowledge regarding the varied appearance of implants and foreign
bodies on MBI is important for interpretation, and familiarity with the appearance of benign and malignant findings will allow
radiologists to more confidently and accurately interpret MBI.
TABLE OF CONTENTS/OUTLINE
Review the appearance of breast implants and other foreign bodies on MBI. Explain the varied appearance of foreign bodies on the
upper/lower detectors. Illustrate pathologic findings seen in patients with implants and foreign bodies in the breast. Demonstrate
the utility of supplemental screening with MBI for patient with silicone injections or implant ruptures.
 BR146-ED-X

Breast Invaders: Atypical Infectious and Parasitic Breast Diseases

All Day Room: NA Digital Education Exhibit

Awards
Magna Cum Laude

Participants
Heni D. Skaf, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Caio D. Pinheiro, MBBS, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Camila F. Corona, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Henrique Durante, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rafael L. Macedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Cecilia S. Goldman, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Erica Endo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Flavia S. Braga, MD, Vitoria, Brazil (Abstract Co-Author) Nothing to Disclose
Marco A. Costenaro, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Dafne C. Andrade, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Ruiz, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Isabelle V. Nisida, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1. Discuss the common presentations and challenges in diagnosing atypical infectious and parasitic diseases of the breast. 2.
Describe the imaging findings of some atypical infectious and parasitosis of the breast, such as paracoccidioidomycosis, filariasis,
cutaneous myiasis, cat-scratch disease and tuberculosis of the breast. 3. Describe the epidemiology, histopathological features and
peculiarities of each disease. 4. Establish a radiologic histopathologic correlation with the clinical aspects of each infection. 5.
Discuss the importance of the multimodality imaging findings in association with histopathological aspects to achieve the essential
diagnosis and proper treatment of each pathology.
TABLE OF CONTENTS/OUTLINE
1 - Common clinical presentations and imaging challenges in diagnosing atypical infections and parasitic breast diseases. 2 -
Specific imaging findings, epidemiology and histopathological features of: - paracoccidioidomycosis, - filariasis, - cutaneous myiasis,
- cat-scratch disease, - tuberculosis of the breast. 4 - Summary and Conclusion.
 BR147-ED-X

Second Look Breast Ultrasound: A Practical Approach

All Day Room: NA Digital Education Exhibit

Participants
Fleur Kilburn-Toppin, MBBChir, MA, Cambridge, United Kingdom (Presenter) Nothing to Disclose
Katy Hickman, MBBS, Cambridge , United Kingdom (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
fleur.kilburn-toppin@addenbrookes.nhs.uk
TEACHING POINTS
The purpose of this exhibit is:To understand indications for second look US after breast MRITo review tips and tricks on techniques
for second look US to maximise lesion detectionTo review the management of cases undergoing second look US
TABLE OF CONTENTS/OUTLINE
- When to do second look US -What am I likely to see on second look US-Techniques for second look US(subsections on breast
positioning, defining location based on MRI, lesion appearances on US based on MRI, use of anatomical landmarks and clips)-What
to do if lesion not found on second look US- Summary of key messages for second look US
 BR148-ED-X

Pathologic Nipple Discharge: Evaluation, Management, and Imaging Findings

All Day Room: NA Digital Education Exhibit

Participants
Alexandra M. Millet, BS,MD, New York, NY (Presenter) Nothing to Disclose
Cecilia L. Mercado, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Chloe M. Chhor, MD, Brooklyn, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Alexandra.millet@nyumc.org
TEACHING POINTS
The purpose of this exhibit is: 1. Identify the common clinical presenting symptoms of pathologic nipple discharge, and contrast this
with the presentation of physiologic nipple discharge. 2. Describe the initial imaging approach and clinical management of the
different causes of pathologic nipple discharge in patients of varying ages. 3. Illustrate the different pathologic causes for nipple
discharge with case examples at mammography, ultrasound, and MRI.
TABLE OF CONTENTS/OUTLINE
I. Clinical symptoms of pathologic discharge versus physiologic discharge II. Causes of pathologic discharge: Benign and malignant
III. Approach to imaging patients with pathologic nipple discharge IV. Clinical management of pathologic discharge V. Case examples
of imaging findings on mammogram, ultrasound, and MRI in patients with pathologic nipple discharge VI. Summary
 BR149-ED-X

Catch Me If You Know It: Common Image Quality Problems in 2D Versus 3D Mammography
All Day Room: NA Digital Education Exhibit

Participants
Mythri M. Reddy, MD, Chicago, IL (Presenter) Nothing to Disclose
Christopher Doyle, MD , Chicago, IL (Abstract Co-Author) Nothing to Disclose
Emily Marshall, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Deepa Sheth, MD, Chicago, IL (Abstract Co-Author) Research Grant, Guerbet SA
Kirti M. Kulkarni, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Ingrid Reiser, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
mythri.reddy@uchospitals.edu
TEACHING POINTS
1. Identifying image quality issues in 2D and 3D mammography before they necessitate a technical recall. 2. Creating awareness in
both technologists and radiologists to adhere to EQUIP-driven criteria and to recognize image quality problems. 3. Case-based
demonstration of common image quality problems in 2D versus 3D mammography imaging.
TABLE OF CONTENTS/OUTLINE
With the advent of full-field digital mammography detectors, image quality in mammography has improved tremendously. Technical
problems are less severe and less frequent, but are more subtle and difficult to identify. This exhibit will highlight common image
quality problems and will compare their appearance in 2D versus 3D mammography. These include, in no particular order, (1) patient
motion/insufficient compression, (2) biopsy clips, (3) large calcifications, (4) EMI artifacts, (5) detector pixel artifacts, (6) detector
line artifacts, (7) grid artifacts. The overall goal of this exhibit is to help technologists and radiologists recognize these problems
while the patient is still in the mammography suite, preventing a technical repeat.
 BR150-ED-X

Architectural Distortion Revisited: Pearls and Pitfalls in the Detection of Architectural Distortion on
Mammography and Spectrum of Pathologies
All Day Room: NA Digital Education Exhibit

Participants
Nishigandha P. Burute, MBBS, MD, Thunder Bay, ON (Abstract Co-Author) Nothing to Disclose
Amer Alaref, MD, Thunder Bay, ON (Presenter) Nothing to Disclose
Anatoly Shuster, MD, Thunder Bay, ON (Abstract Co-Author) Nothing to Disclose
Radu Rozenberg, MD, Ottawa, ON (Abstract Co-Author) Nothing to Disclose
John A. O'Brien, MD, Thunder Bay, ON (Abstract Co-Author) Nothing to Disclose
Richard Bitar, MD, Thunder Bay, ON (Abstract Co-Author) Nothing to Disclose
Colm E. Boylan, MBBCh, FRCR, Hamilton, ON (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
nishirad@gmail.com
TEACHING POINTS
1)Architectural distortion (AD) is the third most common mammographic abnormality indicating breast carcinoma (CA) 2)AD is often
found on prior mammograms of newly detected cancers. 3)Identification of AD is facilitated by diligent inspection of normal
architecture of ductal & lobular units, Coopers ligaments, interspersed islands of fat & distortion of these structures thereof. 4) Fat
necrosis, sclerosing adenosis, complex sclerosing lesions, tubular, invasive lobular & ductal CA may present as AD. 5)Central lucency
in complex sclerosing lesions, extension to the skin surface in postsurgical scars, curvilinear calcifications in fat necrosis &
associated suspicious calcifications in invasive CA help assessment.
TABLE OF CONTENTS/OUTLINE
Familiarity with patterns of normal parenchymal architecture on screening mammography enables detection of AD when present.
Examples of invasive ductal, lobular and tubular CA, radial scars, complex sclerosing lesions, fat necrosis & sclerosing adenosis
among others, presenting as AD will be provided, with an emphasis on missed cancers. Comparing with serial prior mammograms,
attention to previous negative biopsies and challenges encountered in dense breasts will be discussed. A systemic approach
enabling detection of AD, and minimizing the probability of missed cancers in routine radiology practice will be provided.
 BR151-ED-X

Treatment of Breast Fibroadenoma with High Intensity Focused Ultrasound (HIFU)

All Day Room: NA Digital Education Exhibit

Participants
Alana A. Lewin, MD, New York, NY (Presenter) Nothing to Disclose
Kathie-Ann Joseph, New York, NY (Abstract Co-Author) Nothing to Disclose
Jocelyn Acosta, New York, NY (Abstract Co-Author) Nothing to Disclose
Chloe M. Chhor, MD, Brooklyn, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
alana.amarosa@nyumc.org
TEACHING POINTS
1. Describe the theory of high-intensity focused ultrasound (HIFU).2. Discuss the use of HIFU in treating patients with a
symptomatic fibroadenoma (FA) using Echopulse® (Theraclion, Paris, France).3. Demonstrate the clinical and imaging outcomes of
patients with a FA treated with HIFU.
TABLE OF CONTENTS/OUTLINE
1. Theory of HIFU and mechanism of action2. Use of HIFU for treating breast fibroadenoma3. Clinical and imaging outcomes of
patients with a fibroadenoma treated with HIFU
 BR152-ED-X

The Wire and Beyond: A Review of Recent Advances in Breast Imaging Pre-Operative Needle Localization
Technology
All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit
Identified for RadioGraphics

Participants
Megha Madhukar Kapoor, MD, Houston, TX (Presenter) Nothing to Disclose
Miral Patel, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Marion E. Scoggins, MD, Houston, TX (Abstract Co-Author) Institutional Research Grant, General Electric Company
TEACHING POINTS
Review pre-operative breast localization and available methods for localization. Illustrate how potential complications of wire
localization such as wire migration can be avoided by using non-wire localization techniques. Discuss major advantages of the non-
wire localization techniques (device can be placed 5-30 days before surgery thus avoiding potential delays in operating room start
times, less nontarget (healthy) tissue is removed and the approach to localization does not affect surgical approach). Compare and
contrast the different localization techniques
TABLE OF CONTENTS/OUTLINE
A. Overview of pre-operative localization as an aid to surgical excision B. Wire Localization (advantages and disadvantages) i. Case
examples of wire migrated into chest wall, implant rupture, retained wire fragment C. Non-Wire Localization Techniques (advantages
and disadvantages) i. I-125 Radioactive Seed, Case example with lost seed ii. Infrared SAVI SCOUT Radar, Case example iii.
Magnetic Seed, Case example iv. Radiofrequency Identification, Case example D. Summary table comparing different Localization
techniques
 BR153-ED-X

Inflammatory Breast Cancer: A Multimodality Approach

All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Maria T. Fernandez Taranilla, Madrid, Spain (Presenter) Nothing to Disclose
Sara Jimenez Arranz, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Manuel Delgado Marquez, MD, Getafe, Spain (Abstract Co-Author) Nothing to Disclose
Rocio Gonzalez-Tovar, MBBS, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Javier Torrens, Parla, Spain (Abstract Co-Author) Nothing to Disclose
Leisy Sotolongo, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
maiteftr85@gmail.com
TEACHING POINTS
1. To review the pathophysiology of inflammatory breast cancer(IBC) and the major differential diagnoses.2. To describe the role of
imaging not only in diagnosis but also in staging, planning treatment and follow-up.3. To gain an awareness of the importance of
multidisciplinary approach in IBC to designing the best treatment.
TABLE OF CONTENTS/OUTLINE
IBC is an aggressive form of locally advanced breast cancer involving more than a third of the skin of the breast. It is a rare breast
cancer with a highly virulent course and low 5-year survival rate. The "inflammatory" skin changes are due to lymphedema caused
by tumor emboli within the dermal lymphatics.The symptoms arise rapidly compared to non IBC, typically less than 3 months.
Patients present with enlarged,warm, and tender breast as mastitis. Both tissue diagnosis of malignancy and clinical evidence of
infammatory disease are required to confirm the diagnosis. Proper diagnosis and staging of IBC is critical to treatment planning and
requires a multidisciplinary approach that includes imaging. We describes the typical imaging features of IBC at mammography,
ultrasound, MRI and PET/CT. Radiolologists play an integral role as part of the multidisciplinary team caring for patients with IBC and
will play a mayor role in the future with "Radiogenomics" and minimally invasive therapeutic procedures.
 BR154-ED-X

Metastasis to the Breast: Where Do They Come From?

All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Mariana M. Da Costa, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Giselle G. Mello, PhD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Marcela P. Viana, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Cecilia S. Goldman, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rafael L. Macedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marco A. Costenaro, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciano F. Chala, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera L. Aguillar, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Christopher S. Bae, MD, Brasilia, Brazil (Abstract Co-Author) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Dafne C. Andrade, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
mariana.mcosta@grupofleury.com.br
TEACHING POINTS
- Techniques to differentiate extra-mammary metastasis from primary breast lesions. - Imaging features of the main tumors that
metastasize to the breast. - Case studies. - Explain the findings and offer advice on work-up.
TABLE OF CONTENTS/OUTLINE
- Define the types of intra-mammary and extra-mammary metastasis. - Patient workup (clinical history, diagnostic methods and
biopsy to confirm). - Imaging appearance by modality and histological findings as well as potential pitfalls. - Learning which are the
major tumors that metastasize to the breast and the patterns of dissemination. - Describe the management and prognosis of
patients who present with this diagnosis.
 BR155-ED-X

Contrast Enhanced Spectral Mammography (CESM) - Principle, Protocol, Interpretation, and Challenge
All Day Room: NA Digital Education Exhibit

Participants
Juan Huang, Chengdu, China (Abstract Co-Author) Nothing to Disclose
Yun Qin, MD, Chengdu, China (Abstract Co-Author) Nothing to Disclose
Ying Liu, Chengdu, China (Abstract Co-Author) Nothing to Disclose
Huizhi Cao, PhD, Beijing, China (Presenter) Nothing to Disclose
Amiee Chen, Shanghai, China (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Contrast-enhanced spectral mammography(CESM) is known to have high sensitivity for detecting breast cancers and it has the
potential to become an essential modality for assessing breast lesions, using CT contrast to detect breast cancer by providing
morphologic and functional information detecting abnormal anatomic structures. In a multi-center study, we have done more than
400 cases of tissue diagnosis proven breast cancer. The focus of this presentation is to discuss CESM mechanism, value added in
breast cancer screening, staging, and treatment monitoring. The illustration cases focus on the morphology and the enhancement
characteristics of both malignant and benign lesions on CESM.
TABLE OF CONTENTS/OUTLINE
1. To introduce the principle of CESM, contrast protocol. abnormal enhancing characteristics 2. To illustrate cases where CESM
aided in appropriate BIRADS assessment and management of equivocal mammographic lesions 3.To show diagnostic performance of
contrast-enhanced spectral mammography for suspicious malignant microcalcifications (BI-RADS 4) 4. To discuss strategies for
appropriate utilization of CESM in challenging, equivocal, or complicated mammographic abnormalities.
 BR156-ED-X

Breast Cancer MRI Staging: An Easy to Follow Algorithm

All Day Room: NA Digital Education Exhibit

Participants
Celia Astor Rodriguez, MD, Toledo, Spain (Presenter) Nothing to Disclose
Lina M. Cruz Hernandez, ARRT, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Paul M. Aguilar Angulo, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Cristina Romero, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Aitana Palomares, MD, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
Carolina De la Cruz, MD, Getafe, Spain (Abstract Co-Author) Nothing to Disclose
Pilar Sanchez Camacho, Toledo, Spain (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
Breast magnetic resonance imaging (MRI) is the most accurate imaging modality in local staging of breast cancer. Allows radiologist
to describe precisely the T (tumor) and N (nodes) descriptors of the TNM staging. In addition to evaluating the extent of disease in
the ipsilateral breast, MRI could distinguish synchronous cancers in the contralateral breast that were otherwise mammographically
and clinically occult. Radiologist must be familiar with the BI-RADS® and TNM classifications to refer satisfactory information for the
clinicians.
TABLE OF CONTENTS/OUTLINE
The essential findings in breast MRI, following the guidelines of BI-RADS® classification, firstly include parameters that document
breast characterization, then parameters that describe the tumor and finally, other descriptors that assess the extent of disease
such as associated features like skin or pectoralis invasion or the presence of pathologic lymph nodes. We introduce a case based
algorithm, based on BI-RADS® classification and according to TNM staging category, that may facilitate the interpretation of
findings, an adequate staging and the report to a multidisciplinary unit.
 BR157-ED-X

MRI BI-RADS 3: When It Is and When It Is Not

All Day Room: NA Digital Education Exhibit

Participants
Fernanda G. Barbosa, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciano F. Chala, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Giselle G. Mello, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera L. Aguillar, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marcia M. Aracava, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
lfchala@gmail.com
TEACHING POINTS
To review MRI candidates of probably benign (BI-RADS 3) lesions To illustrate challenging clinical situations affecting assessment
and management To present a pictorial review of BI-RADS 3 lesions, and situations of reclassificate benign or suspect, with
teaching points
TABLE OF CONTENTS/OUTLINE
Background - definition - pillars lesion were correctly assessed as probably benign oncological safety of follow-up Which lesions can
be classified as B3? focus mass enhancement non-mass enhancement Patient follow-up protocol B3. Lesions that can not be
classified as B3. Situations to change B2 or B4. Importance of clinical context and complete imaging workup before BIRADS 3
classification
 BR158-ED-X

Male Breast: The Good, the Bad, and the Ugly

All Day Room: NA Digital Education Exhibit

Participants
Marcela P. Viana, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera L. Aguillar, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Barbara H. Bresciani, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Giselle G. Mello, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Mariana M. Da Costa, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciano F. Chala, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marcia M. Aracava, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Mariana B. Mazucato, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
marcelapviana@gmail.com
TEACHING POINTS
Normal anatomy of male breast Review imaging features of gynecomastia and its differentiation from breast cancer. Recognize and
illustrate radiologic and pathologic features of benign and malignant male breast diseases, and how they differs from its female
counterpart Discuss the epidemiology, work-up and treatment of male breast cancer Discuss which breast diseases can affect
transgender women
TABLE OF CONTENTS/OUTLINE
Review the anatomy and histology of the male x female breast, to understand the differential diagnosis of male breast diseases.
Review the different radiographic features of gynecomastia, the most common disease affecting the male breast. Illustrate benign
and malign diseases with pathologic correlation. Review the most prevalent breast tumors in men and their main imaging features
What to do with high risk patients. Breast diseases in transgender women: what are the differences and how to follow up these
patients
 BR159-ED-X

MRI vs Pathology: Evaluation of Pathologic Complete Response of Breast Cancer Subtypes After Neoadjuvant
Chemotherapy - Can We Avoid Breast Surgery in the Era of Personalized Medicine?
All Day Room: NA Digital Education Exhibit

Participants
Maria Ares-Rego, MD, Santiago de Compostela, Spain (Presenter) Nothing to Disclose
Beatriz Fernandez Roduiguez, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Sandra Baleato Gonzalez, MD, PhD, Santiago, Spain (Abstract Co-Author) Nothing to Disclose
Maria V. Trujillo Ariza, MD, Villagarcia de Arosa, Spain (Abstract Co-Author) Nothing to Disclose
Pastora Rivas Pumar, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Maria Dolores Abal Arca, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Roberto Garcia Figueiras, MD, PhD, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Michel Herranz Carnero, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
mariaaresrego@yahoo.es
TEACHING POINTS
Describe the molecular subtypes and review their different biology , clinical features , prognosis and treatment outcomes Describe
the utility of MRI sequences ( Dynamic and Diffussion Weighted Imaging ) in staging, in evaluation of response to neoadyuvant
chemotherapy , and in determining pathologic complete response of breast cancer. Define the accuracy of MRI in determining
pathologic complete response in breast cancer subtypes . Relate a comparative approach of Radiological and Pathologic Response
Criteria Discuss the role of breast MRI in the plannig of surgery and of the therapeutic protocols , with respect to the quantification
of residual disease Review cases of degree of response after neoadyuvant chemotherapy in the breast cancer subtypes
TABLE OF CONTENTS/OUTLINE
Classification of Molecular Subtypes of breast cancer and evaluation the main differences . Imaging features Utility of pre-operative
MRI in staging and in evaluation of response to chemotherapy Correlation between Radiological response criteria ( RECIST) and
Pathologic response criteria ( Miller-Payne) Role of breast MRI in treatment decisions Ilustrative cases
 BR160-ED-X

Beyond Ducts and Lobules: Imaging of Non-Epithelial Tumors of the Breast

All Day Room: NA Digital Education Exhibit

Awards
Certificate of Merit

Participants
Sebastien Moliere, MD, Strasbourg, France (Presenter) Nothing to Disclose
For information about this presentation, contact:
sebastien.moliere@chru-strasbourg.fr
TEACHING POINTS
- Radiation-induced angiosarcoma is a rare but severe lesion. The diagnosis relies on clinical- radiological correlation and skin biopsy
of every cutaneous lesion appearing within the field of radiation therapy - Lipoma is the most frequent non-epithelial tumor of the
breast. The diagnosis is made by imaging, yet a lipoma increasing in size requires biopsy to differentiate it from a well-differenciated
liposarcoma - Breast fibromatosis appears as a spiculated fibrous chest wall mass. It requires specialized care given the high risk of
local recurence after surgery - Diagnosis of hemopathies and metastases of the breast relies on clinical context and the imaging
appearance of well-defined multiple breast masses located in both fibroglandular and subcutaneous tissue - Benign phyllods are
fibro-epithelial tumors with imaging features similar to those of fibroadenoma, except for an increase in size, requiring pathology
examination
TABLE OF CONTENTS/OUTLINE
Part 1. Masses mimicking ductal carcinoma : benign (granular cell tumor, hemangioma, lymphocytic mastitis) and malignant
(sarcomas with focus on angiosarcoma, hemopathies, metastases) Part 2. Masses mimicking a fibroadenoma (phylloid, 'omas') Part
3. Fat-containing masses Part 4. Chest wall masses Partie 5. Cutaneous lesions
 BR161-ED-X

Palpable Lesions of Breast: What the Radiologist Should Know

All Day Room: NA Digital Education Exhibit

Participants
Mariana B. Mazucato, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Bruna M. Thompson, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Giselle G. Mello, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Fernanda G. Barbosa, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciano F. Chala, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera L. Aguillar, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marcia M. Aracava, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
tatianatucunduva@gmail.com
TEACHING POINTS
The perception of a palpable lesion is the second main complaint that leads the patient to the specialist. A palpable lesion
commonly terrifies most patients, and although the majority are related to benign causes, this is the most common symptom
reported by breast cancer patients. Anamnesis, physical examination and imaging tests should work together to correctly identify
which patients can be reassured and which ones need to persist the diagnostic investigation because it is a suspected lesion. Data
from the history and physical examination of the patients, as well as their age, dictate the choice of the best diagnostic tool,
whether mammography, ultrasound, MRI or tomosynthesis. Algorithms: To discuss current imaging guidelines for palpable breast
abnormalities through a series of cases
TABLE OF CONTENTS/OUTLINE
Review the appropriate criteria to select the initial imaging method for a palpable lesion, according to age and clinical information,
including a section of palpable circumscribed masses in women <40y. When the biopsy is imperative? To present illustrative cases of
palpable malignant and benign lesions in the different imaging methods correlating with the pathology. To present algorithms based
on the most common palpable lesions. Illustrate a particular section: lactating women and pregnant.
 BR162-ED-X

Breast Implants: A Pictorial Review of Implant Complications

All Day Room: NA Digital Education Exhibit

Participants
Sumin L. Kong, MD, Winston-Salem, NC (Presenter) Nothing to Disclose
Margaret A. Yacobozzi, MD, Winston-Salem, NC (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
slee2@wakehealth.edu
TEACHING POINTS
-Review types and locations of breast implants-Discuss normal multimodality imaging appearance of breast implants-Discuss
multimodality features of implant complications-Learn imaging optimization for implant complications and detection of breast cancer
TABLE OF CONTENTS/OUTLINE
-Types of breast implants -Saline, gel, cohesive gel silicone (gummy bear) -Single, double-lumen-Location of breast implants -
Retroglandular -Retropectoral-Normal multimodality (MR, MG, US) imaging appearance of breast implants-Multimodality imaging (MR,
MG, US, CT) features of implant complications -Early complications -Infection -Hematoma -Late complications -Implant rupture;
intracapsular vs extracapsular -Implant herniation -Gel bleed -Capsular contracture -Granuloma formation-Discuss imaging
techniques for implant complications-Breast Cancer Detection with Implants -Imaging optimization -Techniques for biopsy
 BR163-ED-X

Ready for BI-RADS Mammography Assessment Categories? Take a Quiz!

All Day Room: NA Digital Education Exhibit

Participants
Daniela Angulo Salazar, MD, Mexico City, Mexico (Presenter) Nothing to Disclose
Paola Gonzalez Balboa, MD, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Christian A. Zamora, MD, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Veronica Benitez Ortiz, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Jonathan Salazar Segovia, MD, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Rosaura E. Fuentes Corona, MD, Mexico, Mexico (Abstract Co-Author) Nothing to Disclose
Jorge Vazquez-Lamadrid, MD, Mexico, Mexico (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
daniela.angulo.salazar@gmail.com
TEACHING POINTS
1. Learn through pictures and cases the most proper BI-RADS mammography category for the lesion of interest2. Provide simple
diagrams of the different lesions in relation to its corresponding category to help the residents to get into this lexicon3. Become
familiar with the relevant points of ACR BI-RADS mammography and hopefully more experienced at time of diagnosis
TABLE OF CONTENTS/OUTLINE
- Introduction- Illustrative cases exposure - Analysis of conflictive categories- BI-RADS mammography categories assessment
resume- Summary
 BR164-ED-X

Pseudoangiomatous Stromal Hyperplasia of the Breast: What the Pathology May Explain from Different
Imaging Findings
All Day Room: NA Digital Education Exhibit

Participants
Carolina D. Rossi, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Daniela G. Giannotti, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Katia Pincerato, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Patricia A. Teixeira, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rafaela C. Bastreghi, MD, Sao Paulo , Brazil (Abstract Co-Author) Nothing to Disclose
Vivian N. Omura, MD, SAO PAULO, Brazil (Abstract Co-Author) Nothing to Disclose
Ligia A. Yamashita, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera Christina C. Ferreira, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Larissa Moyses, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Su J. Hsieh, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marco A. Costenaro, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carla B. Dequi, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Marcos F. Docema, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Giovanni G. Cerri, MD,PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
carolrossi@outlook.com
TEACHING POINTS
The purpose of this exhibit is: 1. To analyze the imaging findings of PASH at Ultrasound, Mammography and MRI. 2. To identify the
interface of PASH lesions with surrounding breast tissue at histological sections, and to correlate the relation of this limit with the
imaging findings. 3. To classify the histological type of PASH as to the predominance of cellularity: hypercellular or hypocellular
(collagenized) and the relation with MRI enhancement. 4. To evaluate the number of fragments obtained and if the biopsy
performed includes the periphery of the lesion, where there is representation of the interface with the surrounding tissue.
TABLE OF CONTENTS/OUTLINE
Imaging aspects of PASH: - Mammography. - Ultrasound. - MRI. Pathological features of PASH: - Correlation with histological and
imaging margins. - Interface with the surrounding breast tissue. - Correlation with cellularity and MRI enhancement. - Type of
biopsy and number of fragments obteined.
 BR169-ED-X

Comparison of DCIS and Invasive Ductal Carcinoma on Digital 2D versus Synthesized 2D Mammography
All Day Room: NA Digital Education Exhibit

Participants
Shreena Shah, MD, New York, NY (Presenter) Nothing to Disclose
Katerina Dodelzon, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Janine T. Katzen, MD, New York, NY (Abstract Co-Author) Travel support, Seno Medical Instruments, Inc
For information about this presentation, contact:
shb9167@nyp.org
TEACHING POINTS
- Recently, the implementation of synthesized 2D (s2D) mammography has been evaluated as a potential replacement for
conventional 2D digital mammography in an effort to reduce patient radiation - Recognize the limitations in using s2D imaging due to
its inferior spatial resolution - Illustrate differences in imaging appearance of ductal carcinoma in situ (DCIS) presenting as
microcalcifications in s2D and DM (digital mammography) with digital breast tomosynthesis (DBT) - Demonstrate differences in
appearance of invasive cancer on s2D and DM - Highlight important similarities and differences in imaging features of DCIS and
invasive cancer in an effort to make readers aware of the advantages and potential limitations of s2D imaging
TABLE OF CONTENTS/OUTLINE
- Summary of the recently published literature that has compared outcomes of screening with S2D/DBT vs DM/DBT - Case-based
pictorial essay of several (10-15) cases of DCIS and invasive carcinoma from our institution demonstrating imaging characteristics
seen on screening DM and s2D - Highlight imaging features of biopsy-proven carcinoma of varying size, distribution, and morphology
at the time of screening examination - Self-assessment with sample cases for readers to compare and contrast these features on
both types of imaging
 ED001-SU

Breast Sunday Case of the Day

Sunday, Nov. 25 7:00AM - 11:59PM Room: Case of Day, Learning Center

AMA PRA Category 1 Credit ™: .50

Participants
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Susan O. Holley, MD, PhD, Raleigh, NC (Abstract Co-Author) Nothing to Disclose
Alexander B. Sevrukov, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Catherine A. Young, MD, JD, Saint Louis, MO (Abstract Co-Author) Research support, Hologic, Inc
Cheryl R. Herman, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Mai A. Elezaby, MD, Madison, WI (Abstract Co-Author) Research Grant, Exact Sciences Corporation
Lonie R. Salkowski, MD,PhD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Urvi A. Tailor, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Ramapriya Ganti, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1) Identify, characterize, and analyze abnormal findings on multimodality breast imaging studies. 2) Develop differential diagnostic
considerations based on the clinical information and imaging findings. 3) Recommend appropriate management for the patients
based on imaging findings.
 VW 23

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Sunday, Nov. 25 10:15AM - 11:25AM Room: Booth 5530

Participants
Terri A. Gizienski, MD, Greenwood Village, CO (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 VW 21

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Self-guided Reading
Sessions: Siemens Healthineers Vendor Workshop
Sunday, Nov. 25 10:30AM - 5:00PM Room: Booth 5530

Program Information
You are invited to our self-guided reading sessions. With syngo® Breast Care workstations configured especially to allow you to
work at your own place at a time that suits you! A series of breast tomosynthesis and contrast enhanced mammography cases
presented as problem cases with a solution enables you to develop and test your reading skills.
 VW 22

Automated Breast Volume Scanner (ABVS) Self-guided Reading Sessions: Siemens Healthineers Vendor
Workshop
Sunday, Nov. 25 10:30AM - 5:00PM Room: Booth 5530

Program Information
With syngo® Ultrasound Breast Analysis (sUSBA) software, self-guided reading sessions with real clinical cases will enable you to
become familiar with the coronal plane while providing practical approaches to interpretation of 3D automated breast ultrasound.
 VW 51

Tomosynthesis Guided Prone Breast Biopsy Solutions: Hologic Vendor Workshop

Sunday, Nov. 25 10:30AM - 11:30AM Room: Booth 5524

Participants
Daniel E. Lehrer, MD, CABA, Argentina (Presenter) Speaker, Hologic, Inc; Institutional research agreement, Siemens AG
Program Information
Clinical benefits of tomosynthesis guided biopsy which includes a hands-on demonstration of the Affirm® Prone Biopsy System and
the Brevera® Breast Biopsy System (Affirm® Prone Biopsy System and Brevera® Breast Biopsy System)
Registration
https://hologicrsna.com
 SSA01

Breast Imaging (Update on Screening)

Sunday, Nov. 25 10:45AM - 12:15PM Room: E450A

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Catherine S. Giess, MD, Wellesley, MA (Moderator) Nothing to Disclose
Debra L. Monticciolo, MD, Temple, TX (Moderator) Nothing to Disclose
Sub-Events
SSA01-01 Interval Cancers in a Large Prospective Breast Tomosynthesis Screening Trial
Sunday, Nov. 25 10:45AM - 10:55AM Room: E450A

Participants
Kristin Johnson, MD, Lund, Sweden (Presenter) Nothing to Disclose
Kristina Lang, MD,PhD, Malmo, Sweden (Abstract Co-Author) Travel support, Siemens AG Speaker, Siemens AG
Debra M. Ikeda, MD, Stanford, CA (Abstract Co-Author) Scientific Advisory Board, Grail, Inc; Reviewer, Siemens AG
Hanna Sartor, MD,PhD, Malmo, Sweden (Abstract Co-Author) Travel support, Siemens AG Speaker, Siemens AG
Ingvar T. Andersson, MD, PhD, Malmo, Sweden (Abstract Co-Author) Nothing to Disclose
Sophia Zackrisson, Malmo, Sweden (Abstract Co-Author) Speaker, AstraZeneca PLC ; Speaker, Siemens AG; Travel support,
AstraZeneca PLC; Travel support, Siemens AG
For information about this presentation, contact:
kristin.johnson@med.lu.se
PURPOSE
To assess interval cancer rate in a large, prospective digital breast tomosynthesis (DBT) screening trial in comparison with a
concurrent screening population and to assess tumor characteristics of interval cancers in DBT-screening.
METHOD AND MATERIALS
The prospective ****Trial, comparing digital breast tomosynthesis (DBT) with digital mammography (DM) in 14,848 women has
shown a significantly increased sensitivity with DBT. Interval cancer rate in the trial was compared with a concurrent screening
population; i.e. women participating in DM screening at the same screening site during the same time period (2010-15, n=100,273
screens). Interval cancers and concurrent screens were identified through linkage with the Radiology Information System, the
**Cancer Registry and the National Quality Register Breast Cancer. Confidence intervals (CI) 95 % were calculated for rates and
difference between rates. Tumor characteristics were retrieved from pathology reports and invasive cancers classified according to
St Gallen subtypes.
RESULTS
In total, there were 22 interval cancers in the ****Trial. The interval cancer rate was 1.5 per 1000 screens [22/14,848] (95% CI
0.9-2.2) in the ****Trial and 1.8 per 1000 screens [179/100,273] (95% CI 1.5-2.1) in the concurrent population. Although the
interval cancer rate was lower in the trial, the difference of 0.3 was not statistically significant (95 % CI -0.5-0.9). Among the
interval cancers in the ****Trial, the mean cancer size was 17 mm (range 2-37 mm), 2 were DCIS (grade 2 and 3), 5 luminal A-
like, 9 luminal B-like HER2-, 2 luminal B-like HER2+ and 4 triple negative.
CONCLUSION
The slightly lower interval cancer rate in the trial might indicate that DBT-screening leads to the detection of clinically relevant
cancers. Still, a relatively large proportion of the interval cancers had unfavorable prognostic characteristics.
CLINICAL RELEVANCE/APPLICATION
Analysis of interval cancers is important in order to elucidate the future value of DBT in screening.
SSA01-02 Does Tomosynthesis Work For Everyone?
Sunday, Nov. 25 10:55AM - 11:05AM Room: E450A

Participants
Ethan O. Cohen, MD, Houston, TX (Presenter) Spouse, Consultant, Medtronic plc; Spouse, Consultant, Novo Nordisk AS; Spouse,
Consultant, Eli Lilly and Company; Spouse, Consultant, AstraZeneca PLC
Rachel E. Perry, MD, Birmingham, AL (Abstract Co-Author) Nothing to Disclose
Ashmitha Srinivasan, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Hilda H. Tso, DO, Houston, TX (Abstract Co-Author) Nothing to Disclose
Kanchan Phalak, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Michele D. Lesslie, DO, Bellaire, TX (Abstract Co-Author) Nothing to Disclose
Karen E. Gerlach, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Jessica W. Leung, MD, Houston, TX (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc;
Speakers Bureau, FUJIFILM Holdings Corporation
For information about this presentation, contact:

ecohen@mdanderson.org
PURPOSE
To compare the performance of full-field digital screening mammograms (FFDM) with and without digital breast tomosynthesis (DBT)
in women with and without breast implants.
METHOD AND MATERIALS
An IRB-approved, HIPAA-compliant retrospective review was performed of 103,070 consecutive screening mammograms obtained
from February 2011 through June 2014. Recall rates (RRs), cancer detection rates (CDRs), and positive predictive values for recall
(PPV1s) were analyzed.
RESULTS
The following data compare FFDM and FFDM-DBT: 67,331 FFDM and 28,835 FFDM-DBT from women without implants yielded RRs of
8.0% and 6.3%, respectively (p<0.00001); CDRs of 4.1 and 5.0 per 1000 exams, respectively (p=0.07); and PPV1s of 5.1% and
8.0%, respectively (p<0.0001). 4325 FFDM and 2579 FFDM-DBT from women with implants yielded RRs of 5.2% and 4.1%,
respectively (p=0.040); CDRs of 1.8 and 2.7 per 1000 exams, respectively (p=0.46); and PPV1s of 3.6% and 6.7%, respectively
(p=0.25). The same data is also used to evaluate the effect of implants on screening: 67,331 FFDM without implants and 4325
FFDM with implants yielded RRs of 8.0% and 5.2%, respectively (p<0.00001); CDRs of 4.1 and 1.8 per 1000 exams, respectively
(p<0.00001); and PPV1s of 5.1% and 3.6%, respectively (p=0.30). 28,835 FFDM-DBT without implants and 2579 FFDM-DBT with
implants yielded RRs of 6.3 and 4.1, respectively (p<0.00001); CDRs of 5.0 and 2.7, respectively (p=0.11); and PPV1s of 8.0 and
6.7, respectively (p=0.63).
CONCLUSION

Tomosynthesis improves the performance of digital screening mammography, while the presence of implants reduces its
performance. Specifically, tomosynthesis improved RRs, CDRs, and PPV1s for all women (with and without implants), though
statistically significant differences were seen only for RRs in women without implants, RRs in women with implants, and PPV1s for
women without implants. Implants were associated with decreased RRs, worse CDRs, and worse PPV1s for all screening exams
(FFDM and FFDM-DBT), but statistically significant differences were seen only for RRs for all screening exams and CDR for FFDM.
Further study with larger populations is warranted.
CLINICAL RELEVANCE/APPLICATION

The benefit of tomosynthesis has been incompletely studied in screening mammography patients with implants. This research
suggests that tomosynthesis is useful for screening women with implants in addition to those without implants.
SSA01-03 Disparities in Screening Mammography Cost-Sharing and Utilization Before and After the Affordable
Care Act (ACA) and the Revised USPSTF Guidelines
Sunday, Nov. 25 11:05AM - 11:15AM Room: E450A

Awards
Trainee Research Prize - Resident

Participants
Soudabeh Fazeli Dehkordy, MD, MPH, San Diego, CA (Presenter) Nothing to Disclose
A. Mark Fendrick, MD, Ann Arbor, MI (Abstract Co-Author) Consultant, Abbott Laboratories; Consultant, AstraZeneca PLC;
Consultant, sanofi-aventis Group; Consultant, F. Hoffmann-La Roche Ltd; Consultant, GlaxoSmithKline plc; Consultant, Merck & Co,
Inc; Consultant, Neocure Group LLC; Consultant, Pfizer Inc; Consultant, POZEN Inc; Consultant, Precision Health Economics LLC;
Consultant, The TriZetto Group, Inc; Consultant, Zanzors; Speakers Bureau, Merck & Co, Inc; Speakers Bureau, Pfizer Inc;
Researcher, Abbott Laboratories; Researcher, AstraZeneca PLC; Researcher, sanofi-aventis Group; Researcher, Eli Lilly and
Company; Researcher, F. Hoffmann-La Roche Ltd; Researcher, GlaxoSmithKline plc; Researcher, Merck & Co, Inc; Researcher,
Novartis AG ; Researcher, Pfizer Inc
Sarah Bell, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
Emily Kobernik, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
Vanessa Dalton, MD, Ann Arbor, MI (Abstract Co-Author) Expert Witness, Bayer AG
Ruth C. Carlos, MD, MS, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

sfazeli@ucsd.edu
PURPOSE

To assess changes in screening mammography cost-sharing and utilization before and after the Affordable Care Act (ACA) and the
revised USPSTF guidelines by race and income level.
METHOD AND MATERIALS

We used patient-level analytic files between 2004 and 2014 from Optum™ Clinformatics™ Data Mart. We included women 40-74
years 1) without a history of breast cancer or mastectomy with 2) at least one year of continuous enrollment in a given plan,
examining out-of-pocket payments and utilization for screening mammography. Trends for screening mammography utilization and
cost sharing elimination over time by race and income level were visually inspected. We then specifically calculated the slopes and
compared trends before and after 2009 and 2010 to assess the impact of ACA implementation and USPSTF guideline revisions on
screening mammography cost sharing elimination and utilization.
RESULTS

We identified an average of 2,173,686 commercially insured women ages 40-74 years with at least 12 months of continuous
We identified an average of 2,173,686 commercially insured women ages 40-74 years with at least 12 months of continuous
enrollment in a given plan per year. Overall, an upward trend was seen in the proportion of women with zero cost sharing over time
among all races and income level. Comparing trends for cost-sharing elimination before and after 2010, a statistically significant
upward but small trend was found among all races and income levels with no racial or income disparities evident. Screening
mammography utilization plateaued or showed a significant decline after 2009 in all income and racial groups except for African
Americans in whom screening rates continued to increase after 2009.
CONCLUSION
Impact of ACA cost-sharing elimination did not differ among various racial and income groups. Among our population of employer-
based insured women, the racial gap in screening mammography appeared to have closed and potentially reversed among African
American women.
CLINICAL RELEVANCE/APPLICATION

It is important to continue monitoring screening mammography utilization as health care policies and guidelines change, as these
changes may affect disparities in screening between different racial and income groups.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Ruth C. Carlos, MD, MS - 2015 Honored EducatorRuth C. Carlos, MD, MS - 2018
Honored Educator
SSA01-04 Screening Mammography: There is Value in Screening Women Aged 75 and Over
Sunday, Nov. 25 11:15AM - 11:25AM Room: E450A

Participants
Stamatia V. Destounis, MD, Scottsville, NY (Presenter) Research Grant, Hologic, Inc; Research Grant, Delphinus Medical
Technologies, Inc
Andrea L. Arieno, BS, Rochester, NY (Abstract Co-Author) Nothing to Disclose
Amanda Santacroce, Rochester, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

sdestounis@ewbc.com
PURPOSE

To review outcomes of screening mammography performed in women 75 and over to determine the rate of cancer diagnosed and
associated histology and surgical excision performed.
METHOD AND MATERIALS
Patients 75 years of age and over who presented for screening mammography and underwent biopsy with resultant malignant
pathology were retrospectively collected and analyzed to record patient demographics, lesion information, pathology results and
treatment. Cases of non-breast malignancy, cases of breast cancer diagnosed by modality other than mammography, and cases
with missing or incomplete records were excluded.
RESULTS
From 2007-2016 there were a total of 679,168 screening appointments, with 3,480 patients diagnosed with screen-detected
cancers (5.1/1000). 68,218 (10%) screenings were performed in women aged 75 and over; 530 of these women were diagnosed
with 560 breast cancers, for a cancer detection rate of 7.8/1000. Average patient age 80.3 (range 75-98). Lesions most frequently
presented as a mass (74%). A large majority (81%) of the malignancies diagnosed were invasive; 55.7% grade 2 or 3. Positive
lymph nodes were reported at surgical excision in 7.0% of patients. Tumor stage was largely stage 0 or 1 (64%); 12% were
determined to be stage II or III. 98% of cancers were surgically excised; twelve cancers were not due to advanced patient age or
overall degraded patient health.
CONCLUSION

For the relatively small percentage of our screening population that women 75 and over comprise (10%), the patients diagnosed in
this population made up 16% of all patients diagnosed with screen-detected cancers, a substantial cancer detection rate
(7.8/1000). Most of the cancers diagnosed were low grade, a significant number were invasive (81%), over half were grade 2 or 3,
and a majority were lower stage (0 or I). Most (98%) underwent surgical excision, suggesting that most women 75 and over are in
good health and want to pursue surgical excision. Screening mammography should be performed in this age group given the
incidence of breast cancer that exists.
CLINICAL RELEVANCE/APPLICATION

Ongoing debate exists regarding the age to cease screening mammography, citing lack of research in the aging population. Our
study demonstrates the value of screening women 75 and over.
SSA01-05 Risk of Breast Cancer After a False-Positive Screening Mammogram in Relation to Mammographic
Abnormality: Potential for Prevention?
Sunday, Nov. 25 11:25AM - 11:35AM Room: E450A

Participants
Rasika Rajapakshe, PhD, Kelowna, BC (Presenter) Nothing to Disclose
Miao Hui, PhD, Singapore, Singapore (Abstract Co-Author) Nothing to Disclose
Brenda A. Farnquist, MD, Chestnut Hill, MA (Abstract Co-Author) Nothing to Disclose
Janette Sam, RT, Vancouver, BC (Abstract Co-Author) Nothing to Disclose
Mikael Hartman, MD,PhD, Singapore, Singapore (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

rrajapak@bccancer.bc.ca
PURPOSE

False positive recall rates have consistently been identified as a harm of organized breast cancer screening. The extent to which
these recalled women are at increased risk of future breast cancer remains unclear.
METHOD AND MATERIALS

All women who had two or more screening mammograms between 1988-2013 in a large organized breast screening program were
included in the study. They were followed until a breast cancer diagnosis, last screen date +5 years, or end of follow-up on Dec 31,
2013, whichever came first. The relative risk (RR) of breast cancer for women with a false-positive test compared with women with
negative tests was estimated with Poisson regression, adjusted for age, and five calendar periods.
RESULTS
A total of 772,289 women with 4.82 million screening mammograms and a median follow up of 11.8 years were included. There were
238,860 women with false positive findings and 26,950 cancers of which 16,084 screen detected and 10,866 non screen detected.
The adjusted RR [Value (95% CI)] of breast cancer after the first false-positive test was 1.73(1.68-1.77) for all, 1.65(1.61-1.70)
for invasive, and 2.13(2.01-2.27) for in situ cancers respectively. The RR remained increased beyond 8 years after the first false-
positive test. Of the 5157 screen detected cancers after the first false positive test, 3358 (65%) were on the ipsilateral breast
while 1799 (35%) were on the contralateral breast. Women with breast density >50% at the time of false positive test had a
twofold risk of breast cancer; RR 2.07(1.99-2.14) while those with breast density <50% had a RR of 1.58(1.54-1.63). When
stratified for mammographic features found on the first false positive mammogram, architectural distortion plus mass had the
highest RR 4.68(3.16-6.93) for invasive cancers while calcifications alone and calcifications plus asymmetry had highest RR
5.57(4.88-6.36) and 4.07(2.49-6.66) for in situ cancers. These findings would require further validation.
CONCLUSION

False positive mammogram correlates with an increased risk of developing breast cancer. Screen detected breast cancers post
false positive mammogram most likely to occur in the ipsilateral breast.
CLINICAL RELEVANCE/APPLICATION

Mammographic features at the time of recall predict the risk of subsequent cancer and may warrant increased surveillance and/or
chemo-prevention.
SSA01-06 Stratification of Ductal Carcinoma in Situ in a Screening Population by Age and Grade Over a 16-Year
Period
Sunday, Nov. 25 11:35AM - 11:45AM Room: E450A

Participants
Angela Sie, MD, Long Beach, CA (Presenter) Nothing to Disclose
Gretchen M. Stipec, MD, Manhattan Bch, CA (Abstract Co-Author) Nothing to Disclose
Maya S. Hsieh, MD, Long Beach, CA (Abstract Co-Author) Nothing to Disclose
Ryan Kobayashi, Long Beach, CA (Abstract Co-Author) Nothing to Disclose
Tiesha Jones, Long Beach, CA (Abstract Co-Author) Nothing to Disclose
Stephen A. Feig, MD, Orange, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
asie@memorialcare.org
PURPOSE

To determine if there is a correlation between age and grade of screening-detected ductal carcinoma in situ (DCIS).
METHOD AND MATERIALS
We performed an IRB approved, retrospective review of screening-detected cases of DCIS at our large, community-based breast
center, diagnosed between 2001 and 2016. All cases of DCIS diagnosed from 2001 through 2016 were collated. Odds ratios were
produced from logistic regression analyses.
RESULTS

DCIS accounted for 20.9% (372/1781) of all cancers detected in our screening practice between 2001 and 2016.The total number
of cases in our study was 372, with a mean age of 60.6 years, median age of 60.0 years, mean size of 2.1 cm, median size of 1.8
cm. Age distribution showed 31.7% (118/372) were <55 years old, and 68.3% (254/372) were >= 55 years old. The tumor grade
was identified in 366 (98.4%) of the cases and included 6.8% (25/366) low grade, 46.4% (170/366) intermediate grade, and 46.7%
(171/366) high grade. Mammographic findings were 80.9 % (301/372) calcifications and 16.9% (63/372) mass or focal asymmetry.
Estrogen receptor (ER)/Progesterone receptor (PR) status was identified in 240 (64.5%) cases and included 81.3% (195/240) ER +
and 72.9% (175/240) PR+ cases. Logistic regression analysis revealed that for each unit decrease in age, the odds ratio of
intermediate versus low grade was 4 times higher (p=0.038), and high versus low grade was 5 times higher (p=0.009). No patient
(0/19) under 45 years old had low grade DCIS. Younger women were slightly more likely to have DCIS that was ER+ and PR+.
CONCLUSION

The vast majority of screening-detected DCIS cases (93.1%) were intermediate or high grade. Younger patients had a statistically
significant greater chance of intermediate or high grade DCIS versus low grade DCIS.
CLINICAL RELEVANCE/APPLICATION

Grade and age have an inverse relationship in the screening-detected DCIS at our institution during the 16 year period studied.
Since screening- detected DCIS in younger patients tends to be high or intermediate grade, screening mammography in younger
patients remains clinically important.
SSA01-07 Frequency and Cancer Yield of Probably Benign Breast Findings in Clinical Practice in the National
Mammography Database (BI-RADS Category 3)
Sunday, Nov. 25 11:45AM - 11:55AM Room: E450A

Participants
Mai A. Elezaby, MD, Madison, WI (Presenter) Research Grant, Exact Sciences Corporation
Colin Longhurst, Madison, WI (Abstract Co-Author) Nothing to Disclose
Priyadarshini Karthik, Reston, VA (Abstract Co-Author) Nothing to Disclose
Debapriya Sengupta, MBBS,MPH, Reston, VA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. Burnside, MD,MPH, Madison, WI (Abstract Co-Author) Dr. Burnside has a research grant from Hologic
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
Wendie A. Berg, MD, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Mythreyi Bhargavan-Chatfield, PhD, Reston, VA (Abstract Co-Author) Nothing to Disclose
Judy Burleson, Reston, VA (Abstract Co-Author) Nothing to Disclose
Cindy S. Lee, MD, Garden City, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
melezaby@uwhealth.org
PURPOSE

The Breast Imaging Reporting and Data System (BI-RADS) assessment category 3 describes a probably benign finding with a low
likelihood of malignancy of <= 2%. Since 2013, the BI-RADS Atlas discouraged its use at screening, which was reinforced in
Medicare's pay-for-performance Physician Quality Reporting System (PQRS) initiative. However, there are sparse data on the
frequency of use and cancer yield of BI-RADS 3 findings in clinical practice. This study assesses the frequency and cancer yield of
probably benign findings in screening and diagnostic mammography in the National Mammography Database (NMD), which is the
largest national database of mammography.
METHOD AND MATERIALS
This HIPAA-compliant and IRB-exempt study retrospectively analyzed data from screening and diagnostic mammograms performed
between 1/1/2009 and 12/31/2015 in the NMD. We calculated the overall frequency and cancer yield of BI-RADS 3 findings in
screening and diagnostic mammography. Cancer yield is defined as the number of breast cancers diagnosed among women with
probably benign findings within the study period. Exams with BI-RADS 3 findings but lacking biopsy or 12-month follow-up were
excluded . Exams from NMD facilities contributing data for < 2 years were excluded to ensure adequate follow-up.
RESULTS

Data from 6,421,365 screening and 1,264,929 diagnostic mammograms performed in 3,345,013 women at 261 NMD facilities in 31
states were analyzed. A total of 215,403 mammograms had probably benign findings, with frequency of 0.3% (20,060/6,421,365) in
screening and 15% (193,850 /1, 264,929) in diagnostic mammograms. Among the 101,025 women with BI-RADS 3 findings and at
least one follow up , 948 (0.94%, 95%CI 0.91, 1.03) women had a diagnosis of malignancy. Rates of BI-RADS 3 use were stable
over the 6 calendar years analyzed.
CONCLUSION

A probably benign, BI-RADS 3, assessment was rarely used in screening but common in diagnostic mammography in the NMD. The
overall cancer yield of probably benign findings was 0.94%, consistent with BI-RADS Atlas threshold of <= 2%. These results
support use of probably benign assessment in clinical practice for findings with low risk of malignancy.
CLINICAL RELEVANCE/APPLICATION

In the NMD, BI-RADS 3 showed appropriately low cancer yield of 0.9%. Proper use of BI-RADS 3 helps reduce cost, morbidity and
patient anxiety while increasing cost-effectiveness of screening.
SSA01-08 Digital Breast Tomosynthesis Improves Performance Metrics of Screening Mammogram in Women
Aged 40 to 54 Compared to Full Field Digital Mammogram
Sunday, Nov. 25 11:55AM - 12:05PM Room: E450A

Participants
Amir Imanzadeh, MD, Shelton, CT (Presenter) Nothing to Disclose
Maryam Etesami, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Sarvenaz Pourjabbar, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Laura J. Horvath, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Michelle Y. Giwerc, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Liane E. Philpotts, MD, New Haven, CT (Abstract Co-Author) Consultant, Hologic, Inc
For information about this presentation, contact:

amir.imanzadeh@yale.edu
PURPOSE

The starting age for screening mammogram has been a controversial subject mostly due to reported low cancer detection rate
(CDR) and high false positive recall rate (RR) in younger women. However, digital breast tomosynthesis (DBT) may improve
performance of screening mammogram and change the justification of screening guidelines. The purpose of this study is to compare
the performance of screening mammogram in women aged 40-54 between full field digital mammogram (FFDM) and DBT.
METHOD AND MATERIALS
In an IRB approved study, we retrospectively analyzed screening mammograms performed in women aged 40-54 at 2 of our imaging
centers from August 2008 through April 2017. We included all FFDM screening mammograms performed before and all DBT
screenings performed after installation of DBT at each site. DBT screening was offered to all women after installation at no
additional charge. RR, CDR, and characteristics of screen-detected cancers were compared between FFDM and DBT screening in 3
age groups: 40-44 (group 1), 45-49 (group 2) and 50-54 (group 3).
RESULTS
16938 FFDM and 28313 DBT were performed in women aged 40-54. In FFDM, RR significantly decreased from group 1 (17.4%) to
group 2 (14.0%) and group 3 (11.3%); however, in DBT screening, RR was only significantly decreased from group 1 (12.2%) to
group 2 (9.1%) and there was no significant difference between group 2 and group 3 (8.4%). RR of all 3 age groups were
significantly lower for DBT compared to FFDM. CDR per 1000 exams in FFDM was significantly lower in groups 1 and 2 (2.4 and 2.3)
compared to group 3 (5.0), but in DBT, there was no significant difference in CDR among the 3 groups (3.0, 4.4, and 3.8 in groups
1,2, and 3, respectively). There was significant increase in CDR from FFDM to DBT in women aged 45-49 (p=0.03). The ratio of
invasive to in-situ carcinomas were similar in FFDM and DBT subgroups.
CONCLUSION
DBT screening decreases RR and increases CDR compared to FFDM in women aged 40-49, and more prominently for 45-49 age
group. As a result, unlike FFDM, there is no significant difference in RR and CDR of ages 45-49 and 50-54 with DBT screening.
CLINICAL RELEVANCE/APPLICATION

Improved performance of DBT screening in women aged 40-49 compared to FFDM screening, may further justify recommendations
for starting screening mammogram at younger age.
SSA01-09 Six Years of Consecutive, Population-Based Screening with Digital Breast Tomosynthesis: Outcomes
by Screening Year and by Screening Round
Sunday, Nov. 25 12:05PM - 12:15PM Room: E450A

Participants
Emily F. Conant, MD, Philadelphia, PA (Presenter) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc; Consultant, iCAD,
Inc; Speaker, iiCME
Samantha P. Zuckerman, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. McDonald, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Susan Weinstein, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Aimilia Gastounioti, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Marie Synnestvedt, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Mitchell D. Schnall, MD, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Rebecca Hubbard, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
emily.conant@uphs.upenn.edu
PURPOSE
Digital breast tomosynthesis (DBT) improves screening outcomes by decreasing recalls and increasing cancer detection. However,
most DBT studies have analyzed prevalence rather than incidence screening. We investigate outcomes from 6 years of consecutive
DBT screening, at both the population level by year (DBT Y1-6) and patient level by round (DBT R1-6).
METHOD AND MATERIALS
71535 consecutive DBT screens in 33245 women were performed from 10/2011 to 9/2017 with 31246, 17801, 11067, 6560, 3553,
and 1308 screens in rounds 1-6, respectively. Recall rate % (RR), cancer detection (CDR), biopsy (BX) and false negative (FN)
rates per 1000 screens for each year (DBTY1-Y6) and each round (R1-R6) were analyzed. Outcomes were compared the prior year
of digital mammography (DM) screening (DMY0, n=10765). Cancer registry data was used to determine FN (defined as cancer
diagnosed <1 yr after negative screen) for DMY0-DBTY5 and DBT R1-R5.
RESULTS
At the Population Level, RR for DBTY1-6 was lower than DMY0 (7.9 vs 10.4%). However, DBT RR increased in DBTY1-3 (8.8, 9.0,
9.2%) before decreasing in DBTY4-6 (7.1, 6.1, 7.4%). CDR increased in DBTY1-3 (5.5, 5.8, 6.5) then decreased in DBTY4-6 (6.3,
5.1, 5.5) however, DBT CDR was higher in each year than DMY0 (4.8). Invasive CDR for DBTY1-4 and DBTY6 (3.9, 4.5, 4.5, 4.9,
respectively) were also higher than DM0 (3.4) however, DBTY5 was lower (2.9). BX increased from DMY0 of 1.9 to 2.2, 1.9, 2.0 in
DBTY1-3; in DBTY4-6, BX rates were lower (1.5, 1.3, 1.5). FN remained without significant change from DMY0 (1.1) to DBT1-5
(0.5, 0.3, 0.4, 0.8, 1.0). By Screening Round, RR decreased from DBTR1 to R5 from 10.6, 7.0, 5.1, 4.5, 4.1 until R6= 4.2%. CDR for
R1-6 were 6.4, 5.5, 6.1, 4.4, 4.2, 3.1, respectively. BX rate decreased after DBTR1 (2.5) for R2-6 (1.4, 1.4, 1.0, 0.8, 1.0). FN were
0.5, 0.5, 1.3, 0.6, 0.8 for R1-5. Whether by year or round, DBT PPV1 was higher than DM.
CONCLUSION

At the population level, DBT screening had higher overall CDR and lower RR than DM with FN rates remaining relatively stable. Little
data exists on outcomes by round of screening, however, our data may help guide DBT screening benchmarks.
CLINICAL RELEVANCE/APPLICATION
Consecutive years of DBT screening show decreased recall and increased PPV1 compared to DM alone. Further long term DBT
outcome data may help guide new, personalized screening algorithms by age, density and risk.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Mitchell D. Schnall, MD, PhD - 2013 Honored Educator
 SSA02

Breast Imaging (Contrast Enhanced Spectral Mammography)

Sunday, Nov. 25 10:45AM - 12:15PM Room: E450B

BR CT OI
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
John M. Lewin, MD, Denver, CO (Moderator) Nothing to Disclose
Maxine S. Jochelson, MD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
SSA02-01 Background Parenchymal Enhancement at Contrast-Enhanced Spectral Mammography (CESM) as a
Breast Cancer Risk Factor
Sunday, Nov. 25 10:45AM - 10:55AM Room: E450B

Participants
Yael Yagil, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Vera Sorin, BMedSc, Ramat Gan, Israel (Presenter) Nothing to Disclose
Renata Faermann, MD, Tel Aviv, Israel (Abstract Co-Author) Nothing to Disclose
Anat Shalmon, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Osnat Halshtok, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Michael Gotlieb, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Miriam Sklair-Levy, MD, Tel -Hashomer, Israel (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
mirisklair@gmail.com
PURPOSE
To assess the extent of background parenchymal enhancement (BPE) at contrast-enhanced spectral mammography (CESM), inter-
reader agreement in BPE classification, and correlation between BPE and breast cancer.
METHOD AND MATERIALS
Between 2012 and 2015 a total of 516 women underwent CESM imaging for screening and diagnostic purposes. BPE on CESM
images was retrospectively, independently and blindly graded by 4 reviewers using the following scale: minimal, mild, moderate or
marked. Inter-reader agreement was estimated using correlation coefficient (ICC). Associations between BPE and clinical factors,
biopsy rate and histopathology results were examined using a multivariate logistic regression analysis.
RESULTS
A total of 412 (80%) of women underwent CESM for screening purposes. Mean age was 53 (range 28-77) years and 86.2-94% had
a breast density BI-RADS score of C-D. Most women (76.4-90.5%) had minimal or mild BPE at CESM. Overall inter-reader agreement
on BPE scores was good (ICC 0.88, 95%CI 0.81-0.92). A total of 122 (24%) biopsies were performed with a malignant
histopathology result in 45 (37%) cases. On a multivariate analysis BPE demonstrated a significant association with age (P=0.004,
OR 0.942, 95%CI 0.905-0.981) and with biopsy performance rate (P=0.006, OR 2.646, 95%CI 1.319-5.307). Moderate or marked
BPE was predictive of a malignant biopsy result (P=0.002, OR 3.105, 95%CI 1.541-6.259).
CONCLUSION
CESM BPE is correlated with age and biopsy rate. Moderate or marked BPE is associated with malignant biopsy results, and hence
may predict an increased risk for breast cancer.
CLINICAL RELEVANCE/APPLICATION
CESM BPE grading may be used as an additional risk assessment tool for breast cancer.
SSA02-02 Contrast Enhanced Digital Mammography (CEDM) Helps to Safely Reduce Benign Breast Biopsies
Sunday, Nov. 25 10:55AM - 11:05AM Room: E450B

Participants
Margarita L. Zuley, MD, Pittsburgh, PA (Presenter) Investigator, Hologic, Inc
David Gur, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Uzma Waheed, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Marie A. Ganott, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Bronwyn Nair, MD, Sewickley, PA (Abstract Co-Author) Nothing to Disclose
Christiane M. Hakim, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Gordon S. Abrams, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
zuleyml@upmc.edu
PURPOSE
One criticism of breast imaging is the harm caused by the relatively high rate of biopsy of benign breast lesions -- particularly
BIRADS 4A and 4B lesions. The purpose of this project is to assess if CEDM during diagnostic evaluation could increase biopsy PPV
for soft tissue density lesions by reducing benign biopsies while not impacting biopsy of cancers.
METHOD AND MATERIALS
This HIPPA compliant IRB approved protocol accrued 57 consenting women aged 34-74 (avg 49) years with 60 BIRADS 4A or 4B soft
tissue lesions scheduled for ultrasound (US), stereotactic or tomosynthesis (DBT) directed biopsy from April 2016-November 2017.
CEDM was performed immediately prior to biopsy. The cohort included 46 masses, 6 asymmetries and 8 distortions. Pathology
confirmed 9 cancers and 51 benign concordant lesions. Four MQSA qualified radiologists reviewed and provided a BIRADS score 3
times for each lesion: first for mammography (M)/DBT only, next with US added and third with CEDM added. Readers recorded if the
lesion enhanced, how enhancement compared to background and background parenchymal enhancement. Differences in BIRADS
ratings were compared.
RESULTS
After M/DBT and US, prior to CEDM, 173/240 (72%) ratings were classified as > BIRADS 4. After viewing CEDM, 60 of these were
re-classified as < BIRADS 3; a 35% average [range 0-59%] reduction in biopsy recommendation (p<0.05). Cancers enhanced in
32/36 (89%) ratings and 32/36 cancers were rated as BIRADS >4 before and after CEDM. Benign lesions enhanced in 77/204 (38%)
(false positives). With US 3/36 cancer and 44/204 benign were converted to BIRADS>4 and 2/36 cancer and 10/204 benign to
BIRADS <3 rating. With CEDM, 1/36 cancer and 8/204 benign were converted to BIRADS>4 and 1/36 cancer and 60/204 benign to
BIRADS<3. Hence 1/36 cancer ratings (2.7%) were adversely affected (false negative) by CEDM.
CONCLUSION
CEDM use during diagnostic evaluation of BIRADS 4A or 4B lesions may result in a significant increase in PPV with minimal impact on
cancer diagnosis rates.
CLINICAL RELEVANCE/APPLICATION

CEDM use during diagnostic evaluation of BIRADS 4A or 4B soft tissue lesions may significantly reduce the number of women
recommended for benign biopsy while missing very few cancers.
SSA02-03 Diagnostic Performance of Contrast-Enhanced Spectral Mammography for Suspicious Malignant
Microcalcifications (BI-RADS 4)
Sunday, Nov. 25 11:05AM - 11:15AM Room: E450B

Participants
Rong Long, Beijing, China (Presenter) Nothing to Disclose
Kun Cao, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Min Cao, MBBCh, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Haijiao Li, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Yingshi Sun, Beijing, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

lucky_6677@126.com
PURPOSE

To assess the diagnostic performance of contrast-enhanced spectral mammography (CESM) for evaluation of suspicious malignant
microcalcifications (BI-RADS 4) comparing with full-field digital mammography (FFDM).
METHOD AND MATERIALS

Patients with mammographic calcifications without associated mass or distortions and were originally reported as BI-RADS 4 during
Jan 2015 to Jan 2018 were retrospectively collected. Lesions that proven through pathological diagnosis either by biopsy or
operation were included in the study and grouped as FFDM or CESM according to the examination they received. The
microcalcification morphology and associated enhancement (CESM group) were reviewed by two radiologists to analyse the
accuracy of the diagnosis . Diagnostic accuracy was assessed respectively for FFDM and CESM versus the results of pathology.
Statistical differences of the two methods were compared using Chi-square test.
RESULTS

48 lesions (13 malignant and 35 benign) from 48 patients were enrolled in FFDM group, and 31 lesions (10 malignant and 21 benign)
from 30 patients were in CESM group. The diagnostic sensitivity, specificity, positive predictive value, negative predictive value and
accuracy were 92.3%, 42.9%, 37.5%, 93.8%and 56.3% for FFDM group, and were 100%, 71.4%, 62.5%, 100% and 80.6% for
CESM group, respectively. The specificity and accuracy of CESM were significantly higher than that of FFDM (p<0.05). All 10
cancers including 8 DCIS in CESM group were judged as enhancement (table 1).
CONCLUSION
Comparing with FFDM, CESM improve the diagnostic performance on BI-RADS 4 mammographic calcifications, especially on
specificity and overall accuracy. The detectability of all DCIS lesions in this small cohort may validate its potential use in previously
"calcification only" disease, but still need further large sample to confirm.
CLINICAL RELEVANCE/APPLICATION

CESM improve the diagnostic performance on BI-RADS 4 mammographic calcifications, and decrease unnecessary biopsies.
SSA02-04 Participants
 Quantitative Objective Evaluation of Contrast-Enhanced Spectral Mammogram in Predicting Response
to Neo-Adjuvant Chemotherapy: A Comparative Study with RECIST 1.1 and Combined Evaluation
Methods
Sunday, Nov. 25 11:15AM - 11:25AM Room: E450B

Amr F. Moustafa, MD, Cairo, Egypt (Presenter) Nothing to Disclose

Rasha M. Kamal, MD, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Mohammed M. Gomaa, MD, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Shaimaa Mostafa, Giza, Egypt (Abstract Co-Author) Nothing to Disclose
Roaa Mubarak, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Mohamed El-Adawy, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Ahmed Abdel-Latif, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Amany M. Helal, MD,PhD, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Mona A. Sakr, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

amrfaroukmoustafa@cu.edu.eg
PURPOSE

Initiating a new objective quantitative tool for evaluation of residual disease after neoadjuvant chemotherapy using CESM in
comparison to RECIST 1.1 and combind evaluation methods.
METHOD AND MATERIALS

The study was approved by the ethical commitee of a multidisciplinary breast cancer hospital. It included 42 patients scheduled for
receiving NAC. They underwent 2 CESM examinations; prior to and after NAC and maximum10 days prior to surgery. All patients
were assessed using the RECIST 1.1 criteria, a combined approach (RECIST+ qualitative subjective assessment) and a new
quantitaive approach using an image analysis software (MATLAB and Simulink, Release 2013b). The technique consists of 3 main
steps: 1-preprocessing 2. extracting the region of Interest (ROI) and 3- Assessing the response to chemotherapy depending on the
analysis of the tumour number of pixels included within the ROI. The difference in the intensity of enhancmenet between the pre
and post NAC enahncement is calculated and compared between the 3 assessment methods in correlation to postoperative
pathology using the Miller-Payne grading. For statistical evaluation, patients were classified into responders and non-responders.
RESULTS

The calculated correlation coefficient when comparing the residual disease on CESM and Miller payne grade using RECIST 1.1, the
combined approach and the proposd quantitative method was 0.59, 0.89 and 0.69 respectively. According to Miller Payne grading
39/42 cases were classified as responders (Miller payne III, IV, and V). Using the new quantitive approach all 39/39 cases (100%)
were considered responders in comparison to 38/39 using the combined approach and 34/39 using the RECIST 1.1 criteria. The
calculated sensitivity, positive and negative predictive values of the quantitative objective evaluation (100, 97.5,100 %
respectively) was higher than the RECIST method (87.2%, 97.1% 28.6%) and the combined response method (97.4%,97.4% and
66.7%).
CONCLUSION

Quantitative objective analysis of CESM allows accurate objective evaluation of the response of breast cancer to chemotherapy
and evaluation of residual tumor prior to surgery.
CLINICAL RELEVANCE/APPLICATION
Objective analysis of CESM is an accurate tool for evaluation of the response of breast cancer post neo-adjuvant chemotherapy
and is recommended as part of pre-operative work up
SSA02-05 Diagnostic Value of Contrast-Enhanced Spectral Mammography in Comparison to MRI in a Population
of Breast Lesions
Sunday, Nov. 25 11:25AM - 11:35AM Room: E450B

Participants
Dong Xing, MBBS, Yantai, China (Abstract Co-Author) Nothing to Disclose
Xiaoxiao Chi I, Yantai, China (Presenter) Nothing to Disclose
Jianjun Dong, Yantai, China (Abstract Co-Author) Nothing to Disclose
Haizhu Xie, Yantai, China (Abstract Co-Author) Nothing to Disclose
Amiee Chen, Shanghai, China (Abstract Co-Author) Nothing to Disclose
Huizhi Cao, PhD, Beijing, China (Abstract Co-Author) Nothing to Disclose
PURPOSE

To evaluate the diagnostic value between contrast-enhanced spectral mammography (CESM) and breast magnetic resonance
imaging (MRI).
METHOD AND MATERIALS

Between July 2017 and February 2018, 235 patients who were suspected of breast abnormalities by clinical examination or
mammography were underwent CESM and MRI examination. The image of CESM and MRI and the pathological specimens were
analyzed. All lesions were evaluated independently by three experienced radiologists. Using histopathological results as the gold
standard, the diagnostic performance of CESM and MRI were investigated. The areas under ROC curves was applied to analyze
diagnostic efficiency. The data on maximum tumor size measurements were gathered on CESM and MRI. The Pearson's correlation
coefficients and 95% confidents intervals between CESM vs. pathology and MRI vs. pathology were calculated.
RESULTS

263 breast lesions were found in 235 patients, in which 177 were malignant and 86 were benign. By evaluating the diagnostic value,
the sensitivity, positive prediction value, negative predictive value, and false-negative from CESM examination was comparable to
that from MRI (91.5%,94.7%,83.7%,8.5% versus 91.5%,90.5%,82.1%,8.5%). Importantly, the accuracy and the specificity were
higher for CESM than that for MRI (81%,89.5% Vs. 80.2%,71.7%) while the the false-positive was lower(10.5% Vs. 19.8%). The
areas under ROC curves of CESM and MRI were 0.950 and 0.939, displaying the equivalent diagnostic efficiency (p=0.48). For the
agreement between measurements, mean tumor size was 3.1 (range 0-16) cm for CESM and 3.4 (range0-17) cm for MRI compared
with 3.2 (range 0-16) cm on histopathological results, the average difference of diameters between CESM, MRI and Histopathologic
size was -0.01, -0.05cm, respectively, with 95% consistency interval range of -0.34 to 0.31, -0.87 to 0.22cm, respectively. The
Pearson's correlation coefficients of CESM versus histopathology (r=0.774, p=0.000) was consistent with MRI (r=0.771, p=0.000).
CONCLUSION
Our results show better accuracy, specificity and the lower false-positive of CESM in breast cancer detection than MRI. CESM
displayed a good correlation with histopathology in assessing the lesion size of breast cancer, which is consistence with MRI.
CLINICAL RELEVANCE/APPLICATION
CESM provides additional enhancement information for diagnosing breast lesions and measuring cancer sizes with high correlation to
surgicohistology.
SSA02-06 Usefulness of Low-Dose Perfusion Breast CT: Quantification of Tumor Vascularity and Prediction of
Histologic Biomarkers in Invasive Breast Cancer
Sunday, Nov. 25 11:35AM - 11:45AM Room: E450B

Participants
Eun Kyung Park, MD,PhD, Ansan, Korea, Republic Of (Presenter) Nothing to Disclose
Bo Kyoung Seo, MD, PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation;
Research Grant, Guerbet SA; Research Grant, Koninklijke Philips NV;
Myoung-Ae Kwon, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Chang Sub Ko, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jaehyung Cha, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Kyu Ran Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ok Hee Woo, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

seoboky@korea.ac.kr
PURPOSE
To investigate the usefulness of low-dose perfusion breast computed tomography (CT) for quantification of tumor vascularity and
for prediction of histologic biomarkers in invasive breast cancer.
METHOD AND MATERIALS
This prospective study was approved by IRB with informed consent. A total of 139 patients with invasive breast cancers were
enrolled. Low-dose perfusion CT was performed in the prone position with a spectral CT (iQon, Philips Healthcare) after contrast
injection (Xenetix350, Guerbet). Effective dose was less than 1.2 mSv. Perfusion parameters were measured using a Philips
Advanced Perfusion and Permeability application prototype in breast cancers, normal breast tissue, and fat; peak enhancement
intensity (HU), perfusion on deconvolution model (mL/min/100/g), mean transit time (sec), time to peak (sec), blood volume
(mL/100/g), permeability (mL/min/100/g), and blood volume permeability on Patlak model (mL/100/g). CT perfusion parameters of
cancers and normal tissue or fat were compared using Mann-Whitney test. Correlation analysis was performed between CT
perfusion parameters of cancers and histologic biomarkers including tumor grade, estrogen receptor (ER), progesterone receptor
(PR), human epidermal growth factor receptor 2 (HER2), and Ki67 using Mann-Whitney or Kruskal-Wallis test.
RESULTS
In breast cancers, peak enhancement intensity, perfusion, blood volume, permeability, and blood volume permeability were
significantly higher, and mean transit time, time to peak were shorter than those values in normal glandular tissues and fat (P<.001
for all). Peak enhancement intensity significantly increased in cancers with ER-, PR-, HER2+, Ki67+ or more than 20 mm (P<.05 for
all). Time to peak decreased in cancers with ER-, PR-, HER2+, Ki67+, high grade, or more than 20 mm (P<.05 for all). Blood volume
permeability increased in cancers with ER-, PR-, Ki67+, or high grade (P<.05 for all). HER2-enriched cancers showed higher peak
enhancement intensity and blood volume permeability than luminal type cancers (P<.02 for all).
CONCLUSION

Low-dose perfusion breast CT can be useful in quantifying tumor vascularity and predicting prognostic biomarkers of invasive breast
cancer.
CLINICAL RELEVANCE/APPLICATION

Low-dose perfusion breast CT can be used to quantify tumor vascularity and to predict biomarkers of invasive breast cancer and
for patients who have difficulty with magnetic resonance imaging.
SSA02-07 Contrast-Enhanced Cone-Beam Breast-CT without Prior Non-Contrast Scan: Can We Reduce
Radiation Exposure While Maintaining Diagnostic Accuracy?
Sunday, Nov. 25 11:45AM - 11:55AM Room: E450B

Participants
Susanne Wienbeck, MD, Goettingen, Germany (Presenter) Nothing to Disclose
Uwe Fischer, MD, Goettingen, Germany (Abstract Co-Author) Nothing to Disclose
Joachim Lotz, MD, Gottingen, Germany (Abstract Co-Author) Nothing to Disclose
Johannes Uhlig, Goettingen, Germany (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
susanne.wienbeck@med.uni-goettingen.de
PURPOSE

Contrast-enhanced cone-beam breast-CT (CE-CBCT) is a novel breast imaging technique with comparably high radiation dose. The
current diagnostic standard includes one non-contrast scan (NC-CBCT) followed by intravenous contrast media injection and a
contrast-enhanced scan (CE-CBCT). Performing only the CE-CBCT scan might reduce radiation exposure. Our study aims to
evaluate whether CE-CBCT alone is comparable to combined NC + CE-CBCT regarding diagnostic accuracy while reducing radiation
exposure.
METHOD AND MATERIALS

This prospective IRB-approved study included 48 women (61 breasts, 100 lesions) with median age 57.9 years (IQR: 49-66 years)
and BI-RADS 4/5 lesions diagnosed on mammography/ultrasound in ACR density types c/d breasts. Two blinded breast radiologists
read CE-CBCT alone versus NC-CBCT + CE-CBCT in consensus. Intra-observer variability was assessed by one reader performing
independent double reading. Sensitivity, specificity and AUC were measured separately for CE-CBCT alone versus NC + CE-CBCT.
RESULTS

Of 100 lesions, 51 were rated as malignant, 6 as high risk and 43 as benign. Histopathological assessment was performed in 63
breast lesions and imaging follow-up over at least 1 year in another 37 lesions. Diagnostic accuracy for both CBCT approaches was
comparable: AUC, sensitivity and specificity showed no significant differences comparing CE-CBCT alone versus NC + CE-CBCT
(AUC: 0.84 vs. 0.83, p=0.643; sensitivity: 0.89 vs. 0.85, p=0.158; specificity: 0.73 vs. 0.76, p=0.655). Inter- and intra-observer
agreement on BI-RADS readings were excellent (ICC=0.76, ICC=0.83, respectively). Radiation dose was significantly lower for CE-
CBCT alone versus NC + CE-CBCT (median average glandular radiation dose 5.9 mGy vs. 11.7 mGy, p<0.001).
CONCLUSION
The diagnostic accuracy of CE-CBCT alone is comparable to that of combined NC + CE-CBCT in ACR type c/d breast. At the same
time, CE-CBCT significantly reduces radiation exposure to the breast. Further research is warranted to confirm these findings in a
larger and generalizable population.
CLINICAL RELEVANCE/APPLICATION

Assessment of CE-CBCT alone yields comparable diagnostic accuracy to combined NC + CE-CBCT and reduces radiation exposure by
up to 50%. Additional acquisition of NC-CBCT might therefore be unnecessary.
SSA02-08 Automatic Classification of Breast Lesions in Contrast Mammography Using Deep Learning in
Conjunction with Multimodal Information: BIRADS Lexicon Features and Raw Image Features
Sunday, Nov. 25 11:55AM - 12:05PM Room: E450B

Participants
Shaked Rose Perek, BSC, Tel Aviv, Israel (Abstract Co-Author) Nothing to Disclose
Miriam Sklair-Levy, MD, Tel -Hashomer, Israel (Presenter) Nothing to Disclose
Gali Zimmerman Moreno, PhD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Arnaldo Mayer, PhD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Anat Shalmon, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Osnat Halshtok, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Renata Faermann, MD, Tel Aviv, Israel (Abstract Co-Author) Nothing to Disclose
Michael Gotlieb, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
Yael Yagil, MD, Ramat Gan, Israel (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

mirisklair@gmail.com
PURPOSE
To assess the combined usage of BIRADS lexicon and pixel data for multimodal automatic classification of breast lesions in dual
energy contrast enhanced spectral mammography (CESM) and evaluate its potential for biopsy sparing in benign breast lesion.
METHOD AND MATERIALS
130 biopsy proven CESM breast lesions, (65 benign and 65 malignant) were manually contoured and described by the BIRADS
lexicon. BIRADS data was encoded by a binary vector for each lesion which, together with the lesion pixels, formed a multimodal
representation. A deep neural network was designed to process pixel data from its entry layer and merge it with BIRADS data in its
deepest layers to better balance between low-level pixel information and high-level BIRADS data that need to be merged. The
network was validated in a 5-folds cross-validation (CV) scheme, to tell apart benign/malignant lesions. In each fold, a different
subset of 25 lesions was used for testing, and the rest for training. This CV was conducted using 3 different configurations, in order
to assess and compare the contributions of different information modalities. (a) BIRADS-only classifier using SVM (BOC), (b) pixel-
only network (PON), (c) and the multimodal BIRADS+pixels network (MBPN).
RESULTS

The results are shown in Fig.1, where blue is benign and red is malignant. The classification score (y axis) reflects malignancy
probability. We seek a threshold, below which there are only benign lesions, so that no malignancy is missed, i.e. sensitivity=100%
(green line Fig.1). With this condition in mind, the maximal specificities (SP) are: (a) BIRADS only, SP=12%; (b) pixel-only network,
SP=37%; (c) multimodal BIRADS and pixel network (MBPN), SP=60%. This means that with MBPN we can safely spare unnecessary
biopsy for 60% of benign lesions without missing any malignancies.
CONCLUSION

This research showed that the combined usage of BIRADS data, provided by the radiologist, with pixel data extracted from CESM
strongly improves the specificity obtained for automatic lesion classification on pixel or BIRADS data alone.
CLINICAL RELEVANCE/APPLICATION

Multimodal lesion classification in CESM may significantly reduce benign breast biopsies, thus reducing cost and improving patient
experience.
SSA02-09 Preoperative Diagnosis of Metastatic Axillary Sentinel Lymph Nodes in Breast Cancer with Quantitative
Parameters Derived from Dual-Energy Spectral CT
Sunday, Nov. 25 12:05PM - 12:15PM Room: E450B

Participants
Chushan Zheng, Guangzhou, China (Presenter) Nothing to Disclose
Xiang Zhang, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jun Shen, MD, Guagnzhou, China (Abstract Co-Author) Nothing to Disclose
PURPOSE
The purpose of this study was to evaluate the diagnostic performance of gemstone spectral imaging (GSI) quantitative parameters
derived from dual-energy spectral computed tomography (DEsCT) for the preoperative diagnosis of metastatic sentinel lymph nodes
(SLNs) in patients with breast cancer.
METHOD AND MATERIALS

This prospective study was approved by the ethics committee, and all patients provided written informed consent. From June 2015
to December 2017, dual-phasic contrast-enhanced DEsCT was performed in 193 female patients with breast cancer. Quantitative
GSI and morphological parameters were compared between metastatic and non-metastatic SLNs. The quantitative parameters were
fitted to univariate and multiple logistic regression models. Their diagnostic abilities were analyzed by receiver operating
characteristic curves and compared by the McNemar test.
CONCLUSION
DEsCT can be used as a complementary means for the preoperative identification of SLN metastases in patients with breast cancer.
CLINICAL RELEVANCE/APPLICATION
The slope of the Hounsfield unit curve in venous phase derived from dual-energy spectral CT, can be used to differentiate
metastatic from non-metastatic axillary sentinel lymph nodes of breast cancer.
 VW 24

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Reading Sessions: Siemens
Healthineers Vendor Workshop
Sunday, Nov. 25 11:40AM - 12:50PM Room: Booth 5530

Participants
Luis Pina, MD, PhD, San Sebastian, Spain (Presenter) Nothing to Disclose
Program Information
Learn about the value of wide-angle Digital Breast Tomosynthesis (DBT) and Contrast Enhanced Mammography (CEM) in the daily
routine from one of our most experienced clinical experts. The differences and respective advantages of the morphological (DBT)
and functional (CEM) breast imaging methods will be discussed. This, all with the flexible assistance of our multi-modality reading
solution syngo® via and the syngo® Breast Care applications.
 VW 52

Triaging Dense Breast Patients in Clinical Practice: Hologic Vendor Workshop

Sunday, Nov. 25 12:00PM - 12:30PM Room: Booth 5524

Participants
Regina J. Hooley, MD, New Haven, CT (Presenter) Consultant, Hologic, Inc
Program Information
The Genius™ 3D Mammography™ is the only mammogram that is FDA approved as superior for women with dense breasts. Attend
this 30 minute session to learn how to triage women with dense breasts using 3D Mammography. (3D Mammography™ Technology)
Registration
https://hologicrsna.com
 BRS-SUA

Breast Sunday Poster Discussions

Sunday, Nov. 25 12:30PM - 1:00PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

FDA Discussions may include off-label uses.

Participants
Bethany L. Niell, MD,PhD, Tampa, FL (Moderator) Nothing to Disclose
Sub-Events
BR220-SD- Breast Density Classification and Follow-Up Decision Support System Using Deep Convolutional
SUA1 Models
Station #1
Participants
Sun Young Park, San Diego, CA (Presenter) Nothing to Disclose
Dustin Sargent, PhD, San Diego, CA (Abstract Co-Author) Nothing to Disclose
Yoel Shoshan, Haifa, Israel (Abstract Co-Author) Employee, IBM Corporation
David Richmond, Newton, MA (Abstract Co-Author) Senior Data Scientist, IBM Watson Health
Ella Barkan, Haifa, Israel (Abstract Co-Author) Nothing to Disclose
Simona Rabinovici-Cohen, Haifa, Israel (Abstract Co-Author) Employee, IBM Corporation
Marwan Sati, PhD, Mississauga, ON (Abstract Co-Author) Employee, IBM Corporation
For information about this presentation, contact:
sypark@us.ibm.com
PURPOSE
This paper presents two new deep learning models that achieve objective breast density classification with high accuracy. The
models are integrated into a system that automates BI-RADS breast density reporting and aids in making additional screening
recommendations for normal patients.
METHOD AND MATERIALS
We have trained two deep learning models that provide objective, accurate, and repeatable breast density classification to address
the issue of inter-physician variability. Our first deep learning model, used for automated BI-RADS breast density reporting, was
trained using four breast density labels (A-almost entirely fatty, B-scattered fibroglandular density, C-heterogeneously dense, D-
extremely dense). Our second model was trained to distinguish between the 'scattered density' and 'heterogeneously dense' classes
to aid in diagnostic follow-up decisions according to ACR guidelines. That is, the second model is a two-class classifier, combining
labels A and B, and labels C and D, into single classes. The two models were combined to produce an optimal follow-up decision.
The networks were trained on a large dataset with pre-processing and data augmentation. Performance was evaluated with five-
fold cross validation, and ROC analysis was performed.
RESULTS
Our models were trained on 6528 MG studies (26112 images) acquired between 2004 and 2016, each with four standard views (L-
CC, R-CC, L-MLO, and R-MLO). The age range of the patients was 24-93 years, with an (A, B, C, D) distribution of (24%, 52%,
20%, 4%). The breast densities in the clinical reports were used as ground truth. Our two-class model achieves training and test
AUCs of (0.98, 0.96), and our four-class model achieves per-class training and test AUCs of (0.98, 0.96), (0.92, 0.88), (0.92,
0.93), and (0.97, 0.96). These initial results outperform existing breast density classification algorithms.
CONCLUSION
We report highly accurate breast density classification using deep learning models trained and evaluated on a large dataset of MG
studies with a wide distribution of patient demographics. We are continuing this work with a multi-site clinical study and a
comparison with inter-physician variance.
CLINICAL RELEVANCE/APPLICATION
We present a diagnostic aid system for automated classification and reporting of BI-RADS density. Breast density assessment is
recommended by the ACR for proper assessment of breast cancer risk.
BR221-SD- Accuracy of the Nodal Staging in Breast Carcinoma Using 18F-FDG-PET/MRI, Comparison with
SUA2 Pathological Findings
Station #2
Participants
Eva Ferdova, MD, Plzen, Czech Republic (Abstract Co-Author) Nothing to Disclose
Jiri Ferda, MD, PhD, Plzen, Czech Republic (Presenter) Nothing to Disclose
Jan Baxa, MD, PhD, Plzen, Czech Republic (Abstract Co-Author) Nothing to Disclose
Ilona Zednikova, Plzen, Czech Republic (Abstract Co-Author) Nothing to Disclose
Ondrej Hes, MD, PhD, Plzen, Czech Republic (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ferda@fnplzen.cz
PURPOSE
to evaluate the sensitivity, specificity, positive and negative predictive value of 18F-PET/MRI in the nodal staging of the breast
carcinoma.
METHOD AND MATERIALS
53 women (age 23-72) underwent the nodal staging of the breast carcinoma using 18F-FDG PET/MRI and all of them underwent
surgery within two weeks with resection of the suspected lymph nodes including/or sentinel lymph node labeled by the99mTc-
nannocolloid application. The examination was performed after intravenous application of 2.5 MBq/kg of 18F-FDG. PET/MRI
examination consist of targeted breast imaging in prone position using dedicated 4-channel breast coil including the dynamic Gd-
enhanced study after application of gadobutrol in the dose of 0.1 mmol/kg. The finding of PET/MRI was compared with the
histological evaluation of the resected lymph nodes.
RESULTS
In 31 women, the nodal status was N0, in 17 N1, in 2 N2 and in 3 N3 respectively. Using the histological evaluations, there were
following findings: 20 true positive, 2 false negative, 22 true negative, 1 false positive lymph node staging. The sensitivity reached
0.91, specificity 0.97, positive predictive value 0.95, and negative predictive value 0.94, respectively.
CONCLUSION
Targeted 18F-FDG-PET/MRI in breast carcinoma enables the valuable assessment of nodal metastatic involvement with high
sensitivity and specificity
CLINICAL RELEVANCE/APPLICATION

To evaluate the sensitivity, specificity, positive and negative predictive value of 18F-PET/MRI in the nodal staging of the breast
carcinoma
BR222-SD- Automated Volumetric Breast Density Estimation: A Comparison with Radiologists' Qualitative
SUA3 Classification
Station #3
Participants
Yuan Tian, Beijing, China (Presenter) Nothing to Disclose
Jing Li, MD, PhD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Erni Li, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Zhang Renzhi, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Ning Guo, Beijing, China (Abstract Co-Author) Nothing to Disclose
Shunan Che, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
Chenglu Ke, Beijing, China (Abstract Co-Author) Nothing to Disclose
Xiaohong R. Yang, Rockville, MD (Abstract Co-Author) Nothing to Disclose
Gretchen Gierach, Bethesda, MD (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
tianyuanlaura@hotmail.com
PURPOSE
The purpose of our study was to assess the agreement of automated volumetric mammographic breast density with the radiologists'
classification by using the BI-RADS density category (5th edition), and to analyze the clinical-radiologic factors that may influence
the discordance between the two density measurements.
METHOD AND MATERIALS
A total of 7971 full-field digital mammograms with standard views were retrospectively analyzed. Breast density measurements
evaluated by radiologists according to BI-RADS category (5th edition) and by an automated volumetric breast density method
(VBDM) which was used to measure VBD (% breast density) and VDG (density grade) were compared. A weighted kappa value was
calculated to assess the degree of agreement among the visual and volumetric assessments of the density category and each
subject was classified into an concordance or discordance group. A number of clinical-radiologic factors including age, history of
breast surgery, indication for mammography, volumes of fibroglandular tissue or total breast and the percentage breast density
were compared between the two groups.
RESULTS
The agreement between breast density evaluations by radiologists and VDG was fair (kappa = 0.346) when using the four-grade
scale (A/B/C/D), and moderate (k value = 0.597) when using the two-grade scale (A-B/C-D). VBD showed a significant positive
correlation with visual assessment by radiologist (Spearman's p=0.691, p < 0.01), but the distribution of density category was
statistically significantly different among visual and volumetric measurements (p < 0.001). Category D was more frequently assigned
by VBDM (43.9%) than by the radiologists (17.1%). Discordant subjects were more likely to be younger (p < 0.001), had undergone
mammography for diagnostic purposes (p = 0.024), and have higher volumetric breast density (p < 0.001) compared with
concordant subjects.
CONCLUSION
More mammograms were classified as dense breast tissue using VBDM, as compared with visual assessments according to the BI-
RADS fifth edition. And age, indication for mammography, and volumetric breast density may contribute to the differences between
assessments by radiologists and by VBDM.
CLINICAL RELEVANCE/APPLICATION

Considering the significant positive correlation between VBD and radiologists' classification, the automated method may be used in
the future to evaluate the quantitative breast density data.
BR223-SD- The Many Faces of Contrast Enhancing Benign Breast Lesions in Digital Mammography
SUA4
Station #4
Participants
Eleni Gioutlaki, MD, Athens, Greece (Presenter) Nothing to Disclose
Christos Tzimas, MD, Athens, Greece (Abstract Co-Author) Nothing to Disclose
Christina Gkali, MD, Athens, Greece (Abstract Co-Author) Nothing to Disclose
Sophia Papaioannou, Athens, Greece (Abstract Co-Author) Nothing to Disclose
Eleni Feida, Athens, Greece (Abstract Co-Author) Nothing to Disclose
Athanasios N. Chalazonitis, MD, MPH, Athens, Greece (Abstract Co-Author) Nothing to Disclose
PURPOSE

CEDM is a quite new breast imaging method commercially available for clinical use.The aim of this study was to compare the degree
and type of enhancement of benign lesions with pathology results.
METHOD AND MATERIALS

Evaluation of CEDM examination is based on enhancement intensity of the lesion according to a qualitative scale categorized in 4-
steps as follows:Type -1(negative enhancement),Type 0(no enhancement),Type 1(low enhancement),Type 2(intense
enhancement).Type 0 and type -1 findings were considered to be probably benign,and type 1 and type 2 findings were considered
to be probably malignant.Within 5 years,1979 females 28 to 85 years,with suspicious finding(s) in DM,U/S,MR,underwent CEDM.29 of
these females had bilateral findings,so 2008 lesions were in total examined.
RESULTS
Of the 2008 breast lesions in 1979 patients in this study, lesions 1363(67,8%)were benign.Among benign lesions,23,2% presented
as enhancement on CEDM.More specific,317 out of 1363,had low or moderate enhancement on
CEDM(type1:171/317,type2:146/317).Also 978 benign lesions,had no enhancement on CEDM(type-
1:74/978,type0:904/978).Moreover,68 cases were non-mass like lesions which were benign.Histopathologically,the 171 benign
lesions with type 1 enhancement were:87 sclerosing adenosis,63 fibroadenomas,5 postoperative scar,3 ductal ectasia,1 paget
disease,2 intramammary lymph node,5 inflammation,5 no specific findings.Concerning,the 146 lesions with type 2
enhancement,histological proof of benign was:64 sclerosing adenosis,63 fibroadenomas,3 intramammary lymph node,4 ductal
ectasia,2 inflammation,2 abscess,8 no specific findings.According to our findings,the specificity of the method is76,7%.
CONCLUSION
Fibroadenomas,and sclerosing adenosis,were more frequently characterized by moderate or medium enhancements on CEDM.It is
very important to recognize the morphological features(as seen in DM)and hemodynamic patterns(as seen in CEDM)of benign lesions
of the breast,in order to eliminate number of false positive enhancing lesions.
CLINICAL RELEVANCE/APPLICATION

The capability of CEDM is to depict tumor angiogenesis in breast cancer and have demonstrated contrast uptake in most malignant
lesions independent of size as small as 1mm.Hence,CEDM may improve clinical decision making in breast cancer diagnosis.Breast
biopsies in lesions which had type 0 or 1 enhancement(according to our recommendation)can be avoided,because these lesions are
very likely to be benign.
BR165-ED- Hematopoietic and Connective Tissue Diseases of the Breast
SUA5
Station #5
Awards
Cum Laude
Participants
Katerina Konstantinoff, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Shani Aharon, BS, Worcester, MA (Abstract Co-Author) Nothing to Disclose
Christine O. Menias, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Catherine M. Appleton, MD, Saint Louis, MO (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc Royalties, Oxford University
Press
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
Various hematologic and connective tissue diseases that arise from dysfunction of hematopoietic stem cells can present in the
breast as primary disease or as a site of systemic involvement, and are important entities to keep in the differential diagnosis in
breast imaging. 1. Review the imaging appearance of hematopoietic and connective tissue diseases that involve the breast 2.
Review the typical presentation and workup of a patient presenting with suspected hematopoietic or connective tissue disease of
the breast 3. Review the differential diagnosis of a patient with hematopoietic disease of the breast
TABLE OF CONTENTS/OUTLINE

A) Appearance of hematopoietic and connective tissue diseases of the breast in different imaging modalities Mammography US
PET/CT MR B) Discussion of hematopoietic and connective tissue diseases of the breast and differential diagnosis of a suspected
hematopoietic lesion of the breast with case examples Benign conditions - Amyloidosis, Rosai-Dorfman, scleroderma,
dermatomyositis, diabetic mastopathy Malignant lesions - Lymphoma, leukemia, myeloid sarcoma, carcinoma C) Presentation and
workup of a patient with a hematopoietic disease of the breast Symptomatic - palpable lump, skin thickening Asymptomatic -
abnormality such as lesion or lymphadenopathy on imaging
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Christine O. Menias, MD - 2013 Honored EducatorChristine O. Menias, MD - 2014
Honored EducatorChristine O. Menias, MD - 2015 Honored EducatorChristine O. Menias, MD - 2016 Honored EducatorChristine O.
Menias, MD - 2017 Honored EducatorChristine O. Menias, MD - 2018 Honored Educator
BR166-ED- Ouch, That Hurts! Breast Trauma: What the Radiologist Needs to Know
SUA6
Station #6
Participants
Alyssa R. Goldbach, DO, Philadelphia, PA (Presenter) Nothing to Disclose
Sana Hava, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Shqiponja Hajdinaj, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Suzanne A. Pascarella, MD, Cherry Hill, NJ (Abstract Co-Author) Nothing to Disclose
Dina F. Caroline, MD, Elkins Park, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

alyssa.goldbach@tuhs.temple.edu
TEACHING POINTS

Breast Trauma is a common and often underreported entity. Accidental injury to the female breast can cause symptoms and signs
which may mimic carcinoma, including skin changes and palpable lumps. Breast trauma can also be painful or disfiguring. We will
reinforce the imaging features of breast trauma including a review of mechanism of injury when applicable and discuss appropriate
follow-up recommendations.
TABLE OF CONTENTS/OUTLINE
The primary goals are to provide a thorough review of imaging findings of breast trauma from our patient database. This will be
achieved by: Illustrating the variety of imaging features of traumatic breast lesions. Describing the appropriate management
recommendations for traumatic breast lesions. Providing a thorough review of traumatic breast pathologies with practical tips for
aiding in diagnosis. Pathologies include gunshot wound, hematoma, hematoma complicated by abscess, seat belt injury, laceration,
and pseudoaneurysm. Pathologic correlation from our patient database will be provided when available. It is important for the
radiologist to recognize the imaging features of traumatic breast injuries as they can create confusion for both the patient and the
clinician. The radiologist is a critical member of the multidisciplinary team working to make the diagnosis and direct patient
management.
BR167-ED- Inflammatory Processes of the Breast a Pictorial Review
SUA7
Station #7
Participants
Maria Soledad Nocetti, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Lucia I. Beccar Varela, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Elizabeth Quiroga, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Veronica E. Grondona, MD, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Vanina Kuznicki, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Flavia B. Sarquis, MD, Vicente Lopez, Argentina (Presenter) Nothing to Disclose
For information about this presentation, contact:

lu.beccar.varela@gmail.com
TEACHING POINTS

• Review the clinical, radiographic and histologic features of commonly encountered inflammatory and reactive breast lesions. •
Learn a diagnostic algorithm when examining women with mastitis, to distinguish between cancer-related and non cancer-related
breast inflammation, since their clinical presentation can be misleading.
TABLE OF CONTENTS/OUTLINE
Mastitis is the inflammation of breast tissue. It always manifests clinically by three cardinal signs of inflammation, which are
redness, heat and pain. From a pathophysiological point of view, mastitis reflects a variety of underlying etiologies. It can be due to
non-infectious inflammation, infection (generally of bacterial origin) but can also be caused by inflammation resulting from malignant
tumor growth. These processes include fat necrosis, mammary duct ectasia, granulomatous lobular mastitis, diabetic mastopathy,
abscess and inflammatory cáncer. The radiologist must be familiar with the radiological signs of breast inflammation, and malignacy
must be recognized and diagnosed without fail. Full use of breast imaging techniques is therefore crucial to ensure diagnosis, and
subsequently to provide the patient with the most efficient treatment.
BR168-ED- Testing the Waters: Multimodality Imaging Evaluation of Peri-Implant Effusions
SUA8
Station #8
Awards
Magna Cum Laude

Participants
Merissa Harris, MD, Dallas, TX (Presenter) Nothing to Disclose
Emily E. Knippa, MD, Durham, NC (Abstract Co-Author) Nothing to Disclose
Nicholas Haddock, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
Sumeet Teotia, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
Sunati Sahoo, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
W. Phil Evans III, MD, Dallas, TX (Abstract Co-Author) Scientific Advisory Board, VuCOMP, Inc
Stephen J. Seiler, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

merissa.harris@phhs.org
merissa.harris@phhs.org
TEACHING POINTS

There are a few, but significant, differences in the differential diagnosis of immediate/early vs. delayed peri-implant effusions.
Breast Implant Associated-Anaplastic Large Cell Lymphoma (BIA-ALCL) can be associated with either saline or silicone implants and
is most frequently seen with textured implant surfaces. When performing an aspiration of a delayed peri-implant effusion, fluid
should be sent for culture and cytologic analysis. Cytology should include Wright Giemsa staining as well as immunohistochemistry
testing for cluster of differentiation (CD) and anaplastic lymphoma kinase (ALK) markers.
TABLE OF CONTENTS/OUTLINE

Multimodality appearance of a peri-implant effusion: Mammography Ultrasound Magnetic Resonance Imaging Effusion Mimics Early
effusion differential considerations: Seroma Hematoma Infection Delayed effusion differential considerations: Non-specific
inflammation Infection Malignant effusion (due to breast cancer) Malignant effusion (due to lymphoma): BIA-ALCL BIA-ALCL:
Background Medical device report statistics Diagnostic evaluation Management considerations Procedural Considerations: Tips for
performing an ultrasound-guided aspiration
BR204-ED- Round Malignancies: A Pictorial Review with Radiologic-Pathologic Correlation
SUA10
Station #10
Participants
Janice Y. Jeon, MD, Washington, DC (Presenter) Nothing to Disclose
For information about this presentation, contact:

janice.l.jeon@gunet.georgetown.edu
TEACHING POINTS

Breast malignancies encompass a host of heterogeneous tumors with a range of appearances on multimodality imaging. The
updated BI-RADS lexicon predicts risk of malignancy based on various descriptors including shape and margins. While oval masses
confer benignity, round masses warrant evaluation to exclude malignancy. Invasive ductal carcinoma (IDC) is the most common
invasive cancer. While most radiologists are familiar with the classic appearance of poorly-differentiated (grade III) irregular IDC,
decreased familiarity with the less common round, non-aggressive appearing well-differentiated (grade I) IDC may lead to
misinterpretation and misdiagnosis. Other cancers often presenting as round masses are the well-differentiated IDC subtypes which
include mucinous, medullary, metaplastic & papillary categories. The aforementioned entities will be presented via pictorial review
with pathology correlation. This exhibit aims to increase awareness of the spectrum of round circumscribed malignancies for which
cancer should not be excluded from the differential considerations.
TABLE OF CONTENTS/OUTLINE

1. Introduction/Background 2. Multimodality (2D FFDM, 3D Tomo, US & MRI) Pictorial Review with Pathology Correlation &
Management A) Invasive Ductal Carcinoma - NOS B) Medullary C) Mucinous D) Metaplastic E) Papillary 3. Take-home Points &
Pearls/Pitfalls
BR013-EB- Beware of the Rare: A Pictorial Review of Unusual Breast Lesions
SUA
Hardcopy Backboard
Participants
Denny Lara Nunez, MD, Mexico City, Mexico (Presenter) Nothing to Disclose
Fernando Candanedo Gonzalez, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Monica Chapa, MD, Mexico, Mexico (Abstract Co-Author) Nothing to Disclose
Mariana Licano, MD, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Antonio Hernandez Villegas, MD, Mexico City, Mexico (Abstract Co-Author) Nothing to Disclose
Nancy Margarita Gutierrez Castaneda, MD, Cuautla, Mexico (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

dennylaran5@gmail.com
TEACHING POINTS

1. Recognize and familiarize with breast lesions that are not seen routinely on mammographic practice. 2. Identify mutlimodality
imaging characteristics of benign and malignant unusual breast lesions. 3. Describe the histopathologic characteristics of rare
tumors involving the breast and make correlation with imaging findings for an accurate diagnosis and management.
TABLE OF CONTENTS/OUTLINE

1. Introduction -General features of breast cancer -Epidemiology 2. A brief review including: -Epidemiology -General features -
Clinical presentation -Multimodality imaging characteristics -Histopathologic analysis of the following lesions will be presented: -
Sarcoma (angiosarcoma and fusocellular) -Primary neuroendocrine tumor -Lymphoma -Metastases from extramammary malignancies
-Fibroadenolipoma -Phyllodes benign tumor, among others 3. Conclusions.
BR008-EB- Where is the Lesion? Mammographic-Sonographic and Breast MR Imaging Correlation
SUA
Hardcopy Backboard
Participants
Jin Hee Moon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Young Joo Won, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Ji-Young Hwang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ji Young Woo, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
geenie-m@hanmail.net
TEACHING POINTS

Breast has very different position according to imaging modality. Lesion detection and accurate correlation in multimodality breast
imaging is very important for preventing false or delayed diagnosis. We will provide process for correlation of lesions in
mammography, ultrasound and breast MRI and show pearls and pitfalls through illustrations and sample cases.
TABLE OF CONTENTS/OUTLINE

1. Introduction 2. Basic concept of triangulation 3. Mammographic-sonographic correlation 4. Sonographic-MRI correlation 5.

Targeted or second look sonography technique 6. MRI imaging navigated sonography system 7. Summary
 VW 01

Comparison between ABUS and Hand-Held Breast Ultrasound in the Screening of Dense Breasts: GE Vendor
Workshop
Sunday, Nov. 25 12:30PM - 1:00PM Room: Booth 8156

Participants
Fernanda P. Pereira, MD, PhD, Rio de Janeiro, Brazil (Presenter) Nothing to Disclose
Program Information
In this presentation, Dr. Fernanda will discuss her results from over 250 patients comparing ABUS (acquired by a trained person,
read by a breast radiologist) with Hand-Held (fully completed by a radiologist) and the impact to cancer detection, recall rate and
false positives. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in
this session. Click the link below to register.
Registration

http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 BRS-SUB

Breast Sunday Poster Discussions

Sunday, Nov. 25 1:00PM - 1:30PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

Participants
Bethany L. Niell, MD,PhD, Tampa, FL (Moderator) Nothing to Disclose
Sub-Events
BR224-SD- Evaluation of Gd-Deposits in Healthy Women Participating in High Risk Screening Program for Early
SUB1 Breast Cancer Detection
Station #1
Participants
Barbara Bennani-Baiti, MD, Wien, Austria (Presenter) Nothing to Disclose
Barbara Krug, Cologne, Germany (Abstract Co-Author) Nothing to Disclose
Daniel Giese, Koln, Germany (Abstract Co-Author) Nothing to Disclose
Kilian Weiss, PhD, Hamburg, Germany (Abstract Co-Author) Employee, Koninklijke Philips NV
Thomas H. Helbich, MD, Vienna, Austria (Abstract Co-Author) Research Grant, Medicor, Inc Research Grant, Siemens AG Research
Grant, C. R. Bard, Inc
Pascal A. Baltzer, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
PURPOSE
To determine whether patients at high risk to develop breast cancer, that undergo contrast enhanced breast MRI at regular
intervals for early breast cancer detection, exhibit brain signal alterations in the dentate nucleus and globus pallidus.
METHOD AND MATERIALS
In this IRB-approved, dual centre randomized, prospective study 73 healthy women with no history of cancer or neurological
disease that had received at least 6 doses of macrocyclic Gadolinium based contrast agents in the course of a national high risk
screening program for the early detection of breast cancer were included. Patients underwent 3T/1.5T MRI of the brain, with a
dedicated head coil, including T1 mapping and mpRAGE sequences. T1 times and T1 signal intensities were measured for dentate
nucleus (ND), pons, globus pallidus (GP), and crus posterior of the capsula interna (CP) bilaterally, employing Horos software. Ratios
of GP to CP as well as ND to pons were calculated for respective signal intensities and further statistical analyses were carried out
with SPSS and Medcalc including multivariate regression analysis.
RESULTS
There were 73 participants (median age 46 +/- 9 years) that had received an average of 9 cumulative dosages of Gd based
macrocyclic contrast agents (median 8, SD 3, range 6 - 23). Spearman's rank correlation coefficient analysis revealed a mild
correlation between age and number of dosages (R 0.31, p < 0.01) but no statistically significant correlations were found for signal
intensity ratios or T1 times in relation to age or number of dosages. T1 times displayed no significant differences between the
analysed brain structures (Fig.1.). ANOVA testing reveiled an adjusted R2 of -0.026 and -0.004 for the number of cumulative
dosages predicting T1 times and signal intensity ratios, respectively, confirming that the number of previous dosages did not affect
T1 signal in globus pallidus or nucleaus dentatus.
CONCLUSION
Neither nucleus dentatus nor globus pallidus display altered T1 signals after high cumulative dosages of macrocyclic Gd-based
contrast agents in healthy women.
CLINICAL RELEVANCE/APPLICATION
Breast MRI being the most sensitive method for breast cancer detection crucially relies on Gd-based contrast agents. These
findings show that the currently employed macrocyclic Gd-based contrast agents do not result in Gd-deposits in the brain of
healthy women participating in a high risk screening program for early breast cancer detection.
BR225-SD- Radiomics Analysis with Contrast-Enhanced Dual-Energy Mammography for the Differentiation of
SUB2 Hormone Receptor Status and Tumor Invasiveness in Breast Cancer Patients
Station #2
Participants
maria adele marino, MD, New York,, NY (Presenter) Nothing to Disclose
Doris Leithner, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Katja Pinker-Domenig, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Janice S. Sung, MD, New York, NY (Abstract Co-Author) Research Grant, Hologic, Inc
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Maxine S. Jochelson, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
marinom@mskcc.org
marinom@mskcc.org
PURPOSE
To investigate the potential of radiomics analysis of contrast-enhanced dual-energy mammography (CEDM) for differentiation of
breast cancer invasiveness and hormonal status.
METHOD AND MATERIALS
In this IRB-approved retrospective data analysis, 53 patients with proven breast cancers were included and underwent pre-
treatment CEDM. There were 58 lesions in 53 patients: 6 non-invasive (ductal carcinoma in situ) and 52 invasive breast cancers.
Among the invasive cancers, 39 were hormone-receptor positive (HR+) and 12 hormone-receptor negative (HR-) Radiomics analysis
was performed using MaZda software (Technical University of Lodz, Poland). Lesions were manually segmented and radiomic
features were derived from the first-order histogram (HIS): co-occurrence matrix (COM), run-length matrix (RLM), absolute gradient
(GRA), autoregressive model (ARM), the discrete Haar wavelet transform (WAV), as well as lesion geometry (GEO). Fisher,
probability of error and average correlation (POE+ACC), and mutual information (MI) coefficients were used for feature selection.
Linear discriminant analysis followed by k-nearest neighbor classification (with leave-one-out cross-validation) was used for
pairwise texture-based separation of subtypes/hormonal status. Radiomics parameters were correlated with tumor histology
(invasive vs non-invasive) and hormonal status (HR+ vs HR-).
RESULTS
Texture analysis yielded the following accuracies for differentiation between invasive/non-invasive breast cancers: 86.2% (based
on COM and WAV features/MI), 84.4% (based on COM, WAV, and RLM features/POE+ACC), and 82.8% (mainly through
COM+WAV/Fisher). For differentiation between HR+ and HR_ cancers, diagnostic accuracy was as follows: 76.4% (based on COM
features/MI), 72.5% (based on COM+HIS+WAV features/POE+ACC), and 66.6% (mainly through COM features/Fisher).
CONCLUSION

Data indicate that radiomics analysis with CEDM might be able to differentiate between non-invasive and invasive breast cancer
and hormone receptor status.
CLINICAL RELEVANCE/APPLICATION
Radiomics analysis with CEDM might provide additional imaging biomarkers for the non-invasive characterization of breast cancer
and thus has the potential to guide treatment decisions.
BR226-SD- Patient, Provider, and Facility Characteristics Associated with Observed Variations in Breast Cancer
SUB3 Screening Outcomes: Findings from A Learning Health System
Station #3
Participants
Nila H. Alsheik, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Firas Dabbous, Park Ridge, IL (Abstract Co-Author) Nothing to Disclose
Zhaohui Su, PhD, Boston, MA (Abstract Co-Author) Employee, OM1
Gregory Donadio, Park Ridge, IL (Abstract Co-Author) Employee, OM1
Richard Gliklich, MD, Cambridge, MA (Abstract Co-Author) Employee, OM1
Scott Pohlman, MSc, BEng, Cambridge, MA (Abstract Co-Author) Employee, Hologic, Inc
Kathleen Troeger, Marlborough, MA (Abstract Co-Author) Employee, Hologic, Inc
Vandana Menon, MD,PhD, Cambridge, MA (Abstract Co-Author) Employee, OM1
Emily F. Conant, MD, Philadelphia, PA (Presenter) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc; Consultant, iCAD,
Inc; Speaker, iiCME
For information about this presentation, contact:

nila.alsheik@advocatehealth.com
PURPOSE

To examine data from a learning health system to identify patient, provider, and imaging facility characteristics associated with
optimal screening outcomes.
METHOD AND MATERIALS
A big data platform was used to integrate EMR, RIS, and tumor registry data to create a learning health system. The analysis
included 411,355 screens, performed 2015 to 2017, from 64 imaging facilities across three large healthcare organizations. The
imaging facilities were stratified into quartiles of increasing recall rates (RR) (Quartile 1 (Q1) range=6.78-8.37%, Q2=8.84-9.98%,
Q3=10.09-10.81%, Q4=10.88-21.00%) and by digital breast tomosynthesis (DBT) conversion (75-100% DBT, 50-75% DBT, 25-50%
DBT, 0-25% DBT). Patient, provider and site characteristics, PPV1, biopsy (BX) rate and cancer detection (CDR) were evaluated
overall and stratified by modality.
RESULTS

There were no consistent trends in distribution of age, ethnicity or breast density across quartiles but facilities in the lower recall
quartiles had more Caucasian patients and were predominantly DBT. Facilities with higher DBT conversion had lower RR than hybrid
or predominantly 2D digital mammography (DM) sites (75-100% DBT: 8.87%, 50-75% DBT: 10.72%, 25-50% DBT:11.39%, 0-25%:
10.11%, p value for trend <0.001). Screening outcomes varied between quartiles. PPV1 decreased from 6.0% in Q1 to 3.3% in Q4;
BX rate and CDR were lowest in Q1 (1.0%, 4.0), Q2 had the highest BX rate and CDR (1.7%, 4.7) and Q4 had a high BX rate and
low CDR (1.7%, 4.2). DBT had higher PPV1, and higher CDR, in Q1 (6.7% vs 3.3%; 4.4 vs 2.6), Q2 (6.1% vs 3.6%; 5.5 vs 3.5) and
Q3 (5.1% vs 3.8%; 5.3 vs 3.8) compared to DM, but not in Q4 (3.2% vs 3.3%; 4.2 vs 4.2).
CONCLUSION

Recall rates were influenced by race and DBT conversion. Across quartiles, DBT had lower RR and higher PPV1 and CDR, except in
Q4. Facilities that were predominantly DBT had consistently lower RR than hybrid DBT/DM sites. These data demonstrate that DBT
offers a more efficient screening option and increasing DBT utilization is associated with improved outcomes. There is an optimal
balance between RR, BX rate and CDR and, as RR increased above 9.18%, there were increased BX but no significant gains in either
PPV1 or CDR.
CLINICAL RELEVANCE/APPLICATION
Digital Breast Tomosynthesis in hybrid sites, and sites with higher conversion to DBT, have better screening outcomes than 2D
Digital Mammography, across all strata of age, race and breast density.
BR227-SD- Breast Imaging (Contrast Enhanced Spectral Mammography)
SUB4
Station #4
Participants
Norran H. Said, MD, FRCR, Cairo, Egypt (Presenter) Nothing to Disclose
Ashraf Selim, MD, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

norranhussein@yahoo.com
PURPOSE

To identify the typical CEDM imaging characteristics and morphological enhancement patterns of different breast cancer
pathological subtypes.
METHOD AND MATERIALS

From Jan 2016 to Jan 2018, 147 patients with pathology proved breast cancer who had undergone CEDM were retrospectively
reviewed. Imaging findings were recorded and divided into mass, non mass enhancement & focus. Morphological enhancement
criteria for mass and non mass enhancement were interpreted as follows; mass enhancement was described by shape (round, oval,
irregular), margin (circumscribed, noncircumscribed : irregular / spiculated), and uptake pattern (rim, homogenous , heterogenous).
Non mass enhancement was described by uptake pattern (homogenous, heterogenous, clumped, clustered ring) and distribution
(focal, linear, segmental, regional, multiple regions, diffuse). All descriptors were correlated with pathology including subtypes;
Triple negative (TN), Estrogen receptor (ER) positive, Human epidermal growth factor receptor 2 (Her 2) positive.
RESULTS

Out of the 147 cases, there were 24 (16.3%) TN, 90 (61.2%) ER positive and 33 (22.4%) Her 2 positive. Enhancement patterns
according to subtype were as follows; TN cases demonstrated 79.2% mass and 20.8% non mass enhancement. ER positive cases
included 72.2% mass, 22.2% non mass enhancement and 5.6% foci. Her 2 positive included 45.5% mass, 48.5 %non mass
enhancement and 6.1% foci. TN tumors demonstrated 68.4% rim enhancement with 73.3% circumscribed margins, and 68.4% round
shape. ER positive demonstrated 78.5% heterogenous enhancement , 63% spiculated margins, and 91% irregular shape. Her 2
positive demonstrated 53.3% heterogenous enhancement with spiculated margins, and 100% irregular shape. P value was found
significant (p<0.001 ). There were no significant differences on the non mass enhancement descriptors and foci.
CONCLUSION

Our findings conclude that the rim enhancement feature on CEDM, along with circumscribed margins and round shape, are
significant morphological descriptors for TN cancers compared to other subtypes. No significant difference was detected in non
mass and focus uptake patterns.
CLINICAL RELEVANCE/APPLICATION

Application of contrast enhancement characteristics of masses by CEDM, allow detailed estimation of pathological subtypes with
radiological-pathological correlation.
BR228-SD- Does Tomosynthesis Work For Everyone?
SUB5
Station #5
Participants
Ethan O. Cohen, MD, Houston, TX (Presenter) Spouse, Consultant, Medtronic plc; Spouse, Consultant, Novo Nordisk AS; Spouse,
Consultant, Eli Lilly and Company; Spouse, Consultant, AstraZeneca PLC
Rachel E. Perry, MD, Birmingham, AL (Abstract Co-Author) Nothing to Disclose
Ashmitha Srinivasan, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Hilda H. Tso, DO, Houston, TX (Abstract Co-Author) Nothing to Disclose
Kanchan Phalak, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Michele D. Lesslie, DO, Bellaire, TX (Abstract Co-Author) Nothing to Disclose
Karen E. Gerlach, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Jessica W. Leung, MD, Houston, TX (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc;
Speakers Bureau, FUJIFILM Holdings Corporation
For information about this presentation, contact:

ecohen@mdanderson.org
PURPOSE

To compare the performance of full-field digital screening mammograms (FFDM) with and without digital breast tomosynthesis (DBT)
in women with and without breast implants.
METHOD AND MATERIALS

An IRB-approved, HIPAA-compliant retrospective review was performed of 103,070 consecutive screening mammograms obtained
from February 2011 through June 2014. Recall rates (RRs), cancer detection rates (CDRs), and positive predictive values for recall
(PPV1s) were analyzed.
RESULTS

The following data compare FFDM and FFDM-DBT: 67,331 FFDM and 28,835 FFDM-DBT from women without implants yielded RRs of
8.0% and 6.3%, respectively (p<0.00001); CDRs of 4.1 and 5.0 per 1000 exams, respectively (p=0.07); and PPV1s of 5.1% and
8.0%, respectively (p<0.0001). 4325 FFDM and 2579 FFDM-DBT from women with implants yielded RRs of 5.2% and 4.1%,
respectively (p=0.040); CDRs of 1.8 and 2.7 per 1000 exams, respectively (p=0.46); and PPV1s of 3.6% and 6.7%, respectively
(p=0.25). The same data is also used to evaluate the effect of implants on screening: 67,331 FFDM without implants and 4325
FFDM with implants yielded RRs of 8.0% and 5.2%, respectively (p<0.00001); CDRs of 4.1 and 1.8 per 1000 exams, respectively
(p<0.00001); and PPV1s of 5.1% and 3.6%, respectively (p=0.30). 28,835 FFDM-DBT without implants and 2579 FFDM-DBT with
implants yielded RRs of 6.3 and 4.1, respectively (p<0.00001); CDRs of 5.0 and 2.7, respectively (p=0.11); and PPV1s of 8.0 and
6.7, respectively (p=0.63).
CONCLUSION
Tomosynthesis improves the performance of digital screening mammography, while the presence of implants reduces its
performance. Specifically, tomosynthesis improved RRs, CDRs, and PPV1s for all women (with and without implants), though
statistically significant differences were seen only for RRs in women without implants, RRs in women with implants, and PPV1s for
women without implants. Implants were associated with decreased RRs, worse CDRs, and worse PPV1s for all screening exams
(FFDM and FFDM-DBT), but statistically significant differences were seen only for RRs for all screening exams and CDR for FFDM.
Further study with larger populations is warranted.
CLINICAL RELEVANCE/APPLICATION

The benefit of tomosynthesis has been incompletely studied in screening mammography patients with implants. This research
suggests that tomosynthesis is useful for screening women with implants in addition to those without implants.
BR170-ED- Imaging of Free Liquid Silicone Injections in the Transgender Breast
SUB6
Station #6
Participants
Emily B. Sonnenblick, MD, New York, NY (Presenter) Nothing to Disclose
Karen A. Lee, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Shabnam Jaffer, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Zil Goldstein, MA, New York, NY (Abstract Co-Author) Nothing to Disclose
Nishi Talati, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Laurie R. Margolies, MD, New York, NY (Abstract Co-Author) Research Consultant, FUJIFILM Holdings Corporation
Susan Boolbol, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Jess Ting, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
emily.sonnenblick@mountsinai.org
TEACHING POINTS

1. Review considerations in breast imaging unique to transgender patients on hormone treatment (HT). 2. Free liquid silicone (FLS)
injections, while not always symptomatic, may cause pain, palpable masses and disfigurement prompting some patients to seek
medical care. 3. FLS breast injections present a challenge in the setting of diagnostic imaging or cancer screening. 4. Specific
patterns and distribution of FLS seen with mammography, ultrasound and MRI correlate with surgical pathology and histology.
TABLE OF CONTENTS/OUTLINE

1. Background • Prevalence, guidelines for transgender care • Expected radiographic and histology correlates of HT for transgender
individuals • Historical review of liquid silicone • Radiographic aspects and complications of FLS used for breast augmentation 2. Pre-
mastectomy evaluation • Role of pre-op imaging/screening for cancer with consideration of trans population breast cancer risk •
Distribution of silicone: superficial location, pectoralis involvement • Fibrotic masses masquerading as cancer • Diffuse silicone
granulomas 3. Diagnostic evaluation • Palpable silicone granulomas • Bloody nipple discharge • Post-op peri-prosthetic collection •
Free silicone on CT • Post-op residual silicone 4. Silicone may migrate, make calcified granulomas and fibrotic masses, obscure
normal tissue, complicating diagnosis of malignancy.
BR171-ED- Breast Implants for Residents Test Your Knowledge in Common and Uncommon Findings
SUB67
Station #7
Participants
Lucia I. Beccar Varela, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Soledad Nocetti, MD, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Veronica E. Grondona, MD, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Elizabeth Quiroga, Vicente Lopez, Argentina (Abstract Co-Author) Nothing to Disclose
Flavia B. Sarquis, MD, Vicente Lopez, Argentina (Presenter) Nothing to Disclose
For information about this presentation, contact:

lu.beccar.varela@gmail.com
TEACHING POINTS
• Learn through pictures and diagrams to recognize different types of breast implants, including saline, double lumen and silicone
gel. • Utilizing clinical cases, analyze the multimodality imaging appearances of common and uncommon complications, including
mammographic, ultrasonographic and MRI findings.
TABLE OF CONTENTS/OUTLINE

An increasing number of patients have breast implants for cosmetic augmentation of the breast, reconstruction after mastectomy
or correction of congenital malformations. Any radiologist who interprets breast imaging studies should be familiar with the normal
and abnormal findings of common implants. Clinical diagnosis of implant rupture is difficult and the aim of imaging is to provide
essential information about tissue and prothesis integrity and detect breast diseases unrelated to implants, such as breast cancer.
This work is a resident primer on successfully recognizing common and uncommon findings related to breast implants, learning
specific strengths and weaknesses of each imaging technique, to make the correct choice for each individual patient.
Participants
BR172-ED- Law and Order: Staying aBREAST of Medicolegal Issues in Mammography
SUB8
Station #8
Michael S. Morrow, DO, Worcester, MA (Abstract Co-Author) Nothing to Disclose
Gopal R. Vijayaraghavan, MD, MPH, Shrewsbury, MA (Presenter) Nothing to Disclose
Suma C. Kannabiran, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
David E. March, MD, Springfield, MA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Review current malpractice trends in Radiology with an emphasis on breast imaging. Highlight pitfalls in mammography that
commonly lead to false negatives. Present best practices/policies regarding disclosing error to patients.
TABLE OF CONTENTS/OUTLINE

Review of recent general trends in medical malpractice across specialties Detail malpractice as it relates to breast imaging Highlight
types of errors in mammography interpretation Present common pitfalls in mammography leading to false negatives Introduce best
practices in mitigating errors in an effort to minimize litigation
BR173-ED- Breast Elastography: How Does It Help? Interactive Case Review Including Helpful Tips for
SUB9 Incorporating Elastography in Your Practice
Station #9
Participants
Ana DiPrete, BA, Providence, RI (Presenter) Nothing to Disclose
Robert C. Ward, MD, Providence, RI (Abstract Co-Author) Nothing to Disclose
Elizabeth Lazarus, MD, Barrington, RI (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ana_diprete@brown.edu
TEACHING POINTS

-Breast elastography utilizes static and dynamic ultrasound techniques to provide information about the "stiffness" of a lesion.-As
malignant lesions tend to be stiffer than benign lesions, qualitative and quantitative assessments are used to score lesions as soft,
intermediate or hard.-Adding elastographic qualities to BI-RADs features improves diagnostic accuracy and aids decision-making
regarding clinical management (i.e. biopsies), especially with BIRADs category 3 and 4a lesions.-Our objective is to provide an
engaging case-based imaging approach with questions and answers to discuss basic principles of elastography, commonly applied
techniques, and the elastographic appearance of benign and malignant lesions to emphasize the implications of adding elastography
to the diagnostic evaluation of breast disease.
TABLE OF CONTENTS/OUTLINE

Users will be faced with diagnostic scenarios and management questions regarding the following areas:1. Comparing strain and
shear wave elastography techniques2. Illustrating how elastographic features of benign and malignant breast disease differ3.
Highlighting the clinical application of elastography and how it improves diagnostic accuracy of breast ultrasound 4. Reviewing the
literature of breast elastography with a focus on how it supplements screening and diagnostic grey scale breast ultrasound
BR174-ED- New Findings on Breast MRI with Benign or Probably Benign Characteristics: Evaluation, Differential,
SUB10 and Management Dilemmas
Station #10
Participants
Sonia P. Sahu, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Sona A. Chikarmane, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Emily Mungovan, BA, Boston, MA (Abstract Co-Author) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
Mirelys Barrios, MD , Boston, MA (Presenter) Nothing to Disclose
For information about this presentation, contact:

schikarmane@bwh.harvard.edu
TEACHING POINTS

1. To understand the criteria for BIRADS 2 and BIRADS 3 lesions on screening breast MRI. 2. To recognize imaging features that
may be new on screening MRI but support a benign or probably benign assessment. 3. To avoid pitfalls in assessing new MRI
findings as benign or probably benign and become familiar with imaging features that should prompt immediate biopsy.
TABLE OF CONTENTS/OUTLINE

1. Provide a brief overview of the qualifying features for BIRADS 2 and BIRADS 3 lesions on MRI. 2. Demonstrate a rich pictorial
review of new lesions and imaging features found on screening MRI that are consistent with a BIRADS 2 or 3 assessment including
fat necrosis, asymmetric and diffuse background parenchymal enhancement, and enhancing foci. 3. Present teaching points to help
guide management of new benign and probably benign findings on screening MRI. 4. Discuss pitfalls in assessing new MRI findings as
benign or probably benign and review imaging features that should prompt immediate biopsy.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
BR175-ED- Imaging Findings Following Image Guided Cryoablation Of Breast Cancer: A New Normal
SUB11
Station #10
Participants
Kenneth R. Tomkovich, MD, Freehold, NJ (Presenter) Consultant, Scion Medical Technologies, LLC;
For information about this presentation, contact:
ktomkovich@princetonradiology.com
TEACHING POINTS

The purpose of this exhibit is:1. To introduce the concept of cryoablation as a primary treatment for certain types of breast
cancers without the need for surgical excision2. To demonstrate the techniques utilized in the ICE 3 trial for successful treatment
of low grade breast cancers with ultrasound guided cryoablation 3. To illustrate different imaging findings as a "new normal" in
patients who have undergone cryoablation as a primary treatment for breast cancer including:Mammographic findings post
cryoablation Ultrasound findings post cryoablation MRI findings post cryoablation 4. To discuss imaging findings that led to post
cryoablation biopsies and why they can be observed if encountered in the future to avoid unnecessary biopsies
TABLE OF CONTENTS/OUTLINE

Cryoablation as a primary treatment for certain types of breast cancersTechniques of ultrasound guided breast cancer cryoablation
Brief description of the ICE 3 trialReview of imaging findings post cryoablation: the "new normal"- Mammography - Ultrasound - MRI-
Biopsy examples Future implications and summary
 VW 53

A Clinical Perspective on Increasing Confidence with Synthesized 2D Imaging Technology: Hologic Vendor
Workshop
Sunday, Nov. 25 1:00PM - 2:00PM Room: Booth 5524

Participants
Jacqueline S. Holt, MD, Wilmington, DE (Presenter) Nothing to Disclose
Program Information
Clinical perspective of Synthesized 2D Imaging Technology to increase reading confidence. This session includes a facilitated review
of relevant cases.
Registration
https://hologicrsna.com
 VW 25

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Sunday, Nov. 25 1:05PM - 2:15PM Room: Booth 5530

Participants
Terri A. Gizienski, MD, Greenwood Village, CO (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 VW 02

Breast Density and Improved Risk Assessment: GE Vendor Workshop

Sunday, Nov. 25 1:30PM - 2:00PM Room: Booth 8156

Participants
Jack Cuzick, London, United Kingdom (Presenter) Nothing to Disclose
Program Information
In this talk, Professor Cuzick will review several large cohort studies demonstrating not only the univariate strength of
mammographic density, but also its near independence and added value to both questionnaire-based risk factors and single
nucleotide polymorphism (SNP) panels. The value of changes in breast density after endocrine therapy as markers of treatment
effectiveness is also briefly reviewed. Registration is required; adding this session to the RSNA calendar tool alone does not secure
your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 RC115

Screening for Breast Cancer

Sunday, Nov. 25 2:00PM - 3:30PM Room: E353C

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Jennifer A. Harvey, MD, Charlottesville, VA (Moderator) Stockholder, Hologic, Inc; Research Grant, Volpara Health Technologies
Limited; Stockholder, Volpara Health Technologies Limited;
For information about this presentation, contact:
jharvey@virginia.edu
zuleyml@upmc.edu
Sub-Events
RC115A Screening Data: Where Are We?
Participants
Debra L. Monticciolo, MD, Temple, TX (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) To review key data for breast cancer screening from randomized controlled trials and observational studies. 2) To understand
the risks and benefits of mammography screening for the woman of average risk. 3) To understand the differences between various
screening recommendations.
RC115B Risk Models
Participants
Jennifer A. Harvey, MD, Charlottesville, VA (Presenter) Stockholder, Hologic, Inc; Research Grant, Volpara Health Technologies
Limited; Stockholder, Volpara Health Technologies Limited;
For information about this presentation, contact:
jharvey@virginia.edu
LEARNING OBJECTIVES
1) Cite pros and cons of different risk models in use. 2) Describe how risk models can be used in practice. 3) List which patients at
risk may not be identified using risk models.
RC115C Personalized Screening Paradigms
Participants
Wendie A. Berg, MD, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
For information about this presentation, contact:
wendieberg@gmail.com
LEARNING OBJECTIVES
1) Distinguish and define average risk, intermediate risk, and high-risk populations. 2) Understand existing recommendations for
supplemental screening beyond mammography and/or screening at an earlier age. 3) Discuss potential strategies for elective
supplemental screening and expected outcomes.
 RC124

The Best of RADIOLOGY in 2018: The Editors of RADIOLOGY Keep You Up to Date
Sunday, Nov. 25 2:00PM - 3:30PM Room: E353A

BR CA CH NR RS
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Sub-Events
RC124A Review of 2018: New Research That Should Impact Your Practice
Participants
David A. Bluemke, MD,PhD, Bethesda, MD (Presenter) Nothing to Disclose
For information about this presentation, contact:

dbluemke@rsna.org
LEARNING OBJECTIVES
1) Identify key publications over the past year that may affect your clinical practice. 2) Evaluate new research developments in
the field of radiological imaging. 3) Describe new developments in radiology that may affect the management of your patients.
ABSTRACT
RADIOLOGY is the leading journal for publications leading to new, important and translatable discoveries in imaging research. In the
past year, there continue to be basic developments in radiology, as well as new guidelines and clinical trials in imaging that affect
your practice. Overall trends for new scientific studies reflect an increasing number of clinical trials being submitted from around the
world in addition to those of North America. Publications from Europe have been prominent in recent years, but new research
programs from countries such as Japan, South Korea and China are developing quickly. Large numbers of study subjects in clinical
trials are now common, and tends to result in more robust demonstration of the efficacy of imaging interventions. Artificial
intelligence applications are becoming commonplace in our publications, as are radiomics studies with increasing large numbers of
study subjects. This seminar will highlight the results of key publications in the past year that are most likely to affect your
practice in the near future, as well as presenting novel topics that are likely to be important to the field over the next 5 years.
RC124B Innovations in Cardiothoracic Imaging in 2018
Participants
Albert De Roos, MD, Leiden, Netherlands (Presenter) Nothing to Disclose
For information about this presentation, contact:
a.de_roos@lumc.nl
LEARNING OBJECTIVES
1) Key publications in cardiothoracic imaging 2018 will be highlighted.
ABSTRACT
Cardiothoracic manuscripts are frequently introducing new technology, acquisition techniques and clinical evaluation. Major
advances in cardiothoracic imaging over the last year published in Radiology will be discussed for their innovation and potential
impact.
RC124C Research and Innovations in Breast Imaging in 2018
Participants
Linda Moy, MD, New York, NY (Presenter) Nothing to Disclose
For information about this presentation, contact:
linda.moy@nyumc.org
LEARNING OBJECTIVES
1) To highlight key publications on breast imaging over the past year. 2) To discuss the implications of these publications for
patient care
RC124D New Developments in Neuroimaging in 2018
Participants
Birgit B. Ertl-Wagner, MD, Toronto, ON (Presenter) Spouse, Stockholder, Siemens AG; ;
For information about this presentation, contact:
BirgitBetina.Ertl-Wagner@sickkids.ca
LEARNING OBJECTIVES
1) Identify key publications over the past year that may affect your clinical practice. 2) Evaluate new research developments in
the field of radiological imaging. 3) Describe new developments in radiology that may affect the management of your patients.
 RC150

MR Imaging-guided Breast Biopsy (Hands-on)

Sunday, Nov. 25 2:00PM - 3:30PM Room: E260

BR MR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Manisha Bahl, MD,MPH, Boston, MA (Presenter) Nothing to Disclose
Rosalind P. Candelaria, MD, Houston, TX (Presenter) Nothing to Disclose
Sarah M. Friedewald, MD, Chicago, IL (Presenter) Consultant, Hologic, Inc; Research Grant, Hologic, Inc;
Brian Johnston, MD, Gilbert, AZ (Presenter) Nothing to Disclose
Jennifer R. Kohr, MD, Seattle, WA (Presenter) Nothing to Disclose
Lizza Lebron, MD, West Harrison, NY (Presenter) Nothing to Disclose
Diana L. Lam, MD, Seattle, WA (Presenter) Grant, General Electric Company
Santo Maimone IV, MD, Jacksonville, FL (Presenter) Research Consultant, GRAIL Inc
Cecilia L. Mercado, MD, New York, NY (Presenter) Nothing to Disclose
Bethany L. Niell, MD,PhD, Tampa, FL (Presenter) Nothing to Disclose
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Elissa R. Price, MD, San Francisco, CA (Presenter) Nothing to Disclose
Jean M. Seely, MD, Ottawa, ON (Presenter) Nothing to Disclose
Toma Omofoye, MD , Houston, TX (Presenter) Nothing to Disclose
Gaiane M. Rauch, MD, PhD, Houston, TX (Presenter) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Presenter) Research support, General Electric Company
Jocelyn A. Rapelyea, MD, Washington, DC (Presenter) Speakers Bureau, General Electric Company; Consultant, Transmed7;
Ryan W. Woods, MD, MPH, Madison, WI (Presenter) Nothing to Disclose
Beatriu Reig, MD, New York, NY (Presenter) Nothing to Disclose
Erin I. Neuschler, MD, Chicago, IL (Presenter) Research Grant, Seno Medical Instruments, Inc; Speaker, Northwest Imaging Forums,
Inc; ; ;
For information about this presentation, contact:
jessica.porembka@utsouthwestern.edu
rstrigel@uwhealth.org
mbahl1@mgh.harvard.edu
jeseely@toh.ca
rwoods@uwhealth.org
jrapelyea@mfa.gwu.edu
dllam@uw.edu
gmrauch@mdanderson.org
sarah.friedewald@nm.org
LEARNING OBJECTIVES
1) Explain why MR-guided breast biopsy is needed for patient care. 2) Identify relative and absolute contraindications to MR-guided
breast biopsy. 3) Describe criteria for MR-guided breast biopsy patient selection. 4) Debate risks and benefits of pre-biopsy
targeted ultrasound for suspicious MRI findings. 5) Understand the basic MR-guided biopsy procedure, protocol and requirements for
appropriate coil, needle and approach selection. 6) Manage patients before, during and after MR-guided breast biopsy. 7) Define
the benefits and limitations of MR-guided vacuum assisted breast biopsy. 8) Apply positioning and other techniques to challenging
combinations of lesion location and patient anatomy for successful MR-guided biopsy.
ABSTRACT
This course is intended to provide basic didactic instruction and hands-on experience for MR-guided breast biopsy. Because of the
established role of breast MRI in the evaluation of breast cancer through screening and staging, there is a proven need for MR-
guided biopsy of the abnormalities that can only be identified at MRI. This course will be devoted to the understanding and
identification of: 1) appropriate patient selection 2) optimal positioning for biopsy 3) target selection and confirmation 4) various
biopsy technologies and techniques 5) potential problems and pitfalls and 6) radiology/pathology concordance. Participants will
spend 30 minutes in didactic instruction followed by 60 minutes practicing MR-guided biopsy using provided phantoms. Various
combinations of full size state-of-the-art breast MRI coils, biopsy localization equipment and needles from multiple different vendors
will be available for hands-on practice. Some stations will have monitors loaded with targeting software. Expert breast imagers from
around the world will be at each of 10 stations to provide live coaching, tips, techniques and advice.
Active Handout:Roberta Marie Strigel
http://abstract.rsna.org/uploads/2018/4426451/Strigel-RSNA-MRIBx-2018-Handout RC150.pdf
 VW 03

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Sunday, Nov. 25 2:30PM - 3:00PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 26

AI-based Mammography Reading (Hands-on session): Siemens Healthineers Vendor Workshop

Sunday, Nov. 25 2:30PM - 3:40PM Room: Booth 5530

Participants
Nico Karssemeijer, PhD, Nijmegen, Netherlands (Presenter) Director and Shareholder, ScreenPoint Medical BV Shareholder, Volpara
Health Technologies Limited Consultant, QView Medical, Inc Shareholder, QView Medical, Inc
Program Information
You will learn about the use of the AI-supported Transpara™* system from Screenpoint in combination with syngo® Breast Care to
support 2D or 3D mammography reading. Transpara™ provides detection and decision support together with an overall exam based
score for prioritization of reading. You will experience a case review session to explore the use of artificial intelligence in
mammography reading. *Transpara™ is pending 510(k) clearance, and is not yet commercially available in the United States.
 VW 54

Advancements in Real-time Breast Specimen Imaging: Hologic Vendor Workshop

Sunday, Nov. 25 2:30PM - 3:00PM Room: Booth 5524

Program Information
Learn how real-time breast specimen imaging can streamline workflow efficiency through specimen verification and automated post-
biopsy specimen handling with the Brevera® Breast Biopsy System. (Brevera® Breast Biopsy System)
Registration
https://hologicrsna.com
 VW 55

Breast Density Assessment: Beyond Radiologist Visual Assessment: Hologic Vendor Workshop
Sunday, Nov. 25 3:30PM - 4:00PM Room: Booth 5524

Participants
Regina J. Hooley, MD, New Haven, CT (Presenter) Consultant, Hologic, Inc
Program Information
A brief session for radiologists interested in learning more about implementing Breast Density Assessment Software. Including
decreasing inter/intra-reader variability and increasing confidence in selecting appropriate patients that may need additional
screening. (Quantra™)
Registration
https://hologicrsna.com
 VW 27

A Practical Approach to Breast Magnetic Resonance Imaging (MRI) Interpretation: An Interactive Session:
Siemens Healthineers Vendor Workshop
Sunday, Nov. 25 3:50PM - 5:00PM Room: Booth 5530

Participants
Susan Weinstein, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Program Information
This interactive session will include both didactic and hands-on case review at workstations equipped with syngo® MR Brevis. A
practical approach to breast MRI interpretation will be discussed as well as utilizing the available sequences and techniques to
improve interpretive skills.
 PS12

Sunday Afternoon Plenary Session

Sunday, Nov. 25 4:00PM - 5:45PM Room: Arie Crow n Theater

BR CA CH GI GU MK NR PD VA
AMA PRA Category 1 Credits ™: 1.75
ARRT Category A+ Credit: 1.75

Participants
Vijay M. Rao, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Sub-Events
PS12A Report of the RSNA Research and Education Foundation
Participants
N. Reed Dunnick, MD, Ann Arbor, MI (Presenter) Royalties, Wolters Kluwer nv; Editor, Reed Elsevier
PS12B Image Interpretation Session
Participants
Donald P. Frush, MD, Durham, NC (Moderator) Nothing to Disclose
John Eng, MD, Cockeysville, MD (Introduction) Nothing to Disclose
Laura W. Bancroft, MD, Orlando, FL (Presenter) Author with royalties, Wolters Kluwer nv; Speaker, World Class CME; Editor, Thieme
Medical Publishers, Inc; Travel support, Thieme Medical Publishers, Inc ; ;
Matthew S. Davenport, MD, Ann Arbor, MI (Presenter) Nothing to Disclose
Tomas C. Franquet, MD, Barcelona, Spain (Presenter) Nothing to Disclose
R. Paul Guillerman, MD, Houston, TX (Presenter) Consultant, Guerbet SA
Christopher P. Hess, MD, PhD, Mill Valley, CA (Presenter) Nothing to Disclose
Andrea Laghi, MD, Rome, Italy (Presenter) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Presenter) Nothing to Disclose
Pamela K. Woodard, MD, Saint Louis, MO (Presenter) Research agreement, Siemens AG; Research, Eli Lilly and Company; Research,
F. Hoffmann-La Roche Ltd; ; ; ; ; ;
For information about this presentation, contact:
woodardp@wustl.edu
 VW 05

Patients Respond Positively to Supplemental Screening with ABUS and Report an Added Sense of Security: GE
Vendor Workshop
Sunday, Nov. 25 4:00PM - 4:25PM Room: Booth 8156

Program Information
Supplemental screening with ultrasound is proving to be an effective tool in detecting mammographically occult breast cancers. The
aim of this study is to identify characteristics of breast cancers detected by supplemental ABUS for women with dense breast
tissue, and to detect small, majority node negative, clinically significant, invasive breast cancers. Registration is required; adding
this session to the RSNA calendar tool alone does not secure your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 06

Hearing from the Voices That Matter: The Patients: GE Vendor Workshop
Sunday, Nov. 25 4:30PM - 5:00PM Room: Booth 8156

Program Information
In this study, patients report low anxiety, greater physical comfort and increased security that ABUS will detect a breast cancer
that mammography may miss in the setting of dense breast tissue. In addition, current patient research, including density
notification and attitudes toward screening, will be presented. Registration is required; adding this session to the RSNA calendar
tool alone does not secure your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 56

True Patient Comfort: Genius™ 3D Mammography™ Exam and SmartCurve™ Breast Stabilization System (In
Spanish): Hologic Vendor Workshop
Sunday, Nov. 25 4:30PM - 5:00PM Room: Booth 5524

Program Information
A clinical perspective on the implementation of Genius™ 3D Mammography™ exam and the SmartCurve™ Breast Stabilization
System without compromising image quality, exam time, dose or workflow. This session includes clinical feedback and tips to
increase patient comfort. (Genius™ 3D Mammography™, SmartCurve™ Breast Stabilization System)
Registration
https://hologicrsna.com
 ED001-MO

Breast Monday Case of the Day

Monday, Nov. 26 7:00AM - 11:59PM Room: Case of Day, Learning Center

AMA PRA Category 1 Credit ™: .50

Participants
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Susan O. Holley, MD, PhD, Raleigh, NC (Abstract Co-Author) Nothing to Disclose
Alexander B. Sevrukov, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Catherine A. Young, MD, JD, Saint Louis, MO (Abstract Co-Author) Research support, Hologic, Inc
Cheryl R. Herman, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Mai A. Elezaby, MD, Madison, WI (Abstract Co-Author) Research Grant, Exact Sciences Corporation
Lonie R. Salkowski, MD,PhD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Urvi A. Tailor, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Ramapriya Ganti, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1) Identify, characterize, and analyze abnormal findings on multimodality breast imaging studies. 2) Develop differential diagnostic
considerations based on the clinical information and imaging findings. 3) Recommend appropriate management for the patients
based on imaging findings.
 SPSC20

Controversy Session: Screening Breast MRI: Abbreviated versus Full Protocol

Monday, Nov. 26 7:15AM - 8:15AM Room: E350

BR MR
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Margarita L. Zuley, MD, Pittsburgh, PA (Moderator) Investigator, Hologic, Inc
For information about this presentation, contact:
zuleyml@upmc.edu
LEARNING OBJECTIVES
1) Understand the acquisition parameters for full vs abbreviated protocols for screening MRI. 2) Understand the literature
surrounding the benefits and limitations of each methodology. 3) Improve your interpretive skills for screening MRI.
Sub-Events
SPSC20A Abbreviated Protocol
Participants
Christopher E. Comstock, MD, New York, NY (Presenter) Nothing to Disclose
For information about this presentation, contact:
comstocc@mskcc.org
LEARNING OBJECTIVES
1) Describe the concept of abbreviated breast MRI (AB-MR) for breast cancer screening. 2) Review the data and current studies
evaluating AB-MR. 3) Discuss the possible benefits and future direction of the use of AB-MR.
SPSC20B Full Protocol
Participants
Bonnie N. Joe, MD, PhD, San Francisco, CA (Presenter) Nothing to Disclose
For information about this presentation, contact:
bonnie.joe@ucsf.edu
LEARNING OBJECTIVES
1) Describe the components of a full protocol breast MRI exam. 2) Understand the benefits and limitations of the full protocol breast
MRI exam in contrast to an abbreviated protocol.
ABSTRACT
This session will review the main components of the full-protocol breast MRI exam, including technical considerations. Benefits and
limitations of the full protocol versus an abbreviated protocol for breast cancer screening will be discussed.
URL
NA
 CS22

Contrast Enhanced Mammography: Incorporating CEM Into Your Clinical Practice: Presented by Hologic, Inc.
Monday, Nov. 26 8:30AM - 9:30AM Room: S102AB

Participants
Bhavika K. Patel, MD, Phoenix, AZ (Presenter) Speaker, Hologic, Inc
PROGRAM INFORMATION
This CME accredited symposium will focus on the implementation and advancement of utilizing contrast enhanced mammography
(CEM) to advance your clinical confidence. Areas of discussion include: 1) Requirements for building contrast-enhanced
mammography into your practice. 2) Proper patient selection for utilizing CEM. 3) Evaluation and interpretation of CEM as a
diagnostic tool.
CME
https://cine-med.com/rsna/
RSVP
https://cine-med.com/rsna/
 MSRO25

BOOST: Breast-Oncology Anatomy (Interactive Session)

Monday, Nov. 26 8:30AM - 10:00AM Room: S103CD

BR OI RO
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Stephanie Markovina, MD, PhD, Saint Louis, MO (Moderator) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Presenter) Research support, General Electric Company
Maria A. Thomas, MD, PhD, Saint Louis, MO (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Understand breast and regional lymph node anatomy. 2) Become familiar with basic anatomic structures and breast pathology
using various imaging modalities. 3) Be familiar with breast and regional lymph node contouring techniques used in radiation
treatment planning for breast cancer. 4) Apply contouring knowledge to inform radiation treatment planning for breast cancer.
 RC215

Breast Series: Hot Topics (The In-Person Presentation is Supported by an Unrestricted Educational Grant
from Hologic)
Monday, Nov. 26 8:30AM - 12:00PM Room: Arie Crow n Theater

AI BR
AMA PRA Category 1 Credits ™: 3.50
ARRT Category A+ Credits: 4.00

FDA Discussions may include off-label uses.

Participants
Linda Moy, MD, New York, NY (Moderator) Nothing to Disclose
Fiona J. Gilbert, MD, Cambridge, United Kingdom (Moderator) Research Grant, Hologic, Inc; Research Grant, General Electric
Company; Research Grant, GlaxoSmithKline plc; Research Consultant, Alphabet Inc
Sub-Events
RC215-01 Radiomics
Monday, Nov. 26 8:30AM - 8:50AM Room: Arie Crow n Theater

Participants
Karen Drukker, PhD, Chicago, IL (Presenter) Royalties, Hologic, Inc
For information about this presentation, contact:
kdrukker@uchicago.edu
Active Handout:Karen Drukker
http://abstract.rsna.org/uploads/2018/18000478/RSNA2018_Drukker_Handout RC215-01.pdf
LEARNING OBJECTIVES
1) Identify the scientific premise, motivation, and increasing role of radiomics in medical imaging. 2) Compare 'conventional'
radiomics methods and deep learning-based radiomics methods. 3) Assess some of the challenges for radiomics-based decision
support systems in becoming powerful players in modern precision medicine.
RC215-02 Quantitative Diffusion-Weighted MRI of Estrogen Receptor-Positive, Lymph Node-Negative Invasive
Breast Cancer: Association between Whole-Lesion Apparent Diffusion Coefficient Metrics and
Recurrence Risk
Monday, Nov. 26 8:50AM - 9:00AM Room: Arie Crow n Theater

Participants
Jin You Kim, MD, Busan, Korea, Republic Of (Presenter) Nothing to Disclose
Lee Hwangbo, MD, Pusan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jin Joo Kim, MD, Busan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Suk Kim, MD, Pusan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

youdosa@naver.com
PURPOSE
To investigate possible associations between quantitative apparent diffusion coefficient (ADC) metrics derived from whole-lesion
histogram analysis and breast cancer recurrence risk in patients with estrogen receptor (ER)-positive, lymph node-negative invasive
breast cancer who underwent the Oncotype DX assay.
METHOD AND MATERIALS
Institutional review board approval was obtained for this retrospective study, which was conducted on 74 women (mean age, 49.3
years) with ER-positive, lymph node-negative invasive breast cancer who underwent the Oncotype DX assay and preoperative
diffusion-weighted MRI from July 2015 to January 2018. Histogram analysis of pixel-based ADC data of whole tumors was performed
by two radiologists using a software tool and various ADC histogram parameters (mean, minimum, maximum, and 5th, 25th, 50th,
75th, and 95th percentile ADCs) were extracted. The ADC difference value (defined as the difference between minimum and
maximum ADC) was calculated to assess intratumoral heterogeneity. Associations between quantitative ADC metrics and Oncotype
DX risk groups (low [recurrence score (RS) <18], intermediate (RS 18-30), and high [RS >30]) were evaluated by receiver operating
characteristic (ROC) curve and logistic regression analyses.
RESULTS
Whole-lesion histogram analysis showed minimum ADCs, maximum ADCs, and ADC difference values were significantly different
between low and non-low (ie, intermediate and high) risk groups (0.604, 1.478, and 0.874 × 10-3mm2/s versus 0.374, 1.687, and
1.321 × 10-3mm2/s, respectively; P<0.001, P=0.010, and P<0.001, respectively). The ADC difference value yielded the largest area
under the ROC curve (0.771; 95% confidence interval [CI]: 0.650, 0.891; P<0.001) for differentiating the two groups. Multivariate
regression analysis showed that the ADC difference value was the only significant factor associated with low Oncotype DX risk
group (adjusted odds ratio = 0.998; 95% CI: 0.996, 0.999; P<0.001).
CONCLUSION
The ADC difference value derived from whole-lesion histogram analysis could be helpful for identifying ER-positive, lymph node-
negative invasive breast cancer patients with low risk of recurrence.
CLINICAL RELEVANCE/APPLICATION
In estrogen receptor-positive, lymph node-negative breast cancer, the ADC difference value derived from whole-lesion histogram
assessments might serve as quantitative biomarkers of recurrence risk.
RC215-03 Radiomic Phenotypes of Tumor Heterogeneity from Pre-Operative DCE-MRI Predict Breast Cancer
Recurrence after 10-Year Follow-Up: Phenotype Discovery and Independent Validation
Monday, Nov. 26 9:00AM - 9:10AM Room: Arie Crow n Theater

Participants
Rhea Chitalia, Philadelphia, PA (Presenter) Nothing to Disclose
Jennifer Rowland, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. McDonald, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Eric A. Cohen, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Aimilia Gastounioti, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Kathleen M. Thomas, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Rebecca Batiste, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Michael D. Feldman, MD, PhD, Philadelphia, PA (Abstract Co-Author) Advisory Board, Inspirata Inc Advisory Board, Koninklijke Philips
NV Advisory Board, XIFIN, Inc
Mitchell D. Schnall, MD, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rhea.chitalia@uphs.upenn.edu
PURPOSE
To validate intrinsic imaging phenotypes of tumor heterogeneity and evaluate their prognostic performance in predicting 10-year
recurrence.
METHOD AND MATERIALS
Pre-treatment DCE-MRI scans of 94 women with primary invasive breast cancer and 10-year follow up data available were
retrospectively analyzed from a clinical trial cohort at our institution (2002-2006). For each woman, a signal enhancement ratio map
was generated for the most representative slice of the primary lesion from which morphologic features were calculated. Radiomic
features (histogram, run-length, structural, and co-occurrence matrix features) were extracted and summarized over tumor
quadrants. Intrinsic phenotypes of tumor heterogeneity were identified via unsupervised hierarchical clustering applied to the
extracted feature vectors, with significant clusters found using Consensus Clustering and the SigClust method. Differences across
phenotypes by hormone receptor status, tumor size, post-surgery therapy, TNM staging, and recurrence outcomes were assessed
using Chi-square and Kruskal-Wallis tests. An independent dataset of 116 women diagnosed with primary invasive breast cancer
(2002-2006), available via The Cancer Imaging Archive, was used to validate phenotype reproducibility. Survival probabilities across
phenotypes were evaluated using Kaplan-Meier curves and phenotype cluster assignments were added to a baseline Cox
proportional hazards model with established histopathologic prognostic factors to predict RFS.
RESULTS

Three significant phenotypes of low, medium, and high heterogeneity were identified in the discovery cohort and reproduced in the
validation cohort (p<0.001). No recurrent cases were found in the low heterogeneity phenotype (p<0.001). Clinical stage, mitotic
grade, lymph invasion, and nuclear grade were different across phenotypes (p<=0.02). Kaplan-Meier curves showed significant
differences (p < 0.001) in RFS probabilities across phenotypes. The augmented model including phenotype assignment had a higher
discriminatory capacity (c-statistic= 0.80) compared to a baseline model with only established prognostic factors (c-statistic=
0.65, p<0.01).
CONCLUSION

Intrinsic imaging phenotypes of tumor heterogeneity can predict 10-year recurrence as validated in an independent dataset.
CLINICAL RELEVANCE/APPLICATION

Radiomic phenotypes could provide a non-invasive characterization of tumor heterogeneity to augment personalized prognosis and
treatment.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Mitchell D. Schnall, MD, PhD - 2013 Honored Educator
RC215-04 Robustness of Computer-aided Diagnosis of Breast Cancer Using Radiomics and Machine Learning
Classification of 1,461 Lesions across Populations in China and the United States
 Monday, Nov. 26 9:10AM - 9:20AM Room: Arie Crow n Theater

Participants
Heather Whitney, PhD, Wheaton, IL (Presenter) Nothing to Disclose
Hui Li, PHD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Yu Ji, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Alexandra V. Edwards, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc; Research Consultant,
Quantitative Insights, Inc
John Papaioannou, MSc, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc
Peifang Liu, MD, PhD, Tianjin, China (Abstract Co-Author) Nothing to Disclose
Maryellen L. Giger, PhD, Chicago, IL (Abstract Co-Author) Stockholder, Hologic, Inc; Shareholder, Quantitative Insights, Inc;
Shareholder, QView Medical, Inc; Co-founder, Quantitative Insights, Inc; Royalties, Hologic, Inc; Royalties, General Electric
Company; Royalties, MEDIAN Technologies; Royalties, Riverain Technologies, LLC; Royalties, Mitsubishi Corporation; Royalties,
Canon Medical Systems Corporation
For information about this presentation, contact:
hwhitney@uchicago.edu
PURPOSE
To assess the performance of computer aided diagnosis (CADx) in breast lesions imaged with DCE-MR in two patient cohorts, one in
China and one in the United States (US), using extracted radiomic features and machine learning classification.
METHOD AND MATERIALS
Dynamic contrast-enhanced magnetic resonance (DCE-MR) images of 1,461 breast lesions (from China, GE scanners: 300 benign
lesions, 302 malignant cancers; from the US, Philips scanners: 268 benign lesions, 591 malignant cancers) were collected under
HIPAA and IRB compliance. The lesions were segmented automatically using a fuzzy c-means method. Thirty-eight radiomic features
describing size, shape, morphology, kinetics, and texture were extracted using previously reported methods. The performance of
CADx for classification between benign lesions and malignant cancers was evaluated with two methodologies: (a) independent
training and testing of the datasets, with each set serving as a training set while the other served as a testing set; and (b) ten-
fold cross validation within each set. Classification was performed using support vector machines with optimization of the
hyperparameters. The area under the ROC curve (AUC) served as figure of merit, with its value and standard error determined using
the conventional binormal model. The AUCs resulting from (a) and (b) were compared within and between each methodology.
Difference in AUC was significantly different when p < 0.05.
RESULTS

When radiomic features extracted from MRIs acquired in China were used to train the machine classifiers and independent testing
was conducted on MRIs acquired in the US, AUC = 0.77 (0.02), while the reverse resulted in AUC = 0.79 (0.02). For cross-
validation within each set, AUC = 0.82 (0.02) for the US database and AUC = 0.80 (0.02) for the China database. AUCs compared
across methodologies failed to show significant difference.
CONCLUSION
Computer aided diagnosis of breast lesions demonstrated potential robustness across independent populations in both independent
training/testing and in cross validation.
CLINICAL RELEVANCE/APPLICATION
Radiomic features extracted from DCE-MRI may be robust for classifying breast lesions as benign or malignant across two cohorts
(one in China, one in US), enhancing translation to clinical use.
RC215-05 Radiogenomics
Monday, Nov. 26 9:20AM - 9:40AM Room: Arie Crow n Theater

Participants
Lars J. Grimm, MD, Durham, NC (Presenter) Editorial Advisory Board, Medscape, LLC; Educational program support, Hologic, Inc
For information about this presentation, contact:

lars.grimm@duke.edu
LEARNING OBJECTIVES

1) Define radiogenomics and describe how it differs from radiomics. 2) Examine the limitations of current radiogenomics research. 3)
Assess the utility of radiogenomics in clinical practice. 4) Develop a framework to evaluate future radiogenomics research.
RC215-06 Proteomic Expression Underlying Quantitative MRI Features in Breast Cancer: A Radioproteomics
Study
Monday, Nov. 26 9:40AM - 9:50AM Room: Arie Crow n Theater

Participants
Ryan M. Hausler, BS, Pittsburgh, PA (Presenter) Nothing to Disclose
Ruimei Chai, Shenyang, China (Abstract Co-Author) Nothing to Disclose
Dooman Arefan, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Jules H. Sumkin, DO, Pittsburgh, PA (Abstract Co-Author) Research Grant, Hologic, Inc; Research Grant, General Electric Company
Min Sun, MD,PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Shandong Wu, PhD, MSc, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

rmh1995@gmail.com
rmh1995@gmail.com
PURPOSE
The complementary analysis of breast cancer via radiology imaging and molecular pathology approaches has spurred radiogenomics
and radioproteomics studies. We performed an investigation of the relationships between quantitative radiomic imaging phenotype
data and underlying proteomic expression, with the goal of improving precise breast cancer diagnosis and cancer behavior
characterization.
METHOD AND MATERIALS

We identified a retrospective cohort of 40 invasive breast cancer patients from a single medical center. Their integrated protein
expression data were obtained from The Cancer Genome Atlas study. The proteomic data was acquired via Reverse Phase Protein
Array (RPPA) to measure the expression of 217 breast cancer related proteins and phospho-proteins. Dynamic Contrast Enhanced
Magnetic Resonance Imaging (DCE-MRI) data of the 40 patients were collected from clinical archive, all acquired with a 1.5T same-
vendor scanner. A set of 30 radiomic imaging features were extracted from automatically-segmented tumor volume in all 40 DCE-
MRIs to capture tumor morphological and contrast enhancement characteristics. Multivariate linear regression was used to map the
associations between each imaging feature with each of the 217 protein expressions, controlling for patient age and cancer stage.
A p value was obtained evaluating the significance of the association and was adjusted for multiple comparisons of the selected
radiomic feature against every protein. Adjusted p values less than 0.05 were recorded.
RESULTS

The average patient age at scan was 38.7±12 years, 10 (25%) of which were pre- with the rest post-menopausal. We found a
variety of expression of cancer related proteins were significantly associated (positively or negatively) with a subset of
morphological and contrast enhancement kinetics related imaging features. For example, ERCC5 (a protein responsible for DNA repair
following UV-induced damage) is negatively associated with the tumor brightness and contrast agent uptake rates. The full
association map is shown in the attached figure.
CONCLUSION

Our study showed that the expression of several cancer related proteins were found to be linearly associated with quantitative
DCE-MRI-derived phenotype features in invasive breast tumors.
CLINICAL RELEVANCE/APPLICATION

Radioproteomic studies of cancer can help to decipher how molecular mechanisms may regulate the development of specific tumor
phenotypes.
RC215-07 Prediction of 21-gene Recurrence Score in Patients with Estrogen Receptor-positive Early-Stage
Breast Cancer Using MRI-based Radiomics Nomogram
Monday, Nov. 26 9:50AM - 10:00AM Room: Arie Crow n Theater

Participants
Nam Joo Lee, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa Jung Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ki Chang Shin, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jong Won Lee, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sae Byul Lee, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

docshin@amc.seoul.kr
PURPOSE

To develop a breast MRI-based radiomics nomogram including pathologic factors which can predict low-risk recurrence score (RS)
on 21-gene RS assay in patients with estrogen receptor-positive early-stage breast cancer (EBC).
METHOD AND MATERIALS

From 2011 to 2017, a total of 547 tumors in 539 patients with EBC who underwent preoperative breast MRI were retrospectively
included in this study. Among them, low-risk was 320 (58.5%), intermediate-risk was 180 (32.9%), and high-risk was 47 (8.6%).
We extracted 744 quantitative MR radiomic features from computerized three-dimensional segmentations of each tumor generated
computer-extracted image phenotypes (CEIP) within the intratumoral regions of early post-contrast T1-weighted images, percent
enhancement (PE) map, signal enhancement ratio (SER) map, and T2-weighted images. We divided 547 cases into a training set
(n=365) and a validation set (n=182). Elastic net was used for feature selection and radiomics score building. Multivariate logistic
regression analysis was used to develop a prediction model, we incorporated the radiomics score and independent pathologic risk
factors and build a radiomics nomogram. Internal validation for an independent validation set (n=182) was performed.
RESULTS

The radiomics score, which consisted of 24 selected CEIPs, was significantly associated with the prediction of recurrence (C-index,
0.769 for training set and 0.745 for validation set). Independent pathologic predictors contained in the nomogram were
progesterone receptor status, nuclear grade, histologic grade, extensive intraductal component, lymphovascular invasion, P53, and
Ki67 status, and their C-index was 0.858 for training set and 0.774 for validation set. Addition of radiomics score to the pathologic
nomogram showed an incremental value of 0.054 and 0.092, respectively. Radiomics nomogram showed good prediction of low-risk
RS, with a C-index of 0.912 for training set and 0.866 for validation set.
CONCLUSION
This study shows that a radiomics nomogram which incorporates the MRI-based radiomics score and pathologic features, can be
used to help the preoperative individualized prediction of low-risk RS in patients with EBC.
CLINICAL RELEVANCE/APPLICATION

Prediction nomogram using breast MRI-based radiomics score and pathologic predictors can be used to facilitate the preoperative
individualized prediction of low-risk RS on 21-gene RS assay in patients with EBC.
RC215-08 Can Histogram Analysis of Dynamic Contrast-Enhanced MRI and Apparent Diffusion Coefficient Map
Predict Molecular Subtypes of Invasive Breast Cancers?
Monday, Nov. 26 10:00AM - 10:10AM Room: Arie Crow n Theater

Participants
Joao V. Horvat, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Doris Leithner, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Blanca Bernard-Davila, MPH,MS, New York, NY (Abstract Co-Author) Nothing to Disclose
Rosa E. Ochoa Albiztegui II, MD, Mexico CIty, Mexico (Abstract Co-Author) Nothing to Disclose
Danny F. Martinez, BSC,MSc, New York, NY (Abstract Co-Author) Nothing to Disclose
Olivia Sutton, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Sunitha Thakur, PhD, MS, New York, NY (Abstract Co-Author) Nothing to Disclose
Katja Pinker-Domenig, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

joaohorvat@gmail.com
PURPOSE

To evaluate if histogram analysis of dynamic contrast-enhanced (DCE) MRI and apparent diffusion coefficient (ADC) maps with
diffusion-weighted imaging (DWI) can predict molecular subtypes of invasive breast cancers.
METHOD AND MATERIALS

In this HIPAA-compliant and IRB-approved study we retrospectively evaluated 91 consecutive patients from January 2011 to
January 2013 with invasive ductal carcinoma of the breast who underwent multiparametric MRI with DCE and DWI at our institution.
The exclusion criteria were 1) lesion smaller than 1 cm, 2) previous treatment for breast cancer, 3) pathology report unavailable,
and 4) poor image quality. One experienced breast radiologist drew a region of interest on DCE MRI and ADC maps on the slice with
the largest diameter of the solid portion of the lesion avoiding cystic areas and biopsy markers. The histogram analysis was
performed and the mean, variance, kurtosis and skewness were calculated. Molecular breast cancer subtypes were derived by IHC
surrogates. Tumors were classified as luminal A if either ER or PR was positive and HER2 was negative, Luminal B if either ER or PR
was positive and HER2 positive, HER2-enriched if ER and PR were negative and HER2 positive and triple-negative if ER, PR and HER2
were negative. Nonparametric Mann-Whitney U test and Kruskal-Wallis were used to compare groups of molecular subtypes. P-
values <0.05 were accepted to be statistically significant.
RESULTS
The histogram analysis of DCE images and ADC maps of 91 breast cancers demonstrated no significant difference among breast
tumor molecular subtypes. Measurements of the mean, variance, kurtosis and skewness were used to compare luminal A/B with
HER-2 enriched/triple-negative cancers, without significant results for both DCE (p-value = 0.405, 0.252, 0.667, 0.809) and ADC
(0.204, 0.081, 0.941, 0.574), respectively. Histogram measurements were also used to compare luminal A with other subtypes and
also demonstrated no significant difference for DCE (0.659, 0.162, 0.516, 0.833) and ADC (0.204, 0.222, 0.495, 0.896).
CONCLUSION

Histogram analysis of DCE MRI and ADC map cannot predict molecular subtypes of invasive breast cancers.
CLINICAL RELEVANCE/APPLICATION

Despite many valuable applications of histogram analysis in diagnostic imaging, it cannot predict molecular subtypes of invasive
breast cancers.
RC215-09 CESM Enhancement Pattern and Intensity and Its Correlation to Breast Cancer Immunophenotype:
Preliminary Results
Monday, Nov. 26 10:10AM - 10:20AM Room: Arie Crow n Theater

Participants
Elzbieta Luczynska, MD, Cracow, Poland (Presenter) Nothing to Disclose
Sylwia Heinze, PhD, Cracow, Poland (Abstract Co-Author) Nothing to Disclose
Joanna Niemiec, Cracow, Poland (Abstract Co-Author) Nothing to Disclose
Agnieszka Adamczyk, Cracow, Poland (Abstract Co-Author)
Wojciech Rudnicki, MD, Krakow, Poland (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
z5luczyn@cyfronet.pl
PURPOSE

The differences in the intensity and pattern of enhancement in CESM between breast carcinomas might result from the differences
in the amount of contrast that leaked out from the blood vessels and timely arrested in the interstitium. The aim of this paper is to
study the expression of podoplanin in cancer stroma and its relation to breast cancer immunophenotype.
METHOD AND MATERIALS
Patients with lesions enhancing on CESM were subjected to biopsy - material obtained during biopsies was histopathologically
verified. In the present study we retrospectively investigated 97 invasive breast carcinomas diagnosed in 94 patients. This study
was performed in compliance with the Declaration of Helsinki and it received the approval of Ethical Committee at the Regional
Medical Chamber. For each tumor enhancing on CESM, the intensity and the pattern of enhancement were evaluated. The
enhancement of contrast agent uptake was qualitatively assessed as weak/medium or strong ,while the pattern as heterogenous or
homogenous. Lymphatic vessels were defined as strongly podoplanin-stained structures with lymphatic vessel characteristics ,
clearly distinguishable from other tissue structures and cells. We classified tumor stroma as: podoplanin-sparse and podoplanin-rich.
RESULTS

Strong enhancement on CESM was found more frequently in: large tumors (pT>1), node-positive carcinomas, in tumors with
podoplanin-sparse stroma vs. tumors with podoplanin-rich stroma . We found no relationship between enhancement on CESM and:
tumor grade, histological type of cancer, breast cancer immunophenotype and Ki-67LI. However, in luminal A tumors strong
enhancement on CESM was insignificantly more frequent as compared to neoplasms with non-luminal A subtype .
CONCLUSION
In our study prognostic significance of selected CESM features was found for the first time: strong and heterogenous enhancement
on CESM was related to poor patients' outcome. In this study, the aforementioned correlation was additionally confirmed by the
relationship between strong enhancement on CESM and nodal involvement or large tumor size.
CLINICAL RELEVANCE/APPLICATION
Our results may suggest that intensity and pattern of enhancement on CESM might bring (together with the results of diagnostic
imaging methods) not only the confirmation of presence or absence of tumor, but also prognostic information.
RC215-10 Development of MRI-based Radiomics Nomogram for the Prediction of Recurrence in Patients with
Luminal-type Breast Cancer: A Nested Case-Control Study
Monday, Nov. 26 10:20AM - 10:30AM Room: Arie Crow n Theater

Participants
Bo Yong Chung, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa Jung Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ki Chang Shin, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
docshin@amnc.seoul.kr
PURPOSE

To determine whether breast MRI-based radiomics nomogram including pathologic factors can predict recurrences or distant
metastasis in patients with luminal-type breast cancer (LTBC).
METHOD AND MATERIALS
From 2006 to 2012, a total of 348 patients with LTBC who underwent preoperative breast MRI were retrospectively included in this
study. Patients with recurrence were 174. Patients without recurrence were matched in terms of age, stage, and type of
chemotherapy, and developed 174 nested case-control pairs. We extracted 804 quantitative MR radiomic features of computerized
three-dimensional segmentations of each cancer generated computer-extracted image phenotypes (CEIP) within the intratumoral
regions of early post-contrast T1-weighted images, percent enhancement (PE) map, signal enhancement ratio (SER) map, and T2-
weighted images. We divided 174 case-control matches into a training set (n=232) and a validation set (n=116). Elastic net was
used for feature selection and radiomics score building. Multivariate logistic regression analysis was used to develop the prediction
model, we incorporated the radiomics score and independent pathologic risk factors and build a radiomics nomogram. Internal
validation for an independent validation set (n=76) was performed.
RESULTS
The radiomics score, which consisted of 14 selected CEIPs, was significantly associated with the prediction of recurrence (C-index,
0.864 for training set and 0.815 for validation set). Independent pathologic predictors contained in the nomogram were
progesterone receptor status, P53, lymphovascular invasion, Ki67 status, and lymph node ratio, and their C-index was 0.695 for
training set and 0.701 for validation set. Addition of radiomics score to the pathologic nomogram showed an incremental value of
0.211 and 0.177, respectively. Radiomics nomogram showed good prediction of recurrence, with a C-index of 0.906 for training set
and 0.878 for validation set.
CONCLUSION
This study shows that a radiomics nomogram which incorporates the MRI-based radiomics score and pathologic features, can be
used to help the individualized prediction of local or distant recurrence in patients with LTBC.
CLINICAL RELEVANCE/APPLICATION

Nomogram using breast MRI-based radiomics score and pathologic predictors can be used to facilitate the individualized prediction
of recurrence in patients with LTBC.
RC215-11 Horizons with Deep Learning
Monday, Nov. 26 10:40AM - 11:00AM Room: Arie Crow n Theater

Participants
Robert M. Nishikawa, PhD, Pittsburgh, PA (Presenter) Royalties, Hologic, Inc; Research Grant, Hologic, Inc; Research Consultant,
iCAD, Inc; Research Grant, Koios Medical
For information about this presentation, contact:
nishikawarm@upmc.edu
LEARNING OBJECTIVES
1) To understand the importance implementing deep learning tools into a breast imager's workflow. 2) To understand applications of
deep learning outside of detection and characterization of breast lesions.
RC215-12 Incorporating Patient Characteristics in Breast Cancer Screening with Deep Convolutional Neural
(DCN) Network
Monday, Nov. 26 11:00AM - 11:10AM Room: Arie Crow n Theater

Participants
Eric Kim, MD, New York, NY (Presenter) Nothing to Disclose
Krzysztof J. Geras, New York City, NY (Abstract Co-Author) Nothing to Disclose
Nan Wu, New York City, NY (Abstract Co-Author) Nothing to Disclose
Yiqiu Shen, New York City, NY (Abstract Co-Author) Nothing to Disclose
Jingyi Su, New York City, NY (Abstract Co-Author) Nothing to Disclose
Sungheon Kim, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Stacey Wolfson, New York, NY (Abstract Co-Author) Nothing to Disclose
Linda Moy, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Kyunghyun Cho, New York City, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

kime18@nyumc.org
PURPOSE

To determine if the addition of patient characteristics obtained from the electronic health records may improve the ability of a DCN
network to detect and classify lesions on screening mammography.
METHOD AND MATERIALS
This is a retrospective study of a DCN network trained on over 250,000 screening mammograms performed at our institution from
2010-2016. The patients were sorted according to the date of their latest exam and divided into training (first 80%), validation
(next 10%), and test (last 10%) sets. In the test phase, only the most recent exam was used for each patient. Patient
characteristics including age, family history of breast cancer, and history of prior examinations were extracted from the radiologist
reports. The original high-resolution images and extracted side information were utilized as inputs by a multi-column DCN network to
classify BI-RADS category. The model was evaluated using area under the receiver operating characteristic curve (AUC) analysis.
Analysis was also performed after stratifying patients by age-group and breast density (dense vs non-dense).
RESULTS

The overall performance of the DCN network improved with the addition of patient characteristics in comparison to using images
alone (AUC 0.750 vs 0.733). This improvement was especially notable for BI-RADS 0 cases, with an AUC of 0.664 vs 0.618.
Performance also generally improved with increasing age, with an average AUC of 0.759 in patients over 70 years of age. Finally,
performance of the model is superior in dense breasts vs non-dense breasts (AUC 0.740 vs AUC 0.707).
CONCLUSION

The performance of DCN networks in evaluating screening mammograms increases with the addition of patient characteristics
information, especially in the abnormal BI-RADS 0 cases which are the most difficult to evaluate.
CLINICAL RELEVANCE/APPLICATION
End-to-end architectures of DCN networks, like ours, support the incorporation of patient characteristics to increase the accuracy
of deep learning algorithms in breast cancer screening.
RC215-13 Detecting Breast Cancer in Mammography: A Deep Learning-Based Computer System versus 101
Radiologists
Monday, Nov. 26 11:10AM - 11:20AM Room: Arie Crow n Theater

Participants
Alejandro Rodriguez-Ruiz, Nijmegen, Netherlands (Abstract Co-Author) Nothing to Disclose
Albert Gubern-Merida, PhD, Nijmegen, Netherlands (Abstract Co-Author) Employee, ScreenPoint Medical
Kristina Lang, MD,PhD, Malmo, Sweden (Abstract Co-Author) Travel support, Siemens AG Speaker, Siemens AG
Mireille Broeders, PhD, Nijmegen, Netherlands (Abstract Co-Author) Nothing to Disclose
Gisella Gennaro, PhD, Padua, Italy (Abstract Co-Author) Nothing to Disclose
Paola Clauser, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Margarita Chevalier, PhD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Thomas H. Helbich, MD, Vienna, Austria (Abstract Co-Author) Research Grant, Medicor, Inc Research Grant, Siemens AG Research
Grant, C. R. Bard, Inc
Tao Tan, Nijmegen, Netherlands (Abstract Co-Author) Research Grant, QView Medical, Inc
Thomas Mertelmeier, PHD, Forchheim, Germany (Abstract Co-Author) Employee, Siemens AG; Stockholder, Siemens AG
Matthew G. Wallis, MD, Cambridge, United Kingdom (Abstract Co-Author) Nothing to Disclose
Ingvar T. Andersson, MD, PhD, Malmo, Sweden (Abstract Co-Author) Nothing to Disclose
Sophia Zackrisson, Malmo, Sweden (Abstract Co-Author) Speaker, AstraZeneca PLC ; Speaker, Siemens AG; Travel support,
AstraZeneca PLC; Travel support, Siemens AG
Ritse M. Mann, MD, PhD, Nijmegen, Netherlands (Abstract Co-Author) Researcher, Siemens AG ; Researcher, Seno Medical
Instruments, Inc; Researcher, Identification Solutions, Inc; Researcher, Micrima Limited; Researcher, Medtronic plc; Scientific
Advisor, ScreenPoint Medical BV; Scientific Advisor, Transonic Imaging, Inc; Stockholder, Transonic Imaging, Inc
Ioannis Sechopoulos, PhD, Atlanta, GA (Presenter) Research Grant, Siemens AG; Research Grant, Canon Medical Systems
Corporation; Speakers Bureau, Siemens AG; Scientific Advisory Board, Fischer Medical
PURPOSE
To compare the stand-alone performance of a computer-based detection system to that of radiologists in detecting breast cancer
on digital mammography (DM).
METHOD AND MATERIALS
Nine multi-reader multi-case (MRMC) study datasets previously used for different performance evaluation purposes in seven
countries were collected. Each dataset consisted of DM exams acquired with systems from four different vendors, multiple
radiologists' assessments per exam (BI-RADS or probability-of-malignancy scores), and ground truth: yielding a total of 2,458 exams
(608 malignant) and interpretations by 101 radiologists (28,373 independent exam interpretations). A deep learning-based computer
system (Transpara, ScreenPoint Medical, Nijmegen, The Netherlands) was used to automatically analyze each exam, resulting in a
score for suspiciousness of cancer (1-100). Independently for each dataset, the area under the receiver operating characteristic
curve (AUC) and the sensitivity at the radiologists' specificity level (case recall) were compared between the computer and
radiologists using MRMC analysis of variance.
RESULTS
The performance of the computer system was not significantly different to that of the average of radiologists in eight of nine
datasets (AUC differences ranged between -2.6% and +2.5%, P>0.329) and was significantly better in the ninth (+4.6%, P=0.036).
At the average specificity of the radiologists, the computer had an equal or higher sensitivity (+0-9%, P>0.083) in all datasets but
one (-13%, P=0.066). Comparing individually, the computer had an AUC and sensitivity higher than 53% and 65% of all radiologists,
respectively.
CONCLUSION
A computer system based on deep learning has an equivalent performance to radiologists for detecting breast cancer in
mammography.
CLINICAL RELEVANCE/APPLICATION

Whether used for decision support (preventing overlook and interpretation errors that are relatively common in the reading of
mammography) or as stand-alone readers, computer systems performing at radiologist-like level might herald a breakthrough in the
breast cancer detection workflow with mammography. In some situations, where there is a lack of experienced breast radiologists,
it might even allow the development or continuation of screening programs.
RC215-14 Improving Accuracy and Efficiency with Concurrent Use of Artificial Intelligence for Digital Breast
Tomosynthesis Screening
Monday, Nov. 26 11:20AM - 11:30AM Room: Arie Crow n Theater

Participants
Emily F. Conant, MD, Philadelphia, PA (Presenter) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc; Consultant, iCAD,
Inc; Speaker, iiCME
Alicia Y. Toledano, DSc, Kensington, MD (Abstract Co-Author) Consultant, iCAD, Inc
Senthil Periaswamy, PhD, Nashua, NH (Abstract Co-Author) Vice President, iCAD, Inc
Sergei V. Fotin, PhD, Nashua, NH (Abstract Co-Author) Principal Scientist, iCAD, Inc; Stockholder, iCAD, Inc
Jonathan Go, Nashua, NH (Abstract Co-Author) Sr. Vice President, iCAD, Inc; ;
Jeffrey W. Hoffmeister, MD, Nashua, NH (Abstract Co-Author) Employee, iCAD, Inc; Stockholder, iCAD, Inc
Justin E. Boatsman, MD, San Antonio, TX (Abstract Co-Author) Consultant, iCad, Inc
For information about this presentation, contact:
Emily.Conant@uphs.upenn.edu
PURPOSE
Screening with Digital Breast Tomosynthesis (DBT) improves accuracy but prolongs reading time when compared to Full-Field Digital
Mammography (FFDM) alone. A reader study evaluated concurrent use of Artificial Intelligence (AI) to shorten reading time, while
maintaining or improving sensitivity and specificity.
METHOD AND MATERIALS
An AI system based on deep convolutional neural networks was developed to identify suspicious soft tissue and calcific lesions in
DBT slices. Findings are outlined in slices, indicating AI's confidence of malignancy with 0-100 scores. A retrospective, fully-
crossed, multi-reader, multi-case designed study compared performance of 24 radiologists reading 260 DBT cases both with and
without AI. The case set included 65 cancer cases with 66 malignant lesions and 65 cases with biopsy-proven benign lesions.
Readings with and without AI occurred in 2 visits separated by a memory washout period of at least 4 weeks. Performance was
assessed by measuring Area Under the ROC Curve (AUC) for malignant lesions with AI versus without AI. Reading time, sensitivity,
specificity and recall rate were also assessed.
RESULTS
Radiologist performance for detection of malignant lesions, measured by mean AUC, increased 0.057 with use of AI (95% CI: 0.028,
0.087; p < 0.01), from 0.795 without AI to 0.852 with AI. Reading time decreased 52.7% with use of AI (95% CI: 41.8%, 61.5%; p
< 0.01), from 64.1 sec without AI to 30.4 sec with AI, using a normalizing transformation to appropriately assess reading times that
were not normally distributed. Sensitivity increased from 77.0% without AI to 85.0% with AI (8.0%; 95% CI: 2.6%, 13.4%; p <
0.01), specificity increased from 62.7% without AI to 69.6% with AI (6.9%; 95% CI: 3.0%, 10.8%; p < 0.01), and recall rate for
non-cancers decreased from 38.0% without AI to 30.9% with AI (7.2%; 95% CI: 3.1%, 11.2%; p < 0.01).
CONCLUSION
Concurrent use of AI improves cancer detection with increases of 0.057 in AUC, 8.0% in sensitivity, and 6.9% in specificity; and
decreases of 7.2% in recall rate and 52.7% in reading time.
CLINICAL RELEVANCE/APPLICATION

Radiologist's concurrent use of AI for DBT with certainty of finding scores increases detection of breast cancer with significant
reduction in reading time while improving sensitivity and specificity.
RC215-15 Breast Cancer Temporal Risk Prediction by Deep Learning and Longitudinal Digital Mammogram
Images
Monday, Nov. 26 11:30AM - 11:40AM Room: Arie Crow n Theater

Participants
Aly A. Mohamed, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
Wendie A. Berg, MD, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Dooman Arefan, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Jules H. Sumkin, DO, Pittsburgh, PA (Abstract Co-Author) Research Grant, Hologic, Inc; Research Grant, General Electric Company
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
Shandong Wu, PhD, MSc, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
wus3@upmc.edu
PURPOSE

Mammographic breast density is a risk factor and recent studies showed deep learning may identify more predictive imaging risk
features than breast density. We performed a study to investigate temporal breast cancer risk prediction by using deep learning
models on longitudinal 'normal' screening mammograms acquired prior to diagnosis of breast cancer.
METHOD AND MATERIALS
We conducted a retrospective case-control study on a cohort of 226 patients (1:1 case-control ratio) who underwent standard
mammographic screening at our institution during 2006-2013. The unilateral cancer cases (61.3±10.3YO) were all newly diagnosed
at 2013 and confirmed by pathology. Asymptomatic cancer-free controls (60.1±10.0 YO) are matched to the cancer cases by age
and year of the cancer-diagnosis imaging. All studied women did not have any prior biopsy or recall on mammography. For all
cohort, a set of sequential prior 'normal' (negative or benign findings) screening mammogram exams acquired during 2006-2012 were
collected (2-8 exams per patient), generating a total of 3263 'normal' images (913 for cancer cases, and 2350 for controls). Those
prior images of the cancer-affected breast (for cancer cases) and side-matched breast (for controls) were used to predict the
outcome (i.e., case/control status). We compared the prediction in terms of three time periods: (A) all priors from 2006 to 2012,
(B) recent priors (1548 images) from 2010 to 2012, and (C) distant priors (1715 images) from 2006 to 2009. The outcome prediction
was based on a pre-trained convolutional neural network model (ResNet-50) that was further fine-tuned on our mammograms. 10-
fold cross-validation and AUC were used to measure model performance.
RESULTS

81% of cancers and 82% of controls were post- with the rest pre-menopausal, and neither menopausal status nor family history of
breast cancer was associated with the outcome. AUC was 0.84 when using all priors, while it was 0.77 or 0.75 when using only the
recent or only the distant priors, respectively.
CONCLUSION
Sequential recent or distant prior 'normal' screening mammograms can predict, and their combination is more predictive of, breast
cancer development using deep learning models.
CLINICAL RELEVANCE/APPLICATION

Deep learning modeling on longitudinally acquired prior 'normal' screening mammogram images through up to 7 years earlier can
enhance temporal prediction of breast cancer development.
RC215-16 Novel Radiomic Descriptor of Tumor Vascular Morphology Identifies Responders to Neo-Adjuvant
Chemotherapy on Pre-Treatment Breast MRI
Monday, Nov. 26 11:40AM - 11:50AM Room: Arie Crow n Theater

Awards
Trainee Research Prize - Medical Student
Participants
Nathaniel Braman, Cleveland, OH (Presenter) Nothing to Disclose
Prateek Prasanna, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Maryam Etesami, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Donna M. Plecha, MD, Strongsville, OH (Abstract Co-Author) Research Grant, Hologic, Inc
Anant Madabhushi, PhD, Cleveland, OH (Abstract Co-Author) Research funded, Koninklijke Philips NV
For information about this presentation, contact:

nathaniel.braman@case.edu
PURPOSE
Despite significant interest in predicting treatment response prior to breast cancer neo-adjuvant chemotherapy (NAC) from DCE-
MRI, prior work has focused on textural patterns of the tumor or parenchyma or deep learning-based approaches that lack direct
biological interpretability. In this work, we introduce functional radiomic descriptors of vascular network disorder (VND) and evaluate
whether differences in the complexity of tumor-associated vasculature on pre-treatment DCE-MRI can discriminate between
patients who do and do not respond to NAC.
METHOD AND MATERIALS
1.5 or 3T DCE-MRI scans of 76 NAC recipients, 24 of whom had surgically confirmed pathological complete response (pCR), were
retrospectively analyzed. Average pixel width and slice thickness were .77 mm and 1.22 mm, respectively. Patients were randomly
divided into training (n=53, 14 pCR) and testing (n=23, 10 pCR) sets. A semi-interactive scheme was employed to segment the
tumor and vascular network. Within a sliding window, vessel orientation was computed for a series of 2-dimensional representations
of the vasculature relative to the tumor centroid. Statistics (mean, median, st. dev, skewness, and kurtosis) of the distribution of
vessel orientations for each representation were computed, yielding 20 VND features total. Top VND features were selected in the
training set using the Wilcoxon rank sum test via three-fold cross validation, then used to train a linear discriminant analysis
classifier to predict response in the test set. Performance was compared against (1) intra- and peri-tumoral texture features and
(2) a 3 layer LeNet convolutional neural network (CNN).
RESULTS
The top 4 VND features distinguished pCR with an AUC=0.75. pCR was characterized by reduced vascular disorder relative to non-
pCR. VND performed comparably or better than other state of the art radiomic approaches, including intra- and peri-tumoral texture
(AUC=.75) and deep learning (AUC=.67). Combining predictions from VND, texture features, and CNN yielded the best response
prediction accuracy (AUC=0.80).
CONCLUSION
VND features, which capture chaotic vessel network architecture, appear to be associated with NAC response and added predictive
value to established radiomic and deep learning approaches.
CLINICAL RELEVANCE/APPLICATION

Quantitative assessment of vessel network architecture as a functional radiomic biomarker could provide interpretable NAC response
prediction in breast cancer.
RC215-17 Using Machine Learning to Assess Tumor Metastatic Lymph Nodes and Ki-67 Expression
Aggressiveness from Breast MRI Using a Large Clinical Dataset of 300 Cancers from China
Monday, Nov. 26 11:50AM - 12:00PM Room: Arie Crow n Theater

Participants
Yu Ji, MD, Chicago, IL (Presenter) Nothing to Disclose
Hui Li, PHD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Alexandra V. Edwards, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc; Research Consultant,
Quantitative Insights, Inc
John Papaioannou, MSc, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc
Peifang Liu, MD, PhD, Tianjin, China (Abstract Co-Author) Nothing to Disclose
Maryellen L. Giger, PhD, Chicago, IL (Abstract Co-Author) Stockholder, Hologic, Inc; Shareholder, Quantitative Insights, Inc;
Shareholder, QView Medical, Inc; Co-founder, Quantitative Insights, Inc; Royalties, Hologic, Inc; Royalties, General Electric
Company; Royalties, MEDIAN Technologies; Royalties, Riverain Technologies, LLC; Royalties, Mitsubishi Corporation; Royalties,
Canon Medical Systems Corporation
For information about this presentation, contact:

yuji710@uchicago.edu
PURPOSE
To evaluate quantitative MRI radiomics in the task of identifying metastatic versus nonmetastatic axillary lymph nodes and Ki-67
expression aggressiveness.
METHOD AND MATERIALS

Our research involved a HIPAA-compliant, DCE-MRI database of 300 breast cancer cases. The average age was 47.2 years with a
standard deviation of 9.6 years and a range from 25 to 77 years with a median of 47 years. The clinical cohort included 48 low Ki-
67 expression (Ki-67 proliferation index < 14%) and 252 cases with high Ki-67 expression (Ki-67 proliferation index >= 14%),
indicating a range of tumor aggressiveness. The cohort also included 93 cases with axillary lymph node metastasis and 201 cases
without metastasis. The images had been obtained with a gadodiamide-enhanced T1-weighted spoiled gradient-recalled acquisition
in the steady state sequence. Primary lesions underwent computerized radiomic analysis in which tumor segmentation and
extraction were automatically conducted on an existing CADx workstation. These computer-extracted features included MRI-based
phenotypes from six categories: size, shape, morphology, enhancement texture, kinetics, and enhancement-variance kinetics.
Radiomic features were input to a Bayesian artificial neural network classifier (BANN) and underwent leave-one-case-out cross
validation. Area under the ROC curve (AUC) served as the figure of merit in the classification tasks.
RESULTS

In the task of identifying Ki-67 expression and lymph node status, the analyses of the various radiomic phenotypes yielded AUCs
ranging from 0.50 (se = 0.05) to 0.69 (se = 0.04). The Ki-67 MRI-based tumor signature produced an AUC value of 0.71 (se =
0.04). In the task of assessing the status of axillary lymph nodes, the radiomics tumor signature yielded an AUC value of 0.67 (se =
0.03). Both signatures were found to be statistically different from random guessing.
CONCLUSION

Quantitative MRI radiomics conducted on depicted primary breast tumors can contribute to identifying aggressive tumors, including
identifying Ki-67 expression and discriminating between metastatic and nonmetastatic lymph nodes, yielding automatic MRI-based
prognostic markers for ultimate use in radiogenomics and patient care.
CLINICAL RELEVANCE/APPLICATION

The ability to assess automatically the potential aggressiveness of tumors may elucidate the characteristics of breast cancers for
radiogenomics and for use in helping clinician estimate prognosis.
 RC221

Advances in CT: Technologies, Applications, Operations-Special Purpose CT

Monday, Nov. 26 8:30AM - 10:00AM Room: S102CD

BR MK CT IR PH
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Ehsan Samei, PhD, Durham, NC (Coordinator) Research Grant, General Electric Company; Research Grant, Siemens AG; Advisory
Board, medInt Holdings, LLC; License agreement, 12 Sigma Technologies; License agreement, Gammex, Inc
Lifeng Yu, PhD, Chicago, IL (Coordinator) Nothing to Disclose
ABSTRACT
CT has become a leading medical imaging modality, thanks to its superb spatial and temporal resolution to depict anatomical details.
New advances have enabled extending the technology to depict physiological information. This has enabled a wide and expanding
range of clinical applications. These advances are highlighted in this multi-session course. The course offers a comprehensive and
topical depiction of these advances with material covering CT system innovations, CT operation, CT performance characterization,
functional and quantitative applications, and CT systems devised for specific anatomical applications. The sessions include
advances in CT system hardware and software, CT performance optimization, CT practice management and monitoring, spectral CT
techniques, quantitative CT techniques, functional CT methods, and special CT use in breast, musculoskeletal, and interventional
applications.
Sub-Events
RC221A Breast CT Applications
Participants
John M. Boone, PhD, Sacramento, CA (Presenter) Patent agreement, Isotropic Imaging Corporation Consultant, RadSite
LEARNING OBJECTIVES
1) Introduce the technology of cone beam breast CT to audience. 2) Show both qualitative parameters describing image quality
and qualitative images. 3) Demonstrate breast CT performance using metrics such as anatomical noise metrics, computer and
human observer studies. 4) Illustrate the future potential of breast CT in diagnostic and screening breast imaging.
RC221B MSK CT Applications
Participants
Wojciech Zbijewski, PhD, Baltimore, MD (Presenter) Research Grant, Carestream Health, Inc; Research Grant, Siemens AG
For information about this presentation, contact:
wzbijewski@jhu.edu
LEARNING OBJECTIVES
1) Explain the technology of musculoskeletal (MSK) cone-beam CT (CBCT). 2) Identify key differences between MSK CBCT and
other orthopedic imaging modalities. 3) Discuss emerging clinical applications of MSK CBCT.
RC221C Interventional CT Applications
Participants
Christopher P. Favazza, PhD, Rochester, MN (Presenter) Nothing to Disclose
 VW 30

A Practical Approach to Breast Magnetic Resonance Imaging (MRI) Interpretation: An Interactive Session:
Siemens Healthineers Vendor Workshop
Monday, Nov. 26 10:15AM - 11:25AM Room: Booth 5530

Participants
Susan Weinstein, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Program Information
This interactive session will include both didactic and hands-on case review at workstations equipped with syngo® MR Brevis. A
practical approach to breast MRI interpretation will be discussed as well as utilizing the available sequences and techniques to
improve interpretive skills.
 VW 28

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Self-guided Reading
Sessions: Siemens Healthineers Vendor Workshop
Monday, Nov. 26 10:30AM - 5:00PM Room: Booth 5530

Program Information
You are invited to our self-guided reading sessions. With syngo® Breast Care workstations configured especially to allow you to
work at your own place at a time that suits you! A series of breast tomosynthesis and contrast enhanced mammography cases
presented as problem cases with a solution enables you to develop and test your reading skills.
 VW 29

Automated Breast Volume Scanner (ABVS) Self-guided Reading Sessions: Siemens Healthineers Vendor
Workshop
Monday, Nov. 26 10:30AM - 5:00PM Room: Booth 5530

Program Information
With syngo® Ultrasound Breast Analysis (sUSBA) software, self-guided reading sessions with real clinical cases will enable you to
become familiar with the coronal plane while providing practical approaches to interpretation of 3D automated breast ultrasound.
 VW 04

Introduction to Qview and QVCAD/ABUS Case Review - The AI Deep Learning Radiology Assist in Reviewing
GE Invenia ABUS Cases: GE Vendor Workshop
Monday, Nov. 26 11:00AM - 12:00PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration

http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 31

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Monday, Nov. 26 11:40AM - 12:50PM Room: Booth 5530

Participants
Terri A. Gizienski, MD, Greenwood Village, CO (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 VW 58

Improving Tomosynthesis Read-times, While Maintaining Clinical Performance: Hologic Vendor Workshop
Monday, Nov. 26 12:00PM - 1:00PM Room: Booth 5524

Program Information
Discussion and case review focuses on Hologic's new innovations to improve workflow efficiency, without sacrificing clinical
outcomes. (3D Mammography™ Technology, Clarity HD, Smart Mapping, SecurView® Workstations)
Registration
https://hologicrsna.com
 BRS-MOA

Breast Monday Poster Discussions

Monday, Nov. 26 12:15PM - 12:45PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

FDA Discussions may include off-label uses.

Participants
Lars J. Grimm, MD, Durham, NC (Moderator) Editorial Advisory Board, Medscape, LLC; Educational program support, Hologic, Inc
Sub-Events
BR229-SD- Ultrasound Radiomics in the Differentiation Between Benign and Malignant Breast Masses:
MOA1 Comparison Between Conventional Ultrasonography and Shear-Wave Elastography
Station #1
Participants
Ji Hyun Youk, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Jin Young Kwak, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Ju Son, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jeong-Ah Kim, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE
To evaluate whether ultrasound radiomics features can differentiate benign and malignant breast mass, compared conventional
ultrasonography (US) with shear-wave elastography (SWE).
METHOD AND MATERIALS
We retrospectively collected 328 pathologically-confirmed breast masses in 296 women who underwent US and SWE before biopsy
or surgery from March to November 2014. Their radiomics features were extracted from US and SWE images by texture analysis
algorithms in Matlab software. A representative SWE image for measuring elasticity of the index mass was selected for analysis,
which had both US and color-coded SWE image displayed in split-screen mode. For a manual lesion segmentation for all masses, an
ROI was delineated around the boundary of the index mass on US image from which the same ROI was copied and pasted to SWE
image by a dedicated breast radiologist. A total of 730 candidate radiomics features including first order statistics, textural features
(GLCM and GLRLM), and wavelet features, were extracted from each image. LASSO regression was used for data dimension
reduction and feature selection. Univariate and multivariate logistic regression analysis were performed to identify independent
radiomics features of differentiating between benign and malignant masses. The area under the receiver operating characteristic
curve (AUC) was calculated.
RESULTS
Of 328 breast masses, 205(62.5%) were benign and 123(37.5%) were malignant. Following radiomics feature selection, 22 features
from US and 6 features from SWE remained. On univariate analysis, all 6 radiomics features from SWE (P<0.0001) and 21 of 22
radiomics features from US (P<0.03) showed significant differences between benign and malignant masses. After multivariate
analysis, 3 radiomics features from US and 2 radiomics features from SWE were independently associated with malignant breast
masses. The AUC of logistic regression model in differentiating between benign and malignant masses was 0.929 for US and 0.992
for SWE (P<0.001).
CONCLUSION
Ultrasound radiomics features can differentiate between benign and malignant breast masses with good performance. SWE showed
significantly better performance than US.
CLINICAL RELEVANCE/APPLICATION
Quantitative ultrasound radiomics features can be used to diagnose breast cancer with good diagnostic performance and SWE
radiomics show superior performance to conventional ultrasonography radiomics.
BR230-SD- Impact of Inflammatory Breast Cancer Heterogeneity for Predicting Survival: Low-Dose Breast CT
MOA2 Texture Analysis
Station #2
Participants
Myoung-Ae Kwon, Ansan, Korea, Republic Of (Presenter) Nothing to Disclose
Bo Kyoung Seo, MD, PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation;
Research Grant, Guerbet SA; Research Grant, Koninklijke Philips NV;
Balaji Ganeshan, PhD, London, United Kingdom (Abstract Co-Author) CEO, Feedback plc; Director, Feedback plc; Director, Stone
Checker Software Ltd; Director, Prostate Checker Ltd
Eun Kyung Park, MD,PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ok Hee Woo, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Kyu Ran Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kwon4715@gmail.com
PURPOSE
Inflammatory breast cancer (iBC) is very aggressive and one of the non-measurable lesions in response evaluation criteria in solid
tumors, therefore measuring tumor size is inadequate to assess prognosis. The purpose of this study was to investigate the
usefulness of low-dose breast CT texture analysis (CTTA) to predict overall survival in patients with iBC.
METHOD AND MATERIALS
This retrospective study was approved by the institutional review board. From 2006 to 2017, fifty-five patients who had iBC and
pre-treatment low-dose breast CT were included. CTTA was performed on pre- and post-contrast CT images using a TexRAD
research software (TexRAD Ltd, www.texrad.com, Cambridge, UK). CTTA comprised a filtration-histogram technique where the
filtration step extracted and enhanced objects of different sizes corresponding to spatial scale filter (0-without filtration; 2mm-fine
texture scale; 3-5mm-medium texture scales; 6mm-coarse texture scale) and variation in density followed by quantification of
texture using histogram based analysis (mean-M, standard deviation-SD, entropy-E, mean of positive pixels-MPP, skewness-S, and
kurtosis-K). Delta texture i.e. difference between post-contrast and pre-contrast texture was also evaluated. The relationship
between CT texture features and overall survival was assessed using Kaplan-Meier survival analysis.
RESULTS
27/55 patients died within a mean follow-up period of 29.7 (0-100) months. Median survival was 32.0 (95% CI: 3.7-60.3) months. A
number of CTTA metrics predicted survival. Higher SD and E without filtration and E across fine to medium texture scales on pre-
contrast CT were significant predictors of poor prognosis (best - E at fine texture scale >=5.55, p=0.003). Lower M and S without
filtration and higher M and E across fine to medium texture scales on post-contrast CT were significant predictors of poor prognosis
(best - E at fine texture scale >=5.91, p=0.0004). Delta texture (Lower E and higher K) without filtration and across fine, medium
and coarse texture scales were significant predictors of poor prognosis (best - Delta K at fine texture scale >=0.08, p=0.0002).
CONCLUSION

Low dose CTTA can be useful to predict survival in patients with iBC.
CLINICAL RELEVANCE/APPLICATION

Filtration histogram based texture analysis increases the utility of low-dose breast CT by reflecting tumor heterogeneity and
demonstrates prognostic potential in inflammatory breast cancer.
BR231-SD- Clinical Impact of Second Opinion Radiology Consultation for Patients with Breast Cancer
MOA3
Station #3
Awards
Student Travel Stipend Award
Participants
Debra S. Whorms, MD, Boston, MA (Presenter) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
Mehra Golshan, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Rachel Freedman, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Craig A. Bunnell, MD, MPH, Boston, MA (Abstract Co-Author) Nothing to Disclose
Emily C. Alper, BA, Boston, MA (Abstract Co-Author) Nothing to Disclose
Katya Losk, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Ramin Khorasani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
dswhorms@gmail.com
PURPOSE
Assess the incidence and clinical significance of:interpretation discrepancies when subspecialty consult review of outside imaging is
performed at a National Cancer Institute (NCI)-designated tertiary cancer center, for patients with newly diagnosed breast cancer.
METHOD AND MATERIALS

This Institutional Review Board-approved retrospective observational study included patients presenting 7/2016-3/2017 to an NCI-
designated comprehensive cancer center for second opinion after a breast cancer diagnosis. Outside and second-opinion radiology
reports of 252 randomly-selected patients were compared by two subspecialty breast radiologists to consensus. A peer review
score was assigned using a modification of the American College of Radiology's RADPEER peer review metric: 1-agree; 2-minor
discrepancy (unlikely clinically significant); 3-moderate discrepancy (may be clinically significant); 4-major discrepancy (likely
clinically significant). Among cases with clinically significant discrepancies, rates of clinical management change (alterations in
management including change in follow-up, use of neoadjuvant therapy, and surgical management as a direct result of image
review) and detection of additional malignancy were assessed through electronic medical record review.
RESULTS

A significant difference (RADPEER 3 or 4) in interpretation was seen in 41 cases (16%, 95% Confidence Interval [CI], 11.7%-
20.8%). The difference led to additional work-up in 38 cases (15%, 95% CI 10.6%-19.5%) and a change in clinical management in
18 cases (7.1%, 95% CI 4.0%-10.2%), including 15 cases with a change in surgical management (6.0%, 95% CI, 3.0%-8.9%)
cases. An additional malignancy or larger area of disease was identified in 12 cases (4.8%, 95% CI, 2.1%-7.4%).
CONCLUSION
Discrepancy between outside and second-opinion radiologists frequently results in additional work-up for breast cancer patients,
changes in treatment plan, and identification of new malignancies.
CLINICAL RELEVANCE/APPLICATION

Formal second opinion imaging consultation in patients with newly diagnosed breast cancer presenting to a multidisciplinary breast
oncology program has significant value in identifying additional malignancy and optimizing treatment approach.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
BR232-SD- Prediction of 21-Gene Recurrence Score in Patients with Estrogen Receptor-positive Early-Stage
MOA4 Breast Cancer Using MRI-based Radiomics Nomogram
Station #4
Participants
Nam Joo Lee, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa Jung Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ki Chang Shin, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jong Won Lee, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sae Byul Lee, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
docshin@amc.seoul.kr
PURPOSE
To develop a breast MRI-based radiomics nomogram including pathologic factors which can predict low-risk recurrence score (RS)
on 21-gene RS assay in patients with estrogen receptor-positive early-stage breast cancer (EBC).
METHOD AND MATERIALS

From 2011 to 2017, a total of 547 tumors in 539 patients with EBC who underwent preoperative breast MRI were retrospectively
included in this study. Among them, low-risk was 320 (58.5%), intermediate-risk was 180 (32.9%), and high-risk was 47 (8.6%).
We extracted 744 quantitative MR radiomic features from computerized three-dimensional segmentations of each tumor generated
computer-extracted image phenotypes (CEIP) within the intratumoral regions of early post-contrast T1-weighted images, percent
enhancement (PE) map, signal enhancement ratio (SER) map, and T2-weighted images. We divided 547 cases into a training set
(n=365) and a validation set (n=182). Elastic net was used for feature selection and radiomics score building. Multivariate logistic
regression analysis was used to develop a prediction model, we incorporated the radiomics score and independent pathologic risk
factors and build a radiomics nomogram. Internal validation for an independent validation set (n=182) was performed.
RESULTS
The radiomics score, which consisted of 24 selected CEIPs, was significantly associated with the prediction of recurrence (C-index,
0.769 for training set and 0.745 for validation set). Independent pathologic predictors contained in the nomogram were
progesterone receptor status, nuclear grade, histologic grade, extensive intraductal component, lymphovascular invasion, P53, and
Ki67 status, and their C-index was 0.858 for training set and 0.774 for validation set. Addition of radiomics score to the pathologic
nomogram showed an incremental value of 0.054 and 0.092, respectively. Radiomics nomogram showed good prediction of low-risk
RS, with a C-index of 0.912 for training set and 0.866 for validation set.
CONCLUSION

This study shows that a radiomics nomogram which incorporates the MRI-based radiomics score and pathologic features, can be
used to help the preoperative individualized prediction of low-risk RS in patients with EBC.
CLINICAL RELEVANCE/APPLICATION

Prediction nomogram using breast MRI-based radiomics score and pathologic predictors can be used to facilitate the preoperative
individualized prediction of low-risk RS on 21-gene RS assay in patients with EBC.
BR233-SD- Volumetric versus Area-based Breast Density Assessment: Comparisons using Fully-Automated
MOA5 Quantitative Measurements in a Large Screening Population
Station #5
Participants
Aimilia Gastounioti, Philadelphia, PA (Presenter) Nothing to Disclose
Meng-Kang Hsieh, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Eric A. Cohen, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Yifan Hu, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
aimilia.gastounioti@uphs.upenn.edu
PURPOSE

Breast density is an important factor affecting the sensitivity of screening mammography due to the masking of tumors within the
dense breast tissue, therefore, substantially influencing supplemental screening recommendations for women. We investigate
associations between fully-automated quantitative measures of area-based (ABD) and volumetric (VBD) breast density versus
categorical (BI-RADS) density assessment and potential implications for supplemental screening.
METHOD AND MATERIALS

We retrospectively analyzed bilateral digital mammograms (Selenia Dimensions, Hologic Inc.) from an entire one-year screening
cohort at our institution (2012-2013; N = 11,107 women). BI-RADS density categories, visually assigned to each study by the
reviewing radiologist, were collected from archived screening reports. Fully-automated software (QuantraTM v2.2; Hologic Inc.)
was used to extract quantitative measures of ABD, VBD and categorical, BI-RADS-like, density categories. Pair-wise correlation (r)
and linear regression were used to investigate the relationship between ABD and VBD measurements, adjusted for age, BMI and
race (41% White, 50% Black, 9% Other/Unknown). Analysis of variance (ANOVA) was performed to evaluate ABD and VBD variation
by the clinical BI-RADS density categories. Agreement between the radiologist's and the software-assessed density categories was
measured via the Cohen's weighted kappa (k).
RESULTS
VBD (12.5% ± 6.9%) and ABD (16.5% ± 16.0%) were strongly correlated (r = 0.95), with VBD being lower than ABD (b = 0.41, 95%
CI = [0.40, 0.41]). Both density measures were significantly different across BI-RADS density categories with an increase as BI-
RADS density increased (ANOVA, p < 0.001). There was moderate agreement between the radiologists' and the software-assessed
breast density categories (k = 0.67, SE = 0.01), on par with reported inter-reader variability between radiologists.
CONCLUSION

VBD measurements are strongly correlated to ABD estimates in digital mammography; yet, reported breast density is likely to be
lower in the volumetric evaluation. The software-generated breast density scores moderately agree with clinical BI-RADS density
readings.
CLINICAL RELEVANCE/APPLICATION
Refinement of density thresholds that prompt supplemental screening is likely to be needed for clinical adoption of automated
density measures, adjusting also for area-based vs. volumetric evaluation.
BR234-SD- Breast Cancer Risk Assessment at the Time of Screening Mammography: Pathology and BIRADS
MOA6 Outcomes
Station #6
Awards
Student Travel Stipend Award

Participants
Diana Murcia, MD, Burlington, MA (Presenter) Nothing to Disclose
Megan E. Mcdevitt, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Maryam Shahrzad, MD, Arlington, MA (Abstract Co-Author) Nothing to Disclose
Cathleen M. Kim, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Jeanette Y. Chun, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Michelle R. McSweeney, DO, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Meera Sekar, MD, Lexington, MA (Abstract Co-Author) Nothing to Disclose
Meaghan Mackesy, MD, Burlington, MA (Abstract Co-Author) Nothing to Disclose
Audrey L. Hartman, MD, MS, Burlington, MA (Abstract Co-Author) Nothing to Disclose
PURPOSE

To evaluate the early experience with breast cancer risk assessment (CRA) in a multicenter breast screening program, utilizing
pathology and BIRADS outcomes for patients stratified into high and low risk groups.
METHOD AND MATERIALS

Patients were offered to complete a commercially available, tablet-based CRA survey at the time of the screening mammogram. The
software utilized 3 risk prediction models; BRCAPRO (high risk >=20%), Tyrer-Cuzick v.7 (high risk >=20%), and BRCAMUTATION
(high risk >=5%). The patient was identified as high risk (HR) if any of the 3 models predicted high risk. Otherwise, the patient was
labeled low risk (LR). A HIPAA compliant and IRB approved retrospective review of all patients offered the survey in a 6 month
period was performed. Inclusion criteria: completion of risk prediction survey, BIRADS 0 on screening mammogram, completion of
additional imaging (diagnostic mammogram and/or focused ultrasound), resulting in BIRADS 4 or 5, and completion of core biopsy
(and surgical biopsy, if indicated).The rate of BIRADS 0, cancer detection rate (CDR), and percentage of invasive cancer were
compared for the HR and LR patients, using a chi-square test. CDR and Positive Predictive Value 3 for breast cancer (PPV3) were
calculated for both HR and LR groups.
RESULTS

A total of 10394 patients underwent screening mammogram, of which 88.5% (9200) elected to complete the CRA survey, of which
14% (1293) of patients were HR, and 86% (7897) were LR. The percent of BIRADS 0 at screening was higher for HR than LR (11.2%
vs 7.7 %, p<0.001). The cancer detection rate (per 1000 patients) was higher for HR than LR (7.8 vs 3.8 p<0.05). The PPV3 was
not statistically different (34.5% for HR, 25.2% for LR, P=0.37). The percent of invasive cancers was also not statistically different
between the HR and LR patients.
CONCLUSION

Patients identified as high risk by breast cancer risk assessment at the time of screening mammography had a higher cancer
detection rate than low risk patients.
CLINICAL RELEVANCE/APPLICATION

Breast cancer risk assessment at screening mammogram may provide information towards personalized screening for breast cancer.
BR176-ED- Contrast-Enhanced Mammography: Current Applications and Future Direction
MOA7
Station #7
Awards
Identified for RadioGraphics

Participants
Kimeya Ghaderi, MD, Boston, MA (Presenter) Nothing to Disclose
Jordana Phillips, MD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Consultant, General Electric
Company
Hannah Perry, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Parisa Lotfi, MD, Newton, MA (Abstract Co-Author) Nothing to Disclose
Tejas S. Mehta, MD, MPH, Boston, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

kghaderi@bidmc.harvard.edu
TEACHING POINTS
Contrast-Enhanced Mammography (CEM) employs a dual energy technique to create images that show abnormal morphology and
density (similar to conventional mammography) and abnormal enhancement (like MRI). CEM is FDA approved for adjunct use in the
diagnostic setting. Diagnostic applications include evaluation of screening callbacks, the symptomatic breast, disease extent in
women with known breast cancer, and response to neoadjuvant chemotherapy. Key clinical indication of CEM is as an MRI
alterative demonstrating similar sensitivity and positive predictive value. Benefits include reduced patient cost and relative ease of
access compared to MRI. Challenges include IV placement, risks associated with contrast in mammography setting, and added time.
TABLE OF CONTENTS/OUTLINE

Background: What is CEM and how are images acquired? FDA approved indications and literature review Screening recall:
Architectural distortion, mass, calcifications Follow-up diagnostic imaging Symptomatic breast Evaluation of disease extent
Response to neoadjuvant chemotherapy Mammographically occult malignancy Troubleshooting When MRI cannot be performed
Supplemental screening Other potential uses: High-risk screening Benefits and Challenges of using CEM for these indications
BR177-ED- Factors Affecting MRI Accuracy in Evaluating of Neoadjuvant Therapy Response in Breast Cancer: A
MOA8 Pictorial Review with Radiopathological Correlation
Station #8
Awards
Certificate of Merit

Participants
Gulgun Engin, Istanbul, Turkey (Presenter) Nothing to Disclose
Semen Onder, MD, ?stanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Inci Kizildag Yirgin, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Ekrem Yavuz, MD, ?stanbul, Turkey (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

gengin@istanbul.edu.tr
TEACHING POINTS
To review tumor shrinkage patterns subsequent to NAC To discuss the factors affecting MRI accuracy in evaluating of NAC
response To improve the knowledge about the detection accuracy of residual disease after NAC by MRI
TABLE OF CONTENTS/OUTLINE
Preoperative neoadjuvant chemotheraphy (NAC) is widely performed for patients with locally-advanced breast carcinomas in order
to control and downstage the primary breast carcinomas. Breast magnetic resonance imaging (MRI) is the most accurate imaging
modality for assessment of tumor response to NAC. However, MRI often over- or underestimate the extent and distribution of
residual carcinomas following NAC. The accuracy of MRI in evaluating residual tumor size and extent depends on tumor type,
morphology, and biomarker status. Cancers with lobular component, non-mass-like enhancement lesions, HR-positive, and HER2-
negative cancers had an increased discrepancy between MRI versus pathology tumor size (Figures 1-5). Improved knowledge about
the detection accuracy of residual disease after NAC by MRI is crucial to help planning of an optimal surgery to achieve a tumor
free margin. This can reduce the re-excision rate and minimize local recurrence. In this pictorial review, the factors affecting MRI
accuracy in evaluating of NAC response are analysed and discussed with detailed radiopathological correlation.
BR178-ED- Pitfalls, Tricks, and Tips in the Sonographic Evaluation of Breast. What the Radiology Residents Need
MOA9 to Know?
Station #9
Participants
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Maria B. Orruma, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Ana G. Luna, CABA, Argentina (Abstract Co-Author) Nothing to Disclose
Diana Herbas Galindo, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Pamela I. Causa Andrieu, MD, La Plata, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

drakarina.pesce@gmail.com
TEACHING POINTS
1. Review of the anatomy and the factors contributing to the difficulties in the evaluation in the breast ultrasound. 2. Describe
1. Review of the anatomy and the factors contributing to the difficulties in the evaluation in the breast ultrasound. 2. Describe
common artefacts and image quality issues. Discuss common pitfalls and limitations. 3. Illustrate several examples of the technical
artefacts, with helpful suggestions to help optimize images and avoid misinterpretation. 4. Learn tips to optimize images techniques
and demonstration of some pitfalls.
TABLE OF CONTENTS/OUTLINE

1. Introduction: We present a review of common pitfalls and troubleshooting. Tips for radiology residents who perform and interpret
breast ultrasound. 2. Overview of breast ultrasound protocol 3- Classification of pitfalls A) Anatomical B) Sonography physics C)
Instrumentation D) Cross-Correlation with mammographic lesions E) Foreign bodies. 4-Sonography Physics: Artefacts in breast
ultrasound a-Gray scale ultrasound b- Artefacts in doppler ultrasound color 5- Tips to optimize imaging techniques, relevant
pathologic findings and demonstration of some pitfalls. 6-Conclusions
BR179-ED- Breast Imaging STAT: A Pictorial Review of Acute Pathology
MOA10
Station #10
Participants
Rutuparna Sarangi, MD, Boston, MA (Presenter) Nothing to Disclose
Anna Rives, MD, PhD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Anastasia F. Barron, DO, Boston, MA (Abstract Co-Author) Nothing to Disclose
Allyson L. Chesebro, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Kitt Shaffer, MD, PhD, Boston, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
anna.rives@bmc.org
TEACHING POINTS
1. Review the types of acute breast pathology commonly seen in the emergency setting. 2. Describe the imaging features of each
entity using case-based examples. 3. Discuss current best practice for management and/or treatment.
TABLE OF CONTENTS/OUTLINE
1. Review clinical and imaging features of mastitis and breast abscess. a. Puerperal b. Non-puerperal c. Granulomatous mastitis d.
Pitfall: inflammatory breast cancer 2. Summarize evidence-based breast abscess management and treatment algorithms for the
radiologist, with emphasis on ultrasound-guided drainage. 3. Present imaging manifestations of accidental breast trauma, including
hematoma and fat necrosis. Pitfall: breast cancer presenting after trauma. 4. Illustrate acute post-procedural complications
following percutaneous breast biopsy and surgery. 5. Describe additional examples of acute breast pathology including Mondor's
disease and inflamed epidermal inclusion cyst.
BR180-ED- Primary Breast Tuberculosis: Imaging Findings
MOA11
Station #11
Participants
Ali H. Baykan, MD, Adiyaman, Turkey (Abstract Co-Author) Nothing to Disclose
Ibrahim Inan, MD, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Hakan S. Sayiner, Adiyaman, Turkey (Abstract Co-Author) Nothing to Disclose
Ebru Hasbay, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Mustafa Goksu, Adiyaman, Turkey (Abstract Co-Author) Nothing to Disclose
Safiye Kafadar, Adiyaman, Turkey (Abstract Co-Author) Nothing to Disclose
Bilge Aydin Turk, Adiyaman, Turkey (Abstract Co-Author) Nothing to Disclose
Sukru Mehmet Erturk, Adiyaman, Turkey (Presenter) Nothing to Disclose
For information about this presentation, contact:

drbaykan@gmail.com
TEACHING POINTS

The purpose of this exhibit is: 1. To provide general information about breast tuberculosis in order to raise awareness. 2. To review
radiologic findings of breast tuberculosis with some sample cases. 3. To discuss difficulties in the diagnosis of breast tuberculosis
and multidisciplinary approach in the management of treatment.
TABLE OF CONTENTS/OUTLINE
1. General information about breast tuberculosis 2. Review of imaging findings with sample cases 3. The role of radiology in the
multidisciplinary approach 4.Summary
BR181-ED- Probably Benign Assessment on Breast MRI: Appropriate and Inappropriate Utilization
MOA12
Station #12
Participants
Mirelys Barrios, MD , Boston, MA (Presenter) Nothing to Disclose
Sona A. Chikarmane, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Emily Mungovan, BA, Boston, MA (Abstract Co-Author) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
schikarmane@bwh.harvard.edu
TEACHING POINTS

ACR BI-RADS Category 3 ('probably benign') should carry <= 2% likelihood of malignancy and was initially defined for a normal risk
screening mammography population. The ACR BI-RADS Lexicon states that probably benign lesions should have a very high
probability of being benign and many approaches for assessment are intuitive. The objectives of this exhibit are: 1. To review the
evolving criteria for findings to be placed in probably benign category on breast MRI, with emphasis on challenges in daily practice.
2. To understand various clinical and patient factors that may affect management and assessment, including high risk patient
populations, younger patient population, and baseline studies.
TABLE OF CONTENTS/OUTLINE

1. To review the published literature on probably benign (BI-RADS 3) lesions on breast MRI. 2. To provide an overview on breast
MRI findings currently accepted as probably benign. 3. To present a pictorial review of upgraded BIRADS 3 lesions, benign and
malignant, with assessment and management lessons. 4. To review appropriate and inappropriate uses of probably benign category
in breast MRI, with emphasis on challenging clinical and patient factors that may affect management.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
 VW 07

Application of ABUS in Diagnosis of Breast Cancer: A Multi-Center Study in China: GE Vendor Workshop
Monday, Nov. 26 12:30PM - 1:00PM Room: Booth 8156

Participants
You-Lin Qiao, Beijing, China (Presenter) Nothing to Disclose
Program Information
In this presentation, Dr. Qiao will discuss his experience as Principal Investigator of a multi-center study in China aimed at
evaluating the diagnosis accuracy of Automated Breast Ultrasound to traditional ultrasound and mammography. Registration is
required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click the link below to
register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 BRS-MOB

Breast Monday Poster Discussions

Monday, Nov. 26 12:45PM - 1:15PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

FDA Discussions may include off-label uses.

Participants
Lars J. Grimm, MD, Durham, NC (Moderator) Editorial Advisory Board, Medscape, LLC; Educational program support, Hologic, Inc
Sub-Events
BR235-SD- Does the Tumor Stiffness of Shear Wave Elastography Correlate with Tumor Hypoxia or Fibrosis of
MOB1 Breast Cancer?
Station #1
Participants
Myoung-Ae Kwon, Ansan, Korea, Republic Of (Presenter) Nothing to Disclose
Bo Kyoung Seo, MD, PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation;
Research Grant, Guerbet SA; Research Grant, Koninklijke Philips NV;
Eun Kyung Park, MD,PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jaehyung Cha, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ok Hee Woo, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Kyu Ran Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sung Eun Song, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kwon4715@gmail.com
PURPOSE
Hypoxic and stiff microenvironments of breast cancer promote metastasis and resistance to therapy. The purpose of this study was
to investigate whether the tumor stiffness parameters of shear wave elastography (SWE) are related with tumor hypoxia or fibrosis
in invasive breast cancer.
METHOD AND MATERIALS

This retrospective study was approved by the institutional review board. From June 2016 to January 2018, eighty-two women with
invasive breast cancers who underwent SWE before treatment were enrolled. We used Aplio 500 US equipment (Toshiba Medical
Systems Corporation, Japan). On the SWE features, average tumor elasticity (Eaverage) and tumor-to-fat elasticity ratio (Eratio)
were extracted. Glucose transporter 1 (GLUT1) for evaluation of tumor hypoxia and Masson's trichrome staining (MT) for fibrosis
were used. Immunostaining and automated digital image analysis were performed for assessment of GLUT1 and MT activities.
Correlations analysis was performed between SWE parameters (Eaverage and Eratio) and GLUT1 or MT activities using spearman's
correlation. Correlations with biomarkers according to molecular subtype, tumor grade, lymph node, hormone receptor, HER2, and
Ki67 were compared using t-test, Mann-Whitney test, and Kruskal-Wallis test with Bonferroni correction.
RESULTS
SWE parameters including Eaverage (r=0.676) and Eratio (r=0.411) were significantly correlated with GLUT1 activities (P<0.001 for
all). On the other hand, SWE parameters were not related with MT activities (P>0.05 for all). Eaverage values were significantly
higher in breast cancers with positive lymph node, negative hormone receptor, high Ki67, and high grade (P<0.03 for all). Eratio
values were higher in breast cancers with high Ki67 and high grade (P<0.05 for all). Eaverage values were different according to
molecular subtypes of breast cancer (P=0.009). Eaverage values were significantly higher in triple negative cancers than those in
luminal A cancers (P=0.03).
CONCLUSION
Tumor stiffness measured by SWE is associated with tumor hypoxia and biomarkers that affect tumor prognosis and it not related to
fibrous components.
CLINICAL RELEVANCE/APPLICATION
Tumor stiffness parameters as measured by SWE is significantly correlated with tumor hypoxia and histologic biomarkers and can be
used to predict prognosis of invasive breast cancer.
BR236-SD- Therapeutic Nipple Sparing Mastectomy (NSM): Tumor-To-Nipple Distance (TND) on Preoperative
MOB2 MRI as Useful Variable in Surgical Patient Selection and Outcomes
Station #2
Participants
Jiyon Lee, MD, New York, NY (Presenter) Nothing to Disclose
Kristin L. Harris, DO, New York, NY (Abstract Co-Author) Nothing to Disclose
Jordan Frey, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Ara Salibian, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Mihye Choi, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Nolan Karp, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Deborah Axelrod, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
jiyon.lee@nyumc.org
PURPOSE
Newer surgical oncological nipple-sparing mastectomy (NSM) enables nipple-areola complex (NAC) preservation. Patient selection
criteria and oncologic outcomes data are being established. We evaluated tumor-to-nipple distance (TND) on preop MR in NSM
patients to help determine cancer recurrence risk factors.
METHOD AND MATERIALS
Retrospective institutional review of 2/2006- 7/2017 yielded 496 therapeutic NSMs in 312 women (mean f/u 48.25 m).
Demographics, outcomes, trends, and cancer recurrences were reviewed. TND was measured on available preop MRIs using axial,
sagittal, or multiplanar reconstruction images. Univariate analysis identified independent risk factors for cancer recurrence. p-values
<0.05 significant.
RESULTS
Mean patient age was 48.66 yo (23-65 y). 74.2% of NSMs were part of bilateral surgery. Most common were: IDC (52.4%) and
DCIS (50.4%); and stage IA (42.5%) and stage 0 (31.3%) disease. Mean tumor size was 1.48 cm. 25.2% of NSMs had multifocal
disease, 11.5% had LVI; 59.9% were ER+, 56.3% were PR+, and 42.5% were HER 2-neu +. SLN biopsy performed in 79.8% of
cases. Positive subareolar biopsy on 6.4% frozen section and 6.7% permanent biopsy. Per NSM, the rate of local in-breast
recurrence was 1.6% (N=8) and regional was 0.6% (N=3). Per patient, rates were 2.6% local, 1.0% regional, and 1.3% (N=4)
distant mets. One local recurrence (12.5%) had positive permanent subareolar biopsy treated with NAC resection. Preop MRI in 171
NSMs showed mean TND 4.78 cm. TND did not signif differ between NSMs with and w/o locoregional recurrence (4.62 vs 4.78 cm;
p=0.8758). However, NSMs with TND <=1 cm (25.0% vs 2.4%, p=0.0031) and <=2 cm (8.7% vs 2.0%; p=0.0218) were signif
associated with locoregional recurrence. In univariate analysis, TND <=1 cm was the only signif risk factor (OR=13.5833, p=0.0385).
Age <50 years (p=0.0503) and multifocal disease (p=0.0820), TND <=2 cm (p=0.1052), and positive permanent subareolar biopsy
(p=0.1094) trended towards association with higher recurrence risk.
CONCLUSION

NSM had locoregional recurrence of 2.0%. TND of <=1 cm and <=2 cm were significantly associated with recurrence risk. TND <=1
cm was significant predictor of locoregional recurrence on univariate analysis.
CLINICAL RELEVANCE/APPLICATION
NSM is oncologically safe with low locoregional recurrence of 2.0% in appropriately selected patients. TND of <=1 cm and <=2 cm
on preoperative MRI can be useful variables in predicting recurrence.
BR237-SD- Determining the Appropriateness of Second Look Targeted Sonography Recommendation Following
MOB3 Detection of Suspicious Non-Mass Enhancement (NME) Detected on Breast Magnetic Resonance
Imaging (MRI)
Station #3
Participants
Jennifer J. Young, MD,MPH, Los Angeles, CA (Presenter) Nothing to Disclose
Iram Dubin, MD , Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Sonya Khan, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Meghan Jardon, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Kara-Lee Pool, MD , Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Melissa M. Joines, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

mjoines@mednet.ucla.edu
PURPOSE

There are currently no evidence-based guidelines detailing the appropriateness of recommending targeted sonography versus
proceeding directly to MRI-guided biopsy for suspicious non-mass enhancement (NME) detected on breast MRI. The purpose of this
study is to determine if characteristics of suspicious NME are associated with detection of appropriate sonographic correlates on
second look ultrasound.
METHOD AND MATERIALS
IRB approved, retrospective review of 4,292 contrast enhanced screening and diagnostic bilateral breast MRI examinations
performed on female patients (ranging in age from 20 to 87 years old) at our institution between January 1, 2012 and December 31,
2015 was performed, focusing on the subset of 134 studies in which targeted sonography was recommended and subsequently
performed for suspicious NME detected on MRI. Multivariate linear regression analysis (p > 0.05) was employed to analyze patient
demographics as well as imaging study findings and characteristics in order to determine the likelihood of identifying an appropriate
sonographic correlate on targeted sonography performed for suspicious NME.
RESULTS

Sonographic correlates were detected in 26% (35/134) of targeted ultrasounds performed for suspicious NME (all subtypes).
Multivariate linear regression analysis demonstrating that suspicious linear NME seen on screening MRI was significantly associated
with negative targeted sonography, coefficient -0.0787454 (p = 0.05, CI [-0.1576416, 0.0001507]), while segmentally distributed
NME was significantly associated with detection of a sonographic correlate, coefficient 0.6343792 (p < 0.0001, CI [0.5294839,
0.7392745]). NME seen in the same quadrant as an enhancing mass on MRI was also significantly associated with detection of a
sonographic correlate, coefficient 0.7905429 (p < 0.0001, CI [ 0.6914279, 0.889658]).
CONCLUSION
This study demonstrates that specific characteristics of suspicious NME are differentially associated with detection of sonographic
correlates on second look ultrasound, findings that can be further studied to guide appropriateness criteria for targeted sonography
after detection of suspicious NME on breast MRI.
CLINICAL RELEVANCE/APPLICATION

Developing appropriateness criteria for targeted sonography for suspicious NME detected on MRI aides patient care by reducing
unnecessary second look ultrasound examinations, health care costs, and delay in time to diagnosis.
BR238-SD- Prediction of Neoadjuvant Chemotherapy Response using Radiogenomics of MR Imaging in Breast
MOB4 Cancer: Correlation of Quantitative and Qualitative MR findings with Results of Multi-Gene Classifier
Station #4
Participants
Yukiko Tokuda, MD,PhD, Suita, Japan (Presenter) Nothing to Disclose
Kaori Minamitani, MD, Suita, Japan (Abstract Co-Author) Nothing to Disclose
Masahiro Yanagawa, MD, PhD, Suita, Japan (Abstract Co-Author) Nothing to Disclose
Yasuto Naoi, MD,PhD, Suita, Japan (Abstract Co-Author) Research Grant, Sysmex Corporation; Speaker, Sysmex Corporation
Shinzaburo Noguchi, MD,PhD, Suita, Japan (Abstract Co-Author) Research funded, Sysmex Corporation Speaker, Sysmex
Corporation
Noriyuki Tomiyama, MD,PhD, Suita, Japan (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation
PURPOSE
To examine the correlation of magnetic resonance imaging (MRI) features with results of multi-gene classifier for predicting
neoadjuvant chemotherapy (NAC) response of breast cancer.
METHOD AND MATERIALS
This study included 120 patients with breast cancer who were classified into response (R) group (n=46) and non-response (NR)
group (n=74) by using the immune-related 23-gene signature for NAC (IRSN23) for NAC sensitivity prediction. All patients had
undergone a dynamic contrast-enhanced breast MRI (DCE-MRI). For quantitative data, the following DCE-MRI features were
measured: volume ratio of each fast, medium and slow to whole mass in the initial phase; volume ratio of each washout, plateau
and persistent to whole mass in the delayed phase; and both kurtosis and skewness of intensity histogram in whole mass on each
phase. Mass size and volume were measured. For qualitative data, two breast radiologists independently interpreted and decided
the findings by consensus reading. The value in examining associations with R- and NR-group were analyzed using univariate logistic
regression (ULR). Significant parameters identified by the URL analysis were included in the multiple logistic regression (MLR)
analysis. Each binary group of quantitative data was designated by a cutoff value decided by receiver-operating characteristic
analysis.
RESULTS

ULR analysis revealed that volume ratio of slow, medium, washout, persistent, medium-washout, medium-plateau, and slow-
persistent volume ratio, volume of the whole mass, skewness in the delayed phase, mass shape, mass margin, and multiple masses
were significant indicators to divide into R- and NR- group. MLR analysis revealed that volume ratio of slow in the initial
phase>34.4% and medium-washout<=4.5%, and round/oval shape were significant indicators associated with R-group (Odds ratio,
3.76, 4.16, and 4.35; 95% confidence interval, 1.44 to 9.86, 1.78 to 9.76, and 1.72 to 11.03; p=0.007, p=0.001, and 0.002,
respectively).
CONCLUSION

Volume ratio of slow >34.4% in the initial phase and medium-washout <=4.5% to the mass, and round/oval shape were found to be
independent indicators associated with R-group.
CLINICAL RELEVANCE/APPLICATION
Both quantitative and qualitative data of dynamic contrast-enhanced breast MRI might contribute to the prediction of neoadjuvant
chemotherapy response of breast cancer using multi-gene classifier.
BR239-SD- Breast Cancer and Breast Imaging Malpractice Litigation: A 10-Year Analysis and Update in Trends
MOB5
Station #5
Awards
Student Travel Stipend Award

Participants
Katerina Konstantinoff, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Catherine M. Appleton, MD, Saint Louis, MO (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc Royalties, Oxford University
Press
Alison R. Gegios, MD, St. Louis, MO (Abstract Co-Author) Nothing to Disclose
Katie M. Miles, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Dawn Hui, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

konstantinoff@wustl.edu
PURPOSE

The purpose of this study is to evaluate factors contributing to breast cancer and breast imaging-related medical malpractice
cases.
METHOD AND MATERIALS

A retrospective analysis of jury verdict and settlement reports in US state and federal courts on the Westlaw legal database was
performed. The database was searched for 'malpractice' and 'breast cancer' related terms from 2005 to 2015. 253 cases were
evaluated for factors including outcome, award amount, alleged causes of malpractice, patient demographics, defendant
specialities, and other alleged factors claimed in the litigation. Data were summarized using descriptive statistics. Logistic regression
was used to evaluate associations between factors and plaintiff award.
RESULTS
Median plaintiff age was 46 (IQR 39, 56). In cases that resulted in plaintiff payment, award amount was $978,858±2,308,598. Delay
in diagnosis was cited as a reason for claimed negligence in 82% of cases. Mean length of delay was 17±13 months. Named
defendants were radiologists (43%), surgeons (27%), obstetrician/gynecologists (26%), and internal medicine/family practice
(15%). Age, defendant type, and cancer stage were not significant predictors of case outcome. Failure to refer to a surgeon was
two-fold (OR [95% CI]: 2.44 [1.085, 5.489]) more likely to be resolved with payment. Cases with a delay in diagnosis of >12
months were two-fold (OR [95% CI]: 2.129 [1.086, 4.175]) more likely to be resolved with payment compared to a delay <12 mos.
Patients who failed to follow up as recommended were two-fold (OR [95% CI]: 2.31 [1.05, 5.10]) less likely to have their case be
resolved with payment.
CONCLUSION

Plaintiffs involved in breast cancer imaging-related malpractice cases tend to be younger than the median age of diagnosis of
breast cancer for US women (62 per NCI Surveillance, Epidemiology and End Results data). Breast cancer-related suits involve
physicians from multiple specialties, radiology being the most common. Delay in diagnosis, lack of surgeon referral, and lack of
recommended follow-up are related to plaintiff payments and may be areas of professional practice to target to prevent over- and
misuse of the medical malpractice system.
CLINICAL RELEVANCE/APPLICATION

Breast cancer imaging-releated medical malpractice remains prevalent and costly for all involved. A better understanding of factors
and trends in malpractice litigation can lead to medical malpractice system improvement.
BR240-SD- Measurement of Breast Density Using Bioimpedance
MOB6
Station #6
Participants
Susan M. Astley, PhD, Manchester, United Kingdom (Presenter) Nothing to Disclose
Elaine Harkness, PhD, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Ryan Halter, Hanover, NH (Abstract Co-Author) Nothing to Disclose
Ethan Murphy, Hanover, NH (Abstract Co-Author) Nothing to Disclose
Ethan Du-Crow, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Annabel L. Davies, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Josh W. Lindsay, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Katherine Graves, Slough, United Kingdom (Abstract Co-Author) Nothing to Disclose
Jessica Ritchie, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Charlotte Day, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Kim Denton, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Sarah Sampson, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Anthony Maxwell, MBChB, FRCR, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Sacha Howell, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Anthony Howell, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
D. Gareth Evans, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

sue.astley@manchester.ac.uk
PURPOSE
Breast density is associated with both risk of developing breast cancer and the effectiveness of mammography as a screening tool.
An accurate and practical measurement method is crucial for personalising screening programmes and assessing the efficacy of
preventive interventions. We evaluate the relationship between breast bioimpedance and volumetric breast density measured from
mammograms. If mammographic density can be estimated using bioimpedance, this would provide a non-invasive, inexpensive and
safe method of assessment.
METHOD AND MATERIALS

Volumetric breast density (VBD) measured by Volpara from mammograms and bioimpedance measurements using a custom Electrical
Impedance (EI) data acquisition device were measured in 211 women commencing a chemoprevention trial. Bioimpedance
measurements at 3.2-102.4 kHz were made with the woman reclining at an angle of approximately 30°. Eight paediatric ECG
electrodes were placed around and equidistant to the nipple. Bioimpedance was recorded from eight combinations of four electrodes
(two delivering current and two measuring potential difference). The bioimpedance recorded from these eight combinations is
averaged for each breast and compared to VBD. Data from women with missing VBD or bioimpedance signatures and those with
poor electrode contact were excluded from the analysis.
RESULTS

EI and VBD data were available for 106 women. As expected, VBD is inversely related to bioimpedance; low VBD breasts with a
greater adipose content have higher impedance than more glandular breasts. Weighted least squares regression analysis suggests
the relationship between EI and VBD is of the form y=AxB, where A=582.926±6.062 Ω-1, B=-1.008±0.0204, and has a Χred2 value
of 14.05.
CONCLUSION
EI has the potential to identify women with high breast density who are at increased risk of cancer or require supplemental
screening, and those with low breast density who could potentially be screened less frequently. Further development of the
technology is necessary to improve reliability.
CLINICAL RELEVANCE/APPLICATION

A safe, practical and inexpensive method of assessing breast density is needed for personalising screening by risk of cancer and
masking, and for assessing the efficacy of preventive interventions.
BR182-ED- Advances in Radiology Education: Development of Subspecialty Specific Entrustable Professional
MOB7 Activities Using a Double Delphi Technique
Station #7
Participants
Monica M. Sheth, MD, Lake Success, NY (Presenter) Nothing to Disclose
Alice Fornari, Manhassett, NY (Abstract Co-Author) Nothing to Disclose
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Katherine A. Klein, MD, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Abstract Co-Author) Nothing to Disclose
Petra J. Lewis, MD, Lebanon, NH (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

msheth@northwell.edu
TEACHING POINTS

1. EPAs provide practicality to our current means of assessing trainee readiness for clinical practice by representing an essential
skill or task performed in daily clinical practice. 2. The developed breast imaging specific EPAs support the General Radiology EPAs
published in 2016 while providing a more refined scope of practice specific to the subspecialty. 3. The Delphi technique is a
methodologically approved consensus driven process used to synthesize and validate expert opinion when evidence is not available.
TABLE OF CONTENTS/OUTLINE
*Introduction: Background; development of EPAs to bridge gap between competency based training and clinical practice by framing
competencies into clinical context; 5 Levels of entrustment; Alignment with ACGME competencies, subcompetencies and
milestones; extension of general radiology EPAs; Need for subspecialty specific EPAs *Methods: Development of EPA list; Validation
via Double Delphi Technique; Refinement by educational theorists; role of facilitator in entire process *Outcome: 8 EPAs specific to
breast Imaging *Challenges *Next Steps: EPA-based online curriculum development and assessments; providing guidance for other
subspecialties in radiology to develop their own EPAs
BR183-ED- Two Faces Breast Cancer: Shooting the Enemy More Than One Time with Biopsy Gun to Unmask It
MOB8
Station #8
Awards
Cum Laude
Participants
Lee Van Diniz, MD, Botucatu, Brazil (Presenter) Nothing to Disclose
Lucas P. Rodrigues, MD, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Eduardo C. Pessoa, PhD,MD, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Julia d. Veloso, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Guilherme B. Neves, MD, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Carla P. Pessoa, PhD,MD, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Joana C. Machado, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Ana Carolina D. Augusto, MD, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
Seizo Yamashita, Botucatu, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

leevandiniz@gmail.com
TEACHING POINTS

Molecular subtypes classification is important to guide treatment, including primary chemotherapy, and developing new therapies.
The purposes of this Education Exhibit are: 1. To discuss about molecular subtypes of breast cancer. 2. To demonstrate an
association between the molecular subtype and ultrasound images. 3.To discuss some cases that we chose to do the biopsy in
more than one location because ultrasound characteristics showed different patterns in the same patient and the histopathology
confirmed different results.
TABLE OF CONTENTS/OUTLINE
1.Introduction. 2. Molecular subtypes of breast cancer. 3. Radiopathological characteristics of molecular subtypes. 4. Show cases
that we chose to do the biopsy in more than one location because the image characteristics showed different patterns in the same
patient. 5. To discuss about possibility of choosing biopsy in two different locations and whether these radiological changes showed
different histopathological characteristics in our cases. 5. To emphasize how the decision in making biopsy in more than one
location can alter the patient's prognosis and treatment.
BR184-ED- Don't Forget the Breast: Incidental Breast Findings on CT
MOB9
Station #9
Participants
Ayushi Singh, DO, New York, NY (Presenter) Nothing to Disclose
Priyanka Kadaba, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Mary M. Salvatore, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Richard H. Stern, PHD, Englewood, NJ (Abstract Co-Author) Nothing to Disclose
Laurie R. Margolies, MD, New York, NY (Abstract Co-Author) Research Consultant, FUJIFILM Holdings Corporation
For information about this presentation, contact:

ayushi.singh@mountsinai.org
TEACHING POINTS

Evaluation of the breast parenchyma on CT can lead to earlier diagnosis of breast cancer Demonstrate malignant and benign lesions
in the breast on CT and which lesions warrant further work-up Demonstrate the appearance of the breast on CT in the
postoperative setting Provide guidelines for interpretation of breast lesions on CT on the basis of the breast imaging lexicon
TABLE OF CONTENTS/OUTLINE

1. Introduction a. Incidence and outcome of breast findings on CT imaging b. Advantages of imaging the breast on CT i. Improved
contrast resolution ii. Larger field of view/cross-sectional capability iii. Better lesions located near the chest wall and medially 2.
Breast Findings a. Malignant i. Invasive ductal carcinoma ii. Invasive lobular carcinoma iii. Invasive ductal and lobular carcinoma iv.
Secondary lymphoma of the Breast v. Inflammatory Carcinoma b. Benign i. Fibroadenoma ii. Benign Calcifications iii. Hamartoma iv.
Lactating breast c. Postoperative findings i. Hematoma/Seroma ii. Fibrous Scar 3. Breast Imaging Lexicon for CT a. Rule of Thumb:
Always include shape, density, pattern of enhancement and associated findings
BR185-ED- You Sow What You Reap: Tumoral Needle Tract Seeding After Percutaneous Breast Biopsy
MOB10
Station #10
Awards
Certificate of Merit

Participants
Heni D. Skaf, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Rafael L. Macedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Caio D. Pinheiro, MBBS, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Cecilia S. Goldman, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Matheus V. Paulino, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Su J. Kim, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana H. Catani, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Vera Christina C. Ferreira, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Erica Endo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciano F. Chala, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

rafajf85@gmail.com
TEACHING POINTS
The purposes of this exhibit are: 1- Review our institution's cases of tumoral needle tract seeding after percutaneous breast
biopsy. 2- Review the prevalence of tumoral needle tract seeding after percutaneous breast biopsy. 3 - Review which tumors are
more susceptible to this complication. 4 - Evaluate the factors that aggravate this phenomenon. 5 - Identify the best ways to
prevent this complication.
TABLE OF CONTENTS/OUTLINE
1 - Review breast tumoral needle tract seeding cases at our radiology department, documented in multiple methods (MRI,
ultrasound and mammography). 2 - Background: epidemiology and clinical aspects of needle tract tumoral seeding after
percutaneous biopsy. 3 - Histopathological and imaging features of the most involved tumors. 4 - Imaging aspects of tumoral local
recurrence. 5 - Identify factors that increase the risk of this complication and how to minimize it.
BR186-ED- Increased Unnaturally? What Lies Beneath Markedly Enlarged Breast
MOB11
Station #11
Participants
Sungmin Moon, Gwangju, Korea, Republic Of (Presenter) Nothing to Disclose
Hyo Soon Lim, MD, Jeollanam-Do, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
So Yeon Ki, Gwangju, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
To aid approach for evaluation of enlarged breast : diffuse breast enlargement or presence of huge mass or other unusual cause of
enlarged breast Provide pictorial review of diseases using multimodality image and discuss clinicopathological findings and
management To know enlarged breast can be associated with benign as well as malignant condition in peripuberty, adult women
and men
TABLE OF CONTENTS/OUTLINE

1. Normal breast development stage 2. Categorization for evaluation of enlarged breast 3. Case based review of diseases (A)
Diffuse breast enlargement Infection - Mastitis Hormonal disturbance - Gynecomastia Other cause - Edema caused by
extramammary origin - Edema caused by mammary procedure (B) Neoplasm Benign - Hamartoma - Fibroadenoma - Giant juvenile
fibroadenoma - Fibrocystic disease - Benign phyllodes tumor Malignant - Invasive ductal cancer - Inflammatory breast cancer -
Malignant phyllodes tumor (C) Extrammamary location of mass: Chest wall - Lipoma - Malignant fibrous histiocytoma 4. Differential
diagnosis: mastitis vs. inflammatory breast cancer 5. Conclusion
BR187-ED- 2018 New Trends! Axillary Ultrasound
MOB12
Station #12
Participants
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Jose Chico, Buenos Aires, Argentina (Presenter) Nothing to Disclose
Carolina Hadad, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Roberto Secco, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Griselda Choque Leniz, MEd,MEd, Longchamps, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

drakarina.pesce@gmail.com
TEACHING POINTS

1-Analyze the role of axillary ultrasound in the patient with breast cancer 2-Discuss the role of axillary ultrasound in the evaluation
of newly diagnosed breast cancer patients given recent changes in the surgical management of the axilla. 3- Emphasize the
common sonographic features of axillary metastasis
TABLE OF CONTENTS/OUTLINE
1-Introduction: Axillary staging in the context of breast cancer is a contentious topic. Ultrasound is the most commonly used
modality for axillary evaluation given its wide availability. Current debate questions whether there is a benefit to diagnosing
metastasis with ultrasound-guided needle biopsy as this may lead to more axillary node dissections in an era of its decreasing role.
2- What is the future of axillary ultrasound in the axillary management of breast cancer? 3- An updated literature review 4- Future
Developments 5-Conclusion
 VW 32

Wide-angle Digital Breast Tomosynthesis Reading Session: Siemens Healthineers Vendor Workshop
Monday, Nov. 26 1:05PM - 2:15PM Room: Booth 5530

Participants
Paula M. Grabler, MD, Oak Park, IL (Presenter) Nothing to Disclose
Brandie L. Fagin, MD, Glenview, IL (Presenter) Nothing to Disclose
Program Information
During this hands-on workshop you will learn to evaluate 2D mammography and 3D Breast Tomosynthesis. An expert tutor will lead
you through cases that will both fascinate and challenge you! All cases have been acquired with Siemens Mammomat Inspiration
and are displayed on our syngo® Breast Care workstations so you will become familiar with the quality of our HD Tomo images and
ease of use of our systems.
 MSRO27

BOOST: Breast-Science Session with Keynote

Monday, Nov. 26 1:30PM - 2:30PM Room: S103CD

BR RO OI
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Kathleen Horst, MD, Stanford, CA (Moderator) Nothing to Disclose
Anna Shapiro, MD, Syracuse, NY (Moderator) Nothing to Disclose
Sub-Events
MSRO27-01 Invited Speaker:
Monday, Nov. 26 1:30PM - 1:50PM Room: S103CD

Participants
Jianling Yuan, MD, PhD, Minneapolis, MN (Presenter) Nothing to Disclose
MSRO27-03 Where Are the RCTs? Analysis of the 2018 American Society for Radiation Oncology (ASTRO)
Evidence-Based Guidelines for Radiation Therapy to the Whole Breast as Treatment for Breast Cancer
Monday, Nov. 26 1:50PM - 2:00PM Room: S103CD

Participants
Norman R. Williams, PhD, London, United Kingdom (Presenter) Travel support, Carl Zeiss AG
For information about this presentation, contact:
norman.williams@ucl.ac.uk
PURPOSE
Early in 2018, the American Society for Radiation Oncology (ASTRO) produced evidence-based guidelines on five key questions for
radiation therapy to the whole breast as treatment for breast cancer [Smith et al PMID: 29545124]. An analysis was made of the
publications supporting these guidelines to determine how many reported level-1 evidence from randomised clinical trials (RCTs), as
this is the standard applied to chemotherapy and adjuvant hormonal therapy.
METHOD AND MATERIALS

All 112 references were scrutinized, and tabulated according to level of evidence (RCT or not), year of publication, country of lead
author, and which of the statements (grouped into five key questions) they addressed.
RESULTS

Of the 33 statements, 12 are not supported by evidence from RCTs. In a further 9 statements, data from RCTs only partly support
the consensus. Therefore, 21/33 (64%) of the statements are not directly supported by evidence from RCTs. There is no evidence
from RCTs to support any of the statements regarding avoiding exposure of cardiac and other normal tissue (key questions 4 and
5). Such exposure has been linked to death from ischemic heart disease [Darby et al PMID: 23484825] and lung cancer [Taylor et al
PMID: 28319436]. There is evidence that the effects of a course of whole breast radiation therapy induces early ECG changes
[Tuohinen et al PMID: 29599341]; biological effects can be detected after a single fraction [Woolf et al PMID: 25045612]. Such
measures could be used in the design of RCTs, particularly of patients with low-risk breast cancer in whom de-escalation of breast
radiation therapy (using accelerated partial breast, intra-operative, etc.) may be warranted to reduce an imbalance in the
efficacy/safety profile [Franco et al PMID: 29616366].
CONCLUSION
The majority of the 2018 ASTRO evidence-based guidelines for use of radiation therapy in breast cancer are not based on level-1
evidence from RCTs. Trials using techniques that minimize exposure to normal tissues are urgently required.
CLINICAL RELEVANCE/APPLICATION
Clinicians and patients should be aware that current guidelines for treatment of breast cancer using radiation therapy are mostly
based on sub-optimal evidence.
MSRO27-04 Clinical Outcomes and Toxicity of Proton Beam Radiation Therapy for Re-Irradiation of Locally
Recurrent Breast Cancer
Monday, Nov. 26 2:00PM - 2:10PM Room: S103CD

Awards
Student Travel Stipend Award
Participants
Prashant Gabani, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Maria A. Thomas, MD, PhD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Beth Bottani, CMD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Jeffrey D. Bradley, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Laura Ochoa, RN,PhD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Imran Zoberi, MD, St. Louis, MO (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
pgabani@wustl.edu
PURPOSE
Repeat radiation therapy (RT) using x-rays for locally recurrent breast cancer results in increased short and long-term toxicity.
Proton beam RT (PBRT) can minimize dose to surrounding organs thereby reducing toxicity. Here, we report the toxicity and
outcomes for women who underwent re-irradiation to the chest wall for locally recurrent breast cancer using PBRT.
METHOD AND MATERIALS
A total of 16 patients with locally recurrent breast cancer who underwent re-irradiation to the chest wall with PBRT between 2014-
2018 were retrospectively analyzed. For their recurrences, 6 patients underwent salvage mastectomy, 8 patients had wide local
excision, and 2 patients had biopsy only. The median dose for the first RT course was 50 Gy, and for the second course, 50.4
Cobalt Gy Equivalent. The target for re-irradiation was chest wall alone in 12 patients and chest wall plus regional nodes in 4
patients. A boost was delivered in 3 (18.8%) patients. Concurrent hyperthermia was used in 10 (62.5%) patients. For systemic
therapy, 4 (25%) patients received chemotherapy and 8 (50%) patients received hormone therapy. Follow up was calculated from
the start of second RT course. Toxicities were based on CTCAE 4.0.
RESULTS

The median age at diagnosis and at recurrence was 49.8 years and 60.2 years respectively. The median time between the two RT
courses was 10.2 (0.7-20.2) years. The median follow up time was 10.6 (1.5-29.1) months. There were no local failures observed
after re-irradiation. Only one patient developed distant metastasis and ultimately died. Grade 3-4 acute skin toxicity was observed
in 5 (31.2%) patients. There were 4 (25%) patients who developed chest wall infections during or shortly (2 weeks) after re-
irradiation. Grade 3-4 fibrosis was observed in only 3 (18.8%) patients. Grade 5 toxicities were not observed. Hyperpigmentation
was seen in 12 (75%) patients. Other RT related toxicities such as pneumonitis, telangiectasia, rib fracture, and lymphedema
occurred in 2 (12.5%), 4 (25%), 1 (6.3%), and 1 (6.3%) patients respectively.
CONCLUSION

Re-irradiation with PBRT for recurrent breast cancer has acceptable toxicities. There was a high incidence of grade 3-4 skin toxicity
and infections, however, they resolved with skin care and antibiotics. Further follow up is needed to determine long-term clinical
outcomes.
CLINICAL RELEVANCE/APPLICATION
PBRT can be safely used for re-irradiation of the chest wall for locally recurrent breast cancer.
MSRO27-05 Carcinosarcoma of the Breast: Treatment Patterns and Survival Outcomes
Monday, Nov. 26 2:10PM - 2:20PM Room: S103CD

Awards
Trainee Research Prize - Resident

Participants
William R. Kennedy, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Prashant Gabani, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Sahaja Acharya, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Maria A. Thomas, MD, PhD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Imran Zoberi, MD, St. Louis, MO (Abstract Co-Author) Nothing to Disclose
PURPOSE

Carcinosarcoma of the breast is a rare yet highly-aggressive tumor accounting for less than 1% of all breast cancers, for which
guidance on optimal management and prognosis are sparse. The purpose of this study is to investigate population-based treatment
patterns and overall survival (OS) outcomes in patients with this diagnosis.
METHOD AND MATERIALS

We queried the National Cancer Database for patients diagnosed with carcinosarcoma (Histology 8980) of the breast. All patients
included were treated with surgery, with or without chemotherapy and/or radiation therapy. Patients with metastatic disease were
excluded. Kaplan-Meier analysis was used to estimate OS. Univariate and multivariate cox analyses were used to determine
predictive factors of OS.
RESULTS
A total of 329 patients from 2004-2012 were identified. Median age at diagnosis was 58 years (range, 24-90). Patients had T1
(21%), T2 (44%), T3 (25%), or T4 disease (10%). Most patients were node-negative at diagnosis (77%). Breast conservation
surgery was utilized in 33% of patients. Chemotherapy was used in 66% of patients. Less than half (44%) of patients received
radiation therapy to a median dose of 50.4Gy (range 35-56 Gy), with a median 10Gy boost used in 76%. With median follow-up of
39.9 months, 3-year overall survival was 74%. Multivariate analysis revealed that T-stage, margin status, and chemotherapy use all
significantly influenced OS. There was a trend towards improved survival with the use of RT (HR 0.66, 95% CI 0.43-1.01, p
=0.053). The 3-yr OS was 80% in patients receiving chemotherapy vs 59% without chemotherapy. The 3-yr OS was 82% in
patients receiving RT vs 66% without RT.
CONCLUSION
Carcinosarcoma of the breast is associated with relatively poor rates of OS. The use of chemotherapy was associated with
Carcinosarcoma of the breast is associated with relatively poor rates of OS. The use of chemotherapy was associated with
improved OS, with a trend towards improved OS with the use of RT.
CLINICAL RELEVANCE/APPLICATION

In the largest study to date investigating outcomes in carcinosarcoma of the breast, adding chemotherapy to surgery improved OS.
A trend toward improved OS was also seen with adjuvant RT.
MSRO27-06 Quantitative Ultrasound Characterization of Radiation-Induced Acute Skin Toxicity in Breast Cancer
Patients Receiving Radiation Therapy: A Feasibility Study
Monday, Nov. 26 2:20PM - 2:30PM Room: S103CD

Participants
Sylvia D. Tang, Johns Creek, GA (Presenter) Nothing to Disclose
Jiwoong Jason Jeong, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
Xiaofeng Yang, PhD, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
Mylin A. Torres, MD, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
Arif N. Ali, MD, MS, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Tian Liu, PhD, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
PURPOSE

Despite technological advances in radiotherapy, high dose of radiation may induce acute skin toxicity in the majority of women
receiving breast-cancer radiotherapy. In current clinical practice, the severity of skin toxicity is often rated by clinicians through
visual inspection and physics examination, which is subjective. The purpose of this study is to investigate the feasibility of
quantitative characterization of radiation-induced acute skin toxicity via ultrasound morphological and texture analysis.
METHOD AND MATERIALS
Twelve patients receiving standard breast radiotherapy were enrolled in the longitudinal ultrasound study. Ultrasound B-mode
images are acquired at various time points: prior to, weekly during, as well as 6 weeks and 3 months post radiotherapy. At each
time point, 4 images (12, 3, 6 and 9 o'clock) were acquired on the irradiated breast and 4 mirror images were acquitted on the
contralateral normal breast. To evaluate radiation-induced skin changes, we performed both morphological (area, height, perimeter
and averaged skin thickness) and textural (contrast, angular second moment (ASM) and inverse difference moment (IDM)) analyses
using ImageJ. Clinical assessment of skin toxicity was performed at each time point.
RESULTS
Changes in skin thickness and texture were observed in 5 patients as early as 1 week during treatment. In 2 cases with most
severe acute toxicity, the average skin thickness of irradiated breast increases more than 175% and 188% at the end of
fractionated therapy in comparison to the untreated contralateral breast, while their slopes in linear regression are 0.87 and 0.70,
respectively. Acute skin toxicity was observed in differences in 5 cases in the angular second moment measurements and in 4
cases of entropy analysis over the patient's temporal treatment course.
CONCLUSION

Radiation-induced skin toxicity in breast cancer patients can be quantitatively assessed by ultrasound-based morphologic and
textural characterization.
CLINICAL RELEVANCE/APPLICATION

Quantitative ultrasound characterization of radiation-induced acute skin toxicity in breast cancer patients receiving radiation
therapy may be of clinical relevance for the optimization of treatment protocols and potential early intervention to prevent long-
term breast toxicity.
 VW 08

Global Perspectives on Risk-based Screening: Panel Discussion: GE Vendor Workshop

Monday, Nov. 26 1:30PM - 2:30PM Room: Booth 8156

Participants
Jack Cuzick, London, United Kingdom (Presenter) Nothing to Disclose
Program Information
A global panel of experts will discuss barriers and opportunities that contribute to the ability to institute risk based breast screening
including workflow impact, risk assessment tools, balance of screening benefits and risks, cost-effectiveness, patient management,
technology needs and more. Registration is required; adding this session to the RSNA calendar tool alone does not secure your
seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 59

Tomosynthesis Guided Prone Breast Biopsy Solutions: Hologic Vendor Workshop

Monday, Nov. 26 1:30PM - 2:30PM Room: Booth 5524

Participants
Daniel E. Lehrer, MD, CABA, Argentina (Presenter) Speaker, Hologic, Inc; Institutional research agreement, Siemens AG
Program Information
Clinical benefits of tomosynthesis guided biopsy which includes a hands-on demonstration of the Affirm® Prone Biopsy System and
the Brevera® Breast Biopsy System (Affirm® Prone Biopsy System and Brevera® Breast Biopsy System)
Registration
https://hologicrsna.com
 VW 33

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Monday, Nov. 26 2:30PM - 3:40PM Room: Booth 5530

Participants
Terri A. Gizienski, MD, Greenwood Village, CO (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 MSRO28

BOOST: Breast-Case-based Multidisciplinary Review (Interactive Session)

Monday, Nov. 26 3:00PM - 4:15PM Room: S103CD

BR OI RO
AMA PRA Category 1 Credits ™: 1.25
ARRT Category A+ Credits: 1.50

Participants
Nora M. Hansen, MD, Chicago, IL (Presenter) Nothing to Disclose
Bethany L. Niell, MD,PhD, Tampa, FL (Presenter) Nothing to Disclose
Jean L. Wright, MD, Baltimore, MD (Presenter) Nothing to Disclose
Cesar A. Santa-Maria, MD, Baltimore, MD (Presenter) Research funded, AstraZeneca PLC; Research funded, Pfizer Inc; Advisory
Board, Polyphor
For information about this presentation, contact:
jwrigh71@jhmi.edu
LEARNING OBJECTIVES
1) Describe the latest advances in breast cancer imaging before, during, and after treatment. 2) Facilitate a multidisciplinary
approach to the diagnosis, management, and treatment of breast cancer.
 SSE01

Breast Imaging (Breast Density and Risk Assessment)

Monday, Nov. 26 3:00PM - 4:00PM Room: E451B

BQ BR
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

FDA Discussions may include off-label uses.

Participants
Jennifer A. Harvey, MD, Charlottesville, VA (Moderator) Stockholder, Hologic, Inc; Research Grant, Volpara Health Technologies
Limited; Stockholder, Volpara Health Technologies Limited;
Ioannis Sechopoulos, PhD, Atlanta, GA (Moderator) Research Grant, Siemens AG; Research Grant, Canon Medical Systems
Corporation; Speakers Bureau, Siemens AG; Scientific Advisory Board, Fischer Medical
Sub-Events
SSE01-01 Surrounding Regions of Tumor in FFDM are Associated with Breast Cancer Prognostic and
Proliferation Markers
Monday, Nov. 26 3:00PM - 3:10PM Room: E451B

Awards
Student Travel Stipend Award
Participants
Dooman Arefan, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
Bingjie Zheng, MD, Zhengzhou, China (Abstract Co-Author) Nothing to Disclose
Ruimei Chai, Shenyang, China (Abstract Co-Author) Nothing to Disclose
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
Jules H. Sumkin, DO, Pittsburgh, PA (Abstract Co-Author) Research Grant, Hologic, Inc; Research Grant, General Electric Company
Shandong Wu, PhD, MSc, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
wus3@upmc.edu
PURPOSE
Radiomic information of segmented tumors in breast images have been shown to be correlated with prediction of certain
markers/surrogates for prognosis. Surrounding regions of tumor may have been affected by tumor development but its effects
remain unclear for similar prediction purposes. We performed an investigation on the radiomic imaging features extracted from both
tumor and surrounding regions in relation to prediction of breast cancer distant recurrence risk and tumor proliferation markers.
METHOD AND MATERIALS
We performed an IRB-approved retrospective study on 119 ER-positive and node-negative invasive breast cancer patients
diagnosed (confirmed by pathology) during 2011-2016. All patients had FFDM scans (including MLO and CC views), Oncotype DX
recurrence risk scores, and proliferation markers (Ki-67) available. Breast tumor was segmented by an expert breast imaging
radiologist, and a varying size (diameter) of surrounding regions outside the segmented tumor were automatically separated using
automated image processing techniques. A total of 23 radiomic features were extracted respectively from tumor and its surrounding
region. The logistic least absolute shrinkage and selection operator (LASSO) regression model was used to estimate the Oncotype
DX risk score and Ki-67 rate, respectively, using the same set of 23 features. AUC and Pearson's correlation coefficient (r) are
performance metrics.
RESULTS
For features extracted from tumor alone, r was 27% (p< 0.05) and 35% (p<0.05) for estimating Oncotype DX and Ki67,
respectively, while the corresponding AUC was 0.77 (High Oncotype DX vs Intermediate and low) and 0.58 (High Ki-67 vs low Ki-
67). When imaging features from the surrounding regions (6 mm outer from tumor boundary) were incorporated additionally, r
increased to 34% (p<0.05) and 47% (p<0.05) for estimating Oncotype DX and Ki-67, respectively, while the AUC was 0.78 and
0.63. The LASSO-selected features included the tumor solidity, surrounding region's skewness and intensity contrast.
CONCLUSION
Inclusion of the surrounding regions of breast tumor in FFDM increased the performance of predicting Oncotype DX recurrence risk
scores and Ki-67 proliferation rate.
CLINICAL RELEVANCE/APPLICATION
Surrounding regions of breast tumor in FFDM may provide additional quantitative information over tumors to enhance prognosis and
proliferation estimation
SSE01-02 Developing a Deep Learning Algorithm to Automatically Classify Breast Density According to 5th
 Edition of ACR BI-RADS in Mammography
Monday, Nov. 26 3:10PM - 3:20PM Room: E451B

Participants
Yiming Ding, Berkeley, CA (Presenter) Nothing to Disclose
Mian Zhong, Berkeley, CA (Abstract Co-Author) Nothing to Disclose
Youngho Seo, PhD, San Francisco, CA (Abstract Co-Author) Consultant, BioLaurus, Inc
Thienkhai H. Vu, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Hari Trivedi, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Dmytro Lituiev, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Amie Y. Lee, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Dexter Hadley, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Bonnie N. Joe, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jae Ho Sohn, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
sohn87@gmail.com
PURPOSE
The current edition of BI-RADS assigns breast density qualitatively based on potential for obscuring breast cancer, rather than a
simple quantitative summation of image intensities. The aim of this study is to develop a deep learning algorithm for breast density
classification based on the new BI-RADS system and then retrospectively apply them to a separate set of digital mammograms to
automatically classify breast density according to BI-RADS 5th Edition.
METHOD AND MATERIALS
A convolutional neural network with ResNet50 architecture was trained on 94,562 screening mammograms performed from 2014 to
2018 from a single institution following the current 5th edition of BI-RADS for mammography. Optimal neural network
hyperparameters were selected via validation accuracy monitoring. The trained model was then applied to a hold-out test set of
size 9,547 from the same institution. We then manually inspected all 13 cases in which the predicted label differed by more than 2
density labels from the ground truth. This algorithm was applied to 433,760 screening mammograms in large scale.
RESULTS

Our deep learning model achieved high sensitivity and specificity in assigning breast density category. Breast density distribution for
the training data was 7,752 (A: 8.20%), 35,656 (B: 37.71%), 40,943 (C: 43.30%), and 10,211 (D: 10.79%). The AUCs of ROC
curves on the test set were 0.97, 0.93, 0.92, and 0.96, respectively for each breast density category. Error analysis revealed that
among the 13 cases where breast density differed by more than 2 classes between actual and predicted labels, 7 were due to
breast implants, 4 were due to incorrect ground truth labels, and 2 remained equivocal. After the model was applied to the 433,760
screening mammograms, the model-predicted breast density distribution came to be 40,109 (A, 9.25%), 151,893 (B, 35.02%),
193,283 (C, 44.56%), and 48,475 (D, 11.17%).
CONCLUSION

Our deep learning algorithm successfully modeled the breast density classification scheme in the 5th edition BI-RADS system. This
was retrospectively applied in large scale to 433,760 mammograms for further inspection.
CLINICAL RELEVANCE/APPLICATION
Qualitative breast density assessment by radiologists is subject to variability. Given widespread adoption of breast density
notification laws in the U.S., automated breast density assessment based on masking can improve consistency of breast density
assignment particularly between categories B & C.
SSE01-03 Using Quantitative Breast Density Analysis to Predict Interval Cancers and Node Positive Cancers in
Pursuit of Improved Screening Protocols
Monday, Nov. 26 3:20PM - 3:30PM Room: E451B

Participants
Elizabeth S. Burnside, MD,MPH, Madison, WI (Presenter) Dr. Burnside has a research grant from Hologic
Lucy M. Warren, PhD, Guildford, United Kingdom (Abstract Co-Author) Nothing to Disclose
Louise S. Wilkinson, MBBCh,FRCR, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Kenneth C. Young, PhD, Guildford, United Kingdom (Abstract Co-Author) Nothing to Disclose
Jonathan Myles, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Stephen W. Duffy, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
PURPOSE
This study investigates whether quantitative breast density can predict interval cancers and node positive screen detected
cancers in order to serve as a biomarker to consider more aggressive screening to improve early detection.
METHOD AND MATERIALS

We conducted a case-control study of 1204 women drawn from the U.K. NHS Breast Screening Program aged 50-74 including 599
cases (comprising 302 screen detected cancers, 297 interval cancers; 239 node positive, 360 node negative) and 605 controls.
Each woman had prior digital mammograms and 70% had unprocessed images. A radiologist assessed breast density using a visual
analog scale (VAS) from 0 to 100 and BI-RADS 5th Edition density categories. Volpara software (V1.5.1) calculated fibroglandular
volume (FGV) and volumetric density grade (VDG) on unprocessed images. Logistic regression determined whether the breast
density measures could predict mode of detection (screen detected or interval); node-negative cancers; and node-positive
cancers, all vs. controls.
RESULTS

FGV predicted both screen-detected (p<0.01) and interval cancers (p<0.01) compared to controls. VDG, VAS and BI-RADS
FGV predicted both screen-detected (p<0.01) and interval cancers (p<0.01) compared to controls. VDG, VAS and BI-RADS
predicted interval cancers (all p<0.01) but not screen-detected cancers (p=0.16, p=0.18, p=0.46 resp.). FGV demonstrated
impressive risk stratification with an age-adjusted relative risk (RR) of the 4th quartile compared to the 1st quartile of 3.7 overall,
2.8 for screen detected, and 5.3 for interval cancers. VDG also had notable risk stratification with an age-adjusted RR of 3.6 for
interval cancers (Table). FGV predicted node-negative cancers as compared to controls (p<0.01) while BI-RADS, VAS, and VDG did
not (p=0.07, p=0.09, and p=0.47 resp.). FGV, BI-RADS, and VDG predicted node-positive cancers (all p<0.01) while VAS did not
(p=0.14).
CONCLUSION
FGV predicts interval, screen detected, node-positive and node-negative cancers compared to controls and provides remarkable
stratification the RR of interval cancers. BI-RADS and VDG predict interval and node positive cancers. VAS only predicts interval
cancers. The quantitative and automated nature of FGV and VDG and notable risk stratification based on RR indicates that these
variables may be promising biomarkers.
CLINICAL RELEVANCE/APPLICATION

By predicting mode of detection and nodal status, FGV may be a biomarker for more intensive screening. By predicting interval
cancers, BI-RADS, VAS, and VDG may act as supplementary biomarkers.
SSE01-04 Fully-Automated Volumetric Breast Density Estimation from Digital Breast Tomosynthesis: A Case-
Control Comparison to Area-Based Density Measures from Digital Mammography
Monday, Nov. 26 3:30PM - 3:40PM Room: E451B

Participants
Aimilia Gastounioti, Philadelphia, PA (Presenter) Nothing to Disclose
Meng-Kang Hsieh, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
PURPOSE
To investigate associations between breast cancer and fully-automated volumetric density measures extracted with digital breast
tomosynthesis (DBT), while also comparing to area-based density measures from digital mammography (DM).
METHOD AND MATERIALS
We retrospectively analyzed contralateral combo DM-DBT studies (Selenia Dimensions, Hologic Inc.) from 174 women with unilateral
breast cancer and 696 controls matched to cases on age, ethnicity and screening exam date at a 4:1 ratio. The publicly available
'LIBRA' software (v1.0.4) was adapted and used to estimate absolute dense volume (DV) and volumetric percent density (VPD) from
DBT, and absolute dense area (DA) and area percent density (APD) from DM images. Quantra (v2.2; Hologic Inc.) was also applied
to DM images, allowing to evaluate APD as well as to infer DV and VPD via physics-based models. Associations between the
different density measures and breast cancer were evaluated via logistic regression after adjustment for age, ethnicity and body
mass index (BMI). Area under the curve (AUC) of the receiver operating characteristic was used to assess case-control
discriminatory capacity, where model performance was compared using the DeLong's test, and odds ratios (ORs) for each of the
density measures were estimated.
RESULTS

All density measures had a significant association with breast cancer (OR = 1.24-2.40) after adjustment for age, ethnicity and BMI.
Models based on volumetric density measures had significantly (p < 0.05) larger case-control discriminatory capacity (AUC = 0.59-
0.63) than models considering area density (AUC = 0.57-0.59). Maximum breast cancer association was observed for DV with
slightly (p = 0.440) improved performance when extracted with DBT (AUC = 0.63) relative to DV inferred from DM (AUC = 0.61).
CONCLUSION

Fully-automated, quantitative, volumetric evaluation of breast density from DBT is feasible and can result in larger case-control
discriminatory capacity than area-based density measures from conventional planar mammography. Associations with breast cancer
can potentially further improve when volumetric density evaluation is performed with the DBT reconstructed breast volume
compared to physics-based models applied to DM.
CLINICAL RELEVANCE/APPLICATION

Our results further elaborate important clinical implications of breast density measures estimated with DBT, which may result in
improved measures of breast density in breast cancer risk assessment.
SSE01-05 Predicting Masking Risk in Mammography
Monday, Nov. 26 3:40PM - 3:50PM Room: E451B

Participants
James G. Mainprize, PhD, Toronto, ON (Presenter) Institutional research agreement, General Electric Company
Olivier Alonzo-Proulx, Toronto, ON (Abstract Co-Author) Institutional research agreement, General Electric Company
James Patrie, MS, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Jennifer A. Harvey, MD, Charlottesville, VA (Abstract Co-Author) Stockholder, Hologic, Inc; Research Grant, Volpara Health
Technologies Limited; Stockholder, Volpara Health Technologies Limited;
Martin J. Yaffe, PhD, Toronto, ON (Abstract Co-Author) Research collaboration, General Electric Company; Shareholder, Volpara
Health Technologies Limited; Co-founder, Mammographic Physics Inc; Research Consultant, BHR Pharma LLC
For information about this presentation, contact:
james.mainprize@sri.utoronto.ca
PURPOSE
Masking in mammography is the reduction of lesion conspicuity by surrounding and overlying dense tissue. Masking risk is increased
in dense breasts, leading to reduced sensitivity of breast screening. We have developed a masking index that can predict the
likelihood of a masked or missed cancer and could be used in a screening program to stratify women at greatest risk of masking to
alternative or supplementary imaging modalities to mammography.
METHOD AND MATERIALS
The study population were cancer cases collected (2003-2013) a case-control study used to develop a breast cancer risk model
incorporating density measures. Cancers were classified as screen-detected cancers (SDC) found on a screening mammogram and
non-screen detected cancers (NSDC) found by clinical symptoms or other imaging. The study had ethics board approval with
informed consent All SDC found on baseline images were excluded. Inclusion as NSDC required at least one prior negative screening
exam within two years of diagnosis. Images were analyzed with in-house algorithms and by volumetric breast density (VBD)
software. The aim in this study was to create an index that differentiated mammograms which allowed for detection (SDC) from
those for causing masking or missed lesions (NSDC). To avoid the influence of the lesion itself, only the contra-lateral breast images
were used.
RESULTS

The study included 90 NSDC cases and 186 SDC controls. Univariate masking indices based on BMI, age, BI-RADS density, VBD or
mean detectability yielded areas under ROC (AUC) of 0.61, 0.65, 0.67, 0.72 and 0.75 (±0.06 95% confidence) respectively. For
cancers found within one year, the detectability AUC improved to 0.81.
CONCLUSION
Age and BMI are relatively weak predictors of masking risk whereas VBD and detectability measures have better performance.
Further, adding textural measures improve predictions slightly, suggesting that the masking effects of anatomic noise and texture
are informative. In future, we will validate the model in an independent population and test the result on normal mammograms to
predict impact on a stratified screening program.
CLINICAL RELEVANCE/APPLICATION

A reliable masking index to predict when mammography will underperform would be a valuable tool in a stratified screening program
which could be used to redirect women with highly masked mammograms to alternative or adjunct screening strategies such as
tomosynthesis, MR or ultrasound.
SSE01-06 Quantitative MRI Background Parenchymal Enhancement and Contralateral Breast Cancer Risk
Monday, Nov. 26 3:50PM - 4:00PM Room: E451B

Participants
Lei Zhang, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
Xiaosong Chen, Shanghai, China (Abstract Co-Author) Nothing to Disclose
Yue Liang, Shanghai, China (Abstract Co-Author) Nothing to Disclose
Aly A. Mohamed, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Yiwei Tong, Shanghai, China (Abstract Co-Author) Nothing to Disclose
Kunwei Shen, shanghai, China (Abstract Co-Author) Nothing to Disclose
Shandong Wu, PhD, MSc, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

wus3@upmc.edu
PURPOSE
MRI background parenchymal enhancement (BPE) has been shown to be correlated with the risk of developing breast cancer in
breast-cancer free women. History of breast cancer is a significant risk factor of contralateral breast cancer. We investigated the
association between quantitative MRI BPE measures and risk of contralateral breast cancer development in a case-control setting
METHOD AND MATERIALS

A retrospective case-control (1:2 case-control ratio) study was performed using breast DCE-MRI scans from 135 newly diagnosed
unilateral breast cancer patients (confirmed by pathology) from 2010-2016: 35 women had future contralateral breast cancer
development after the initial breast cancer diagnosis and 70 were age- and year-of-MRI matched controls that remained breast
cancer-free in the contralateral breasts after at least 1-year follow-up. The MRIs acquired at the initial diagnosis of the unilateral
breast cancer were analyzed using published automated computer algorithms, generating two quantitative BPE measures computed
from the first post-contrast series: the absolute BPE volume (|BPE|) and its relative amount over the whole breast volume (BPE%).
Volumetric amounts of fibroglandular (dense) tissue (|FGT| and FGT%) were also automatically quantified from the MRI. Conditional
logistic regression was performed to assess these BPE and FGT measures as predictors of the contralateral breast cancer
development.
RESULTS

Average age was 52.2±13.1 (range 28-81) for cancers and 52.8±12.9 (range 28-80) for controls. 51% (or 49%) are post-
menopause for cancers (or controls). Invasive other-cancer history was found in less than 10% of the patients for both cancer
cases and controls. None of the |FGT|, FGT%, |BPE|, BPE% measures is statistically significant (all p>0.05) in terms of the
difference between the cancer group and control group. Logistic regression showed that odds ratios of these four measures are
very close to 1 with no statistical significance.
CONCLUSION

This pilot study showed that MRI BPE quantified in the contralateral breast of unilateral breast cancer patients is not associated
with the future development of the contralateral breast cancers.
CLINICAL RELEVANCE/APPLICATION

BPE may be affected bilaterally even in unilateral breast cancer patients. Contralateral breast cancer risk markers can guide
treatment selection and risk reduction, meriting further investigation.
 SSE02

Breast Imaging (Patient-Centric Care)

Monday, Nov. 26 3:00PM - 4:00PM Room: E450B

BR HP
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Rachel F. Brem, MD, Washington, DC (Moderator) Board of Directors, iCAD, Inc; Board of Directors, Dilon Technologies, Inc ; Stock
options, iCAD, Inc ; Stockholder, Dilon Technologies, Inc; Consultant, Dilon Technologies, Inc; Consultant, ClearCut Medical Ltd ;
Consultant, Delphinus Medical Technologies, Inc
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Moderator) Nothing to Disclose
Sub-Events
SSE02-01 Communicating Mammography Results: What Method and How Quickly Do Women Want Their
Screening Mammogram Results?
Monday, Nov. 26 3:00PM - 3:10PM Room: E450B

Participants
Julia Staschen, BS, Richmond, VA (Abstract Co-Author) Nothing to Disclose
Nghiem Pham, BA, Midlothian, VA (Presenter) Nothing to Disclose
Alicia Johns, Richmond, VA (Abstract Co-Author) Nothing to Disclose
Biren A. Shah, MD, Glen Allen, VA (Abstract Co-Author) Royalties, Wolters Kluwer nv; Royalties, Springer Nature
For information about this presentation, contact:
phamn@vcu.edu
PURPOSE
Expectations for when and how to receive results for mammograms are uncertain and the goal of this study is to explore
expectations for patient care with regards to receiving mammogram results. The purpose of the study was to understand the
majority preference with regards to the wait time for screening mammogram results, whether prompt communication of mammogram
results was of importance to patients, whether the time frame to schedule an additional imaging follow-up appointment after an
abnormal screening mammogram was of importance to patients, and how patients preferred to be given their screening mammogram
results. From the survey, investigators wanted to determine whether any quality practice improvements are necessary at their
academic breast imaging centers in order to improve the communication of screening mammogram results.
RESULTS

There were 2,245 patients who participated in the survey. A majority of patients preferred to receive results on Friday (N=1,868,
85.4%). Most individuals preferred to schedule their follow up appointments soon after their initial appointment, preferring either the
next day or within 1-2 days. Finally, over half of the sample preferred to be contacted via a phone call, with letter and text
messaging being the next most preferred methods. The preference for receiving results on Friday was evaluated by each of the
patient characteristics. The responses by the patients for preference for receiving results on Friday were significantly different by
ethnicity, education, and clinic. In particular, patients of other ethnicities besides African Americans and Caucasians responded
with the highest percent to receiving results on Friday and Caucasians were the least inclined to prefer to receive results on Friday.
Individuals with some college and college degrees were more likely to not prefer receiving results on Friday than those with no
formal education or high school graduates. The preferred time for scheduling a follow up was assessed by each of the patient
characteristics. Across all the patient characteristics, patients preferred to schedule their follow up appointment the next day.
Lastly, first choice of contact was examined across each of the patient characteristics. Phone call was the overwhelming choice for
method of contact for all patient characteristics.
CONCLUSION
The findings suggest that the patient population surveyed have a preference for their wait time, which is either to wait or receive
them within 24 hours. Patients preferred to receive results on Friday, and the most frequent choice for scheduling a follow up
appointment was the next day. A phone call was the preferred first choice for method of contact with e-mail being the least
preferred. These suggestions can help clinics and providers make changes to how they communicate their results. These findings
may help to streamline results for patients who prefer a shorter wait time, and highlight the overwhelming patient preference to
receive their abnormal screening mammogram results on a Friday. The findings may also suggest how to contact patients in a
different way, as phone calls were the preferred method of contact. The strong preference of patients receiving their screening
mammogram results more promptly should help trigger alternative methods toward improving communication between the radiologist
and the patient.
METHODS
Patients The study population was patients aged 18 or older, who completed an anonymous paper survey consisting of eight
questions when coming in for a routine mammogram screening. The study period lasted from September to November 2017 at two
academic breast-imaging centers. Measures The primary objective of the study was to summarize the results from a survey
examining the communication of mammography results from the patient population. The association between patient demographics
and four outcomes, preferred wait time, preferred scheduling time for a follow-up appointment, preference for receiving screening
results on Friday, and preferred methods of contact was investigated. Parameters of the survey are summarized in attached Table
1. Statistical Methods Frequencies and percentages for clinic site, age, ethnicity, education, and insurance were calculated.
Pearson chi-square analyses were conducted to determine the association between wait time preference and each patient
demographic, between preferred scheduling time for follow-up appointment and each patient demographic, between preference for
receiving screening results on Friday and each patient demographic, and between the first choice of method of contact and each
patient demographics. SAS version 9.4 was used for all analyses.
PDF UPLOAD
http://abstract.rsna.org/uploads/2018/18003961/18003961_3t99.pdf
SSE02-02 City Patterns of Screening Mammography Uptake and Disparity Across the United States
Monday, Nov. 26 3:10PM - 3:20PM Room: E450B

Awards
Trainee Research Prize - Resident
Participants
Eric Kim, MD, New York, NY (Presenter) Nothing to Disclose
Linda Moy, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Yiming Gao, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
C. A. Hartwell, BS, New York, NY (Abstract Co-Author) Nothing to Disclose
James S. Babb, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Samantha L. Heller, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kime18@nyumc.org
PURPOSE
Rural disparities in screening utilization are known to be affected by lack of access to care providers; city-level screening
mammography disparity has been less well evaluated, although over 30 million adult women live in the 500 largest cities. Our
purpose is therefore to evaluate disparities in screening in US cities.
METHOD AND MATERIALS

This descriptive study used public data from the 500 Cities project--500 largest cities, 103,020,808 individuals--which includes self-
reported screening uptake (127,298 women ages 50-74) from the Behavioral Risk Factor Surveillance System (BRFSS). Uptake was
matched with BRFSS and American Community Survey national population/economic census data variables expected to impact
screening: geographic region, health insurance, median household income, obesity, race, combined preventive care services (flu,
pneumococcal shots, colorectal screening), and pap smear use. Cities with incomplete data were excluded, yielding 490 cities,
34,629,163 women. Univariable and multivariable analyses were performed. All statistical tests were conducted at the two-sided
5% significance level using SAS 9.3 software (SAS Institute, Cary, NC).
RESULTS

Mean city screening mammography utilization was 77.7% (62.8%-88.9%). Utilization was highest in New England cities (p<0.002),
significantly positively correlated with pap smear (r=0.75), other preventive services (r=0.3), household income (r=0.44), %Asian
race (r=0.36), weakly with %Black race (r=0.10); significantly negatively correlated with obesity (r=-0.36), poverty (r=-0.30),
%White race (r=-0.29), no insurance (r=-0.27), (p<0.05 for all); not significantly correlated with population size (p=0.651).
Multivariable analysis demonstrated Pap smear use, Asian race, private insurance, and geographic region to be significant
independent predictors of utilization.
CONCLUSION
Screening mammography utilization varies across large cities in the U.S. with highest uptake in New England. Although the literature
focuses on rural screening disparities, disparities also exist even across large cities without physical barriers to screening.
CLINICAL RELEVANCE/APPLICATION

Mammographic screening disparities exist at the city level; identifying predictors of uptake may aid in targeting areas and
populations for screening education and intervention.
SSE02-03 Aligning Insurance Benefit Design with Patient Preference for Out-of-Pocket Cost Payments
Monday, Nov. 26 3:20PM - 3:30PM Room: E450B

Participants
Paniz Charkhchi, MD, Ann Arbor, MI (Presenter) Nothing to Disclose
Aaron Scherer, PhD, Iowa City, IA (Abstract Co-Author) Nothing to Disclose
Angela Fagerlin, Salt Lake City, UT (Abstract Co-Author) Nothing to Disclose
A. Mark Fendrick, MD, Ann Arbor, MI (Abstract Co-Author) Consultant, Abbott Laboratories; Consultant, AstraZeneca PLC;
Consultant, sanofi-aventis Group; Consultant, F. Hoffmann-La Roche Ltd; Consultant, GlaxoSmithKline plc; Consultant, Merck & Co,
Inc; Consultant, Neocure Group LLC; Consultant, Pfizer Inc; Consultant, POZEN Inc; Consultant, Precision Health Economics LLC;
Consultant, The TriZetto Group, Inc; Consultant, Zanzors; Speakers Bureau, Merck & Co, Inc; Speakers Bureau, Pfizer Inc;
Researcher, Abbott Laboratories; Researcher, AstraZeneca PLC; Researcher, sanofi-aventis Group; Researcher, Eli Lilly and
Company; Researcher, F. Hoffmann-La Roche Ltd; Researcher, GlaxoSmithKline plc; Researcher, Merck & Co, Inc; Researcher,
Novartis AG ; Researcher, Pfizer Inc
Ruth C. Carlos, MD, MS, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
panizcharkhchi@gmail.com
PURPOSE

Alternative payment models encourage physicians to accept payment bundles. Their patients continue to pay fee-for-service (FFS)
out of pocket costs, contrary to their preference for predictable out of pocket costs for a care episode. Using a hypothetical
screening mammography diagnostic episode of care, we assessed patient preference for FFS vs bundled copayments (copays).
METHOD AND MATERIALS
After IRB approval, we recruited a population-based cross-sectional survey of women 40-75 years old through Survey Sampling
International. Participants read a hypothetical scenario describing a screening mammography episode of care where 12% of women
undergoing screening needed a follow-up breast ultrasound and 2% a breast biopsy. Estimated out-of-pocket (OOP) costs for these
services and the characteristics of OOP payment types (FFS and a hypothetical bundled copay) were also described. We assessed
OOP payment type preference and knowledge (7 true/false items, see Fig. 1), additional test cost worry (7-point scale from not at
all to very worried), likelihood of mammogram use by OOP payment type and willingness to pay (WTP) for a breast screening
episode bundle.
RESULTS
Participant (n=1,236) characteristics are described in Fig. 1. 82.9%preferred bundled copays over FFS and 70.8% answered at
least half of the knowledge questions correctly. While most participants (74.5%) indicated they would get screened regardless of
OOP payment type, 13.8%said they would only get screened with bundled copays and 3.4% with FFS. Sizeable percentages of
participants were worried about costs associated with the ultrasound (42.0%) and biopsy (67.7%). Median WTP was $81 (25-
75IQ=$34-$215) for a bundle to avoid paying additional OOP for possible ultrasound or biopsy.
CONCLUSION

Given worry over the cost over potential additional tests, participants preferred a predictable OOP cost for a breast screening
episode, despite the low likelihood of additional testing. Although most participants would not defer mammography based on OOP
payment type, a significant minority were more likely to screen if bundled copays were available. Insurers should consider
incorporating these preferences in future benefits.
CLINICAL RELEVANCE/APPLICATION
Designing insurance plans to include bundled copays may decrease financial barriers to breast cancer work-up and diagnosis after
screening mammography.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Ruth C. Carlos, MD, MS - 2015 Honored EducatorRuth C. Carlos, MD, MS - 2018
Honored Educator
SSE02-04 Patient, Radiologist, and Examination Characteristics Affecting Screening Mammography Recall Rates
in a Large Academic Practice
Monday, Nov. 26 3:30PM - 3:40PM Room: E450B

Participants
Catherine S. Giess, MD, Wellesley, MA (Presenter) Nothing to Disclose
Aijia Wang, Boston, MA (Abstract Co-Author) Nothing to Disclose
Ivan Ip, MD, MPH, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Ronilda Lacson, MD, PhD, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Sarvenaz Pourjabbar, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Ramin Khorasani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

cgiess@bwh.harvard.edu
PURPOSE

To evaluate patient, radiologist, and examination characteristics affecting individual screening mammography recall rates in a large
academic breast imaging practice.
METHOD AND MATERIALS

This Institutional Review Board approved retrospective study included all screening mammography examinations in female patients
interpreted by thirteen breast imaging specialists at an urban academic center and two outpatient imaging centers from 10/1/2012-
5/31/2015. Patient demographics were extracted via electronic medical record. A natural language processing algorithm captured
breast density, BI-RADS assessment, and current and prior screening examination findings. Radiologists' annual screening volumes,
years of clinical experience, and percentage of time doing breast imaging were calculated. Risk aversion, stress from uncertainty,
and malpractice concerns were derived via online survey. Univariate and multivariate analyses assessed patient, radiologist, and
examination characteristics associated with likelihood of mammography recall. Pearson product-moment correlation coefficient
assessed relationship between cancer detection rate and recall rate.
RESULTS

Overall, 5,678 (9.3%) of 61,198 screening examinations were recalled. In multi-variate analysis, patient and radiologist
characteristics associated with higher odds of recall included patient's age < 50 years (p<0.0001); calcification (p<0.0001), mass
(p<0.0001), and higher density category (p<0.0001) on prior mammogram; baseline examination (p<0.0001); annual reading volume
< 1250 screening exams (p=0.0282); and <10 years experience (p=0.0036). Radiologist's risk aversion, stress from uncertainty,
malpractice concerns, or cancer detection rates were not associated with higher recall rates (r=-0.36, p=0.23).
CONCLUSION
In addition to patient and examination factors, screening recall variations were explained by radiologist's annual reading volume and
In addition to patient and examination factors, screening recall variations were explained by radiologist's annual reading volume and
experience. Interventions targeting radiologist factors may reduce unwarranted variation in screening recall and improve patient's
care experience.
CLINICAL RELEVANCE/APPLICATION

Efforts to reduce screening recall rate variability and false positive recalls might include increasing annual screening interpretive
volumes and review of uncertain screening findings for less experienced radiologists by more experienced breast imaging specialists.
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
SSE02-05 Preventive Care: How Comorbidities Affect Mammography Screening Rates
Monday, Nov. 26 3:40PM - 3:50PM Room: E450B

Awards
Student Travel Stipend Award
Participants
Cindy Yuan, MD,PhD, Chicago, IL (Presenter) Nothing to Disclose
Kirti M. Kulkarni, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Brittany Z. Dashevsky, MD,DPhil, Chicago, IL (Abstract Co-Author) Nothing to Disclose
PURPOSE

Here we evaluate the impact of comorbid conditions and age on screening mammography utilization.
METHOD AND MATERIALS

Data was retrospectively drawn from the 2011-2015 Medical Expenditure Panel Survey, which contained 40,752 women over the
age of 40. Utilization was defined as a screening mammogram within the previous one or two years, analyzed separately. A logit
model was employed to evaluate correlation of comorbidities with utilization. Statistical significance was defined by a p-value <0.05
by two-sided test.
RESULTS
Of the 36,575 women in the final sample, 45.9%, 43.6%, 3.9%, and 5.7% reported a history of hypertension (HTN), hyperlipidemia
(HLD), prior heart attack (MI) and prior stroke (CVA), respectively. Among women without a comorbid condition, baseline annual
mammography utilization was 47.3%. HTN and HLD were correlated with increased annual utilization [2.5 and 6.8 percentage points
(pp)]. In comparison, prior MI and CVA were correlated with decreased annual utilization (-8.2 and -1.5 pp, not statistically
significant in the latter). Results were similar for biennial utilization (3.0 and 7.6 pp increased utilization with HTN and HLD,
respectively, and -6.5, and 0.1 pp decreased utilization for MI and CVA, respectively).
CONCLUSION

Screening utilization was increased in patients with HTN and HLD, and decreased in patients with prior MI and CVA. An
understanding of how comorbid conditions influence screening may help better target specific populations and improve overall
utilization of preventive care.
CLINICAL RELEVANCE/APPLICATION
HTN and HLD, comorbidities that require regular physician visits, are correlated with increased screening. However, prior MI or CVA,
are correlated with decreased screening, possibly reflecting the effect of decreased life expectancy and increased morbidity. This is
especially relevant as age-based guidelines fail to take into account individual comorbidities and life expectancy.
SSE02-06 Contrast-Enhanced Spectral Mammography (CESM) in the Screening Setting: Patient Preferences and
Attitudes
Monday, Nov. 26 3:50PM - 4:00PM Room: E450B

Participants
Matthew M. Miller, MD, PhD, Charlottesville, VA (Presenter) Nothing to Disclose
Kathy L. Repich, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Carrie M. Rochman, MD, Charlottesville, VA (Abstract Co-Author) Research Consultant, Theraclion
Brandi T. Nicholson, MD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Jonathan Nguyen, MD, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
James Patrie, MS, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Roger T. Anderson, Charlottesville, VA (Abstract Co-Author) Nothing to Disclose
Jennifer A. Harvey, MD, Charlottesville, VA (Abstract Co-Author) Stockholder, Hologic, Inc; Research Grant, Volpara Health
Technologies Limited; Stockholder, Volpara Health Technologies Limited;
For information about this presentation, contact:

matthew.miller@virginia.edu
PURPOSE
Contrast-enhanced spectral mammography (CESM) is an emerging imaging tool that has been shown to have greater sensitivity
than conventional mammography and equal sensitivity with improved specificity relative to breast MRI in the diagnostic setting.
Ongoing studies are evaluating CESM performance in the screening setting, but little is known regarding whether patients would
agree to its use and the associated risks in screening. Our study aims to evaluate the attitudes and concerns of patients regarding
the use of CESM in a screening setting.
METHOD AND MATERIALS

In this IRB-approved HIPAA-compliant prospective study, patients with prior mammograms demonstrating heterogeneous or
extremely dense breasts presenting for screening mammography were invited to complete a survey. Patients were asked to rate
their perception of personal breast cancer risk compared to peers and their level of concern related to screening callbacks,
radiation exposure, and contrast allergies, and then identify which factors might deter them from getting adjunct screening exams
such as CESM.
RESULTS

512 patients with dense breasts undergoing screening mammography completed surveys. While 27% of surveyed patients reported
previously having been called back from screening for a diagnostic workup, a majority (63%) expressed little or no concern for
callbacks leading to additional imaging or biopsy. Most patients (63%) felt it was likely or very likely that cancer could be missed on
their mammogram, but only 9% had undergone adjunct screening exams in the past 3 years. The most commonly cited deterrents
to undergoing adjunct screening were cost (69%), pain (35%), and concern for an increased likelihood of having a biopsy or surgery
recommended (32%). When asked to select from several hypothetical adjunct screening modality choices, patients reported a
strong preference (63%) for a test that is most likely to detect their cancer, even if this would require IV-line placement. Only 5%
preferred a common but less sensitive test that did not require IV-line placement.
CONCLUSION
Our study suggests that women with dense breasts may accept CESM as an adjunct screening exam and may actually prefer it
over screening MRI or US given its relatively high sensitivity and low cost.
CLINICAL RELEVANCE/APPLICATION
Women with dense breasts may accept CESM as an adjunct screening exam given its sensitivity and low cost.
 SSE23

Physics (Breast X-Ray Imaging)

Monday, Nov. 26 3:00PM - 4:00PM Room: S502AB

AI BR PH SQ
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

FDA Discussions may include off-label uses.

Participants
Srinivasan Vedantham, PhD, Tucson, AZ (Moderator) Research collaboration, Koning Corporation
Hilde Bosmans, PhD, Leuven, Belgium (Moderator) Co-founder, Qaelum NV Research Grant, Siemens AG
Sub-Events
SSE23-01 Radiation Dose Reduction in Digital Breast Tomosynthesis (DBT) by Means of Neural Network
Convolution (NNC) Deep Learning
Monday, Nov. 26 3:00PM - 3:10PM Room: S502AB

Participants
Junchi Liu, MS, Chicago, IL (Abstract Submitter) Nothing to Disclose
Amin Zarshenas, MSc, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Syed Ammar Qadir, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Limin Yang, MD, PhD, Iowa City, IA (Abstract Co-Author) Nothing to Disclose
Laurie L. Fajardo, MD, MBA, Park City, UT (Abstract Co-Author) Consultant, Hologic, Inc; Consultant, Siemens AG; Consultant,
FUJIFILM Holdings Corporation;
Kenji Suzuki, PhD, Chicago, IL (Presenter) Royalties, General Electric Company; Royalties, Hologic, Inc; Royalties, MEDIAN
Technologies; Royalties, Riverain Technologies, LLC; Royalties, Canon Medical Systems Corporation; Royalties, Mitsubishi
Corporation; Royalties, AlgoMedica, Inc
For information about this presentation, contact:
jliu118@hawk.iit.edu
PURPOSE
To reduce cumulative radiation exposure and lifetime risks for radiation-induced cancer from breast cancer screening, we developed
novel NNC deep learning for radiation dose reduction in DBT.
METHOD AND MATERIALS
Our original NNC deep learning employed patched-based neural network regression in a convolutional manner to convert lower-dose
(LD) to higher-dose (HD) tomosynthesis images. We trained NNC with quarter-dose (25% of the standard dose: 12mAs at 32kVp)
raw-projection images and corresponding "teaching" higher-dose (HD) images (200% of the standard dose: 99mAs at 32kVp) of a
breast cadaver phantom acquired with a DBT system (Selenia Dimensions, Hologic). Once trained, NNC no longer requires HD
images. It converts new LD images to images that look like HD images; thus the term "virtual" HD (VHD) images. We reconstructed
tomosynthesis slices on a research DBT system. To determine a dose reduction rate, we acquired 4 studies of another test
phantom at 4 different doses (1.35, 2.7, 4.04, and 5.39mGy entrance dose). Structural SIMilarity(SSIM) index was used to evaluate
the image quality. For further testing, we collected half-dose (50% of the standard dose: 32±14 mAs at 33±5 kVp) and full-dose
(100% of the standard dose: 68±23mAs at 33±5kvp) images of 51 clinical cases with the DBT system at Univ. of Iowa Hospitals &
Clinics. We evaluated resulting images in a blinded observer study with 35 breast radiologists to rate and distinguish blinded VHD
and real full-dose DBT images.
RESULTS
NNC converted quarter-dose images (1.35mGy; SSIM: 0.88) of the testing cadaver phantom to VHD images with image quality
(SSIM:0.97) equivalent to 119% dose images (6.41mGy), achieving 79% dose reduction. In our blinded observer study, 21(60%) of
35 breast radiologists either preferred VHD images over real full-dose images or could not distinguish between the two. The
difference in image quality between the two was not statistically significant (P=0.37). The time required to process each study was
0.48 sec. on a GPU (GTX Titan Z, Nvidia).
CONCLUSION
Blinded observer study with 35 radiologists demonstrated that VHD images converted by our deep-learning technology were
equivalent to full-dose DBT images. Our cadaver phantom experiment demonstrated 79% dose reduction.
CLINICAL RELEVANCE/APPLICATION
Substantial radiation dose reduction would benefit patients by reducing the lifetime risk of radiation-induced cancer from DBT
screening.
SSE23-02 Centralized Physics QC Program for the ECOG-ACRIN TMIST Trial of Breast Tomosynthesis
 Monday, Nov. 26 3:10PM - 3:20PM Room: S502AB

Participants
Martin J. Yaffe, PhD, Toronto, ON (Presenter) Research collaboration, General Electric Company; Shareholder, Volpara Health
Technologies Limited; Co-founder, Mammographic Physics Inc; Research Consultant, BHR Pharma LLC
Etta D. Pisano, MD, Charleston, SC (Abstract Co-Author) Researcher, Freenome Holdings Inc; Researcher, Real Imaging Ltd;
Researcher, Therapixel; Researcher, DeepHealth, Inc; Researcher, ToDos
Aili K. Maki, BEng, Toronto, ON (Abstract Co-Author) Research collaboration, General Electric Company; Contractor, Mammographic
Physics, Inc
James G. Mainprize, PhD, Toronto, ON (Abstract Co-Author) Institutional research agreement, General Electric Company
Gordon Mawdsley, BS, Toronto, ON (Abstract Co-Author) Director, Medical Physics Incorporated Research collaboration, General
Electric Company
Sam Shen, Toronto, ON (Abstract Co-Author) Employee, Mammographic Physics Inc; Research collaboration, General Electric
Company
Ruth C. Carlos, MD, MS, Ann Arbor, MI (Abstract Co-Author) Nothing to Disclose
Kathy D. Miller, MD, Indianapolis, IN (Abstract Co-Author) Nothing to Disclose
Christopher E. Comstock, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
martin.yaffe@sri.utoronto.ca
PURPOSE
To describe and provide preliminary results from a remote-monitoring QC program developed to provide assessment of quality and
rapid feedback in a screening trial. The program is being used in the randomized TMIST trial of screening with breast tomosynthesis
versus digital mammography. TMIST is expected to include 125 sites in the US and Canada and will recruit 164,986 women who will
be imaged up to 5 times over 4 years.
METHOD AND MATERIALS
The QC program is based on imaging of phantoms by the technologist at each site and digital transmission to a central analysis
server. Phantoms assess signal and noise properties, artifacts, spatial resolution and geometric fidelity of the imaging system. The
analysis is performed automatically with results made available to technologists on a password protected web site. Technical
information from the DICOM header, stripped of personal identifiers, from every clinical image is available for analysis of doses,
exposure factors and compression parameters..
RESULTS
As of April 2018, initial QC data from 87 units at the first 29 TMIST sites were available, including de-identified screening
mammogram header data from 60 units at the first 25 sites. The most frequent technical problems were due to electronic
interference, dustlike artifacts and the compression force being reported in the header as '0. Problems were also noted due to
duplication of image submission from the same individual as separate cases and noncompliance with the QC protocol. In addition, it
was noted that digital detectors were occasionally replaced without technical documentation. This was accompanied by changes in
signal-to-noise performance. Based on 881 examinations, the mean dose (CC + MLO) was 4.2 mGy for 2D digital mammograms and
8.2 mGy for tomosynthesis. The presentation will report on results up to November, 2018.
CONCLUSION

Use of a centralized remote data collection QC system reduces technologist labor at the site and reduces subjectivity in testing.
This approach enables consistent analysis and rapid reporting of QC results.
CLINICAL RELEVANCE/APPLICATION
Sensitivity and specificity of breast cancer detection depend critically on the technical image quality. The credibility of the results
from the TMIST trial requires that the image quality of both modalities is verified. In addition, experience from this trial will provide
data to help define the essential elements of the standard QC program for tomosynthesis.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Ruth C. Carlos, MD, MS - 2015 Honored EducatorRuth C. Carlos, MD, MS - 2018
Honored Educator
SSE23-03 Visual Grading Characteristics Analysis of Propagation-Based X-Ray Phase Contrast Mammography
Monday, Nov. 26 3:20PM - 3:30PM Room: S502AB

Participants
Seyedamir Tavakoli Taba, Sydney, Australia (Abstract Co-Author) Nothing to Disclose
Sarah J. Lewis, PhD,MEd, Sydney, Australia (Abstract Co-Author) Nothing to Disclose
Patrycja Baran, Parkville, Australia (Abstract Co-Author) Nothing to Disclose
Matthew Dimmock, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Mikkaela McCormack, Heidelberg, Australia (Abstract Co-Author) Nothing to Disclose
Sheridan Mayo, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Yakov Nesterets, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Christopher Hall, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Jane Fox, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Zdenka Prodanovic, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Darren Lockie, FRANZCR, Southbank, Australia (Abstract Co-Author) Software support, Hologic, Inc
Harry Quiney, Parkville, Australia (Abstract Co-Author) Nothing to Disclose
Daniel Hausermann, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Giuliana Tromba, PhD, Trieste, Italy (Abstract Co-Author) Nothing to Disclose
Andrew Peele, Clayton, Australia (Abstract Co-Author) Nothing to Disclose
Timur Gureyev, Parkville, Australia (Abstract Co-Author) Nothing to Disclose
Patrick C. Brennan, PhD, Lidcombe, Australia (Presenter) Nothing to Disclose
For information about this presentation, contact:

amir.tavakoli@sydney.edu.au
PURPOSE

While all current x-ray based breast imaging modalities rely on minimal differences in soft tissue x-ray attenuation (absorption
contrast), phase-contrast imaging has the capacity to also visualise variations in x-ray refraction (phase contrast). For x-ray
energies typically used in breast imaging, the phase contrast can be substantially larger than the absorption contrast, presenting
an opportunity to improve soft tissue visualisation especially in mammographically dense breasts. The goal of this study was to
evaluate the radiological quality of images produced by the x-ray propagation-based phase-contrast computed tomography (PB-CT)
technique at two different x-ray energies in comparison to absorption-based CT images collected at the same radiation dose (4
mGy).
METHOD AND MATERIALS

Twenty-seven synchrotron-based CT images of a full-size breast mastectomy specimen were reconstructed. Nine images were
absorption-based CT at 32 KeV, nine images were PB-CT at 32 KeV and nine were PB-CT at 38 KeV. A group of breast specialist
radiologists and medical imaging experts compared the radiological quality of the three sets of images based on various image
quality criteria. Visual grading characteristics (VGC) analysis was conducted and VGC curves were obtained. The area under the
VGC curve (0<=AUCVGC<=1) was calculated as the measure of the difference in image quality between two compared sets of
images.
RESULTS
The results show that the radiological quality ratings of PB-CT 32 KeV images were significantly higher than absorption-based CT
images (AUCVGC=0.879, p<=.001) and PB-CT 38 KeV images (AUCVGC=0.795, p<=.001). The image quality ratings were not
significantly different between PB-CT 38 KeV images and absorption-based CT images (AUCVGC=0.567, p=.076).
CONCLUSION

Phase-contrast PB-CT mammography can be used to produce images with substantially higher radiological quality compared to
conventional absorption-based images, but this advantage appears to be dependent on beam energy. The results from this study
should provide a strong basis for future experimental and clinical protocols for further optimisation of this novel and promising
approach to breast imaging.
CLINICAL RELEVANCE/APPLICATION
PB-CT of the breast is expected to deliver improved image quality compared to current x-ray modalities and become a viable
method for early diagnosis of breast cancer in the future.
SSE23-04 Evaluation of American College of Radiology (ACR) Mammography Accreditation Phantom Image
Quality of a Grid-Less and Software-Based Scatter Correction Technology
Monday, Nov. 26 3:30PM - 3:40PM Room: S502AB

Participants
Anzi Zhao, MS, Cleveland, OH (Presenter) Nothing to Disclose
Katie Hulme, MS, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
PURPOSE
To evaluate the quality of ACR mammography accreditation phantom images acquired with a grid-less and software-based scatter
correction technology - Progressive Reconstruction Intelligently Minimizing Exposure (PRIME).
METHOD AND MATERIALS

3 Siemens Mammomat Inspiration units with PRIME were utilized in this study. The same ACR phantom was imaged on all units. 20
2D phantom images were acquired on each unit using a phototimed technique (W/Rh, 28kVp, AEC segmentation off, dose level
'normal', exam tag 'QC RAW'), of which 10 were acquired with grid in position and 10 were acquired with PRIME. Mode of acquisition
was varied in a random order. 10 additional PRIME images were acquired on one unit with a resolution test pattern to assess spatial
resolution. Contrast-to-noise ratio (CNR), signal-to-noise ratio (SNR), and standard deviation (SD) of phantom images were
evaluated using the method in Siemens quality control manual. Incident air kerma and average glandular dose (AGD) were measured
and calculated for each exposure. A total of 60 phantom images were scored by 4 qualified medical physicists and 2 experienced
mammography technologists on a diagnostic workstation in clinical viewing conditions, and using ACR phantom evaluation guidelines
with demographics hidden.
RESULTS

With PRIME, all images failed CNR criteria (>=2) with significantly lower CNR and higher SD than grid-based images by as much as
43% and 23%, respectively; SNR was reduced by 2-4%; spatial resolution was unaffected at 7 mm/lp; AGD was reduced by up to
16%. Visual scoring by 6 viewers resulted in no significant difference between the two types of images. Minor degradation on
average score of masses on PRIME images compared to grid-based images (4.1 vs 4.2) was noted on one unit. All viewers agreed
on the notable difference in image appearance and noise texture when PRIME was employed.
CONCLUSION
PRIME didn't penalize ACR phantom scoring, although there was a significant degradation of CNR on PRIME images because of the
increased noise. The clinical implications of differences in noise texture warrant further investigation. Although PRIME offers
moderate dose savings, clinicians should still be aware of potential effects on image appearance.
CLINICAL RELEVANCE/APPLICATION

PRIME technology corrects scatter radiation and enables grid-less full field digital mammography at lower patient average glandular
PRIME technology corrects scatter radiation and enables grid-less full field digital mammography at lower patient average glandular
dose, with comparable image quality.
SSE23-05 Mammographic Compression Variability Increased after Removing Real-Time Pressure Indicator
Monday, Nov. 26 3:40PM - 3:50PM Room: S502AB

Participants
Monique G. van Lier, MSc, Amsterdam, Netherlands (Presenter) Employee, SigmaScreening BV
Jerry E. De Groot, PhD, Amsterdam, Netherlands (Abstract Co-Author) Employee, SigmaScreening BV
Woutjan Branderhorst, PhD, Amsterdam, Netherlands (Abstract Co-Author) Employee, SigmaScreening BV
Laura J. Schijf, MD, Amsterdam, Netherlands (Abstract Co-Author) Nothing to Disclose
Cornelis A. Grimbergen, PhD, Amsterdam, Netherlands (Abstract Co-Author) Founder, SigmaScreening BV Employee, SigmaScreening
BV Board Member, SigmaScreening BV Patent holder, SigmaScreening BV
Gerard J. den Heeten, MD,PhD, Nijmegen, Netherlands (Abstract Co-Author) Founder, SigmaScreening BV; Scientific Advisor,
SigmaScreening BV; Patent Holder, SigmaScreening BV; Stock options, Volpara Health Technologies Limited; Medical Advisory
Board, Volpara Health Technologies Limited
For information about this presentation, contact:

m.g.vanlier@amc.nl
CONCLUSION
When replacing a paddle with a pressure indicator, in a group of technicians familiar with the indicator, by a conventional paddle,
the variability increased significantly leading to more unfavorable over- and under-compression.
Background
A certain level of breast flattening in mammography is needed to obtain a high quality image. Generally accepted and quantifiable
standards do not exists. Recent studies show that the level of compression pressure at exposure influences screening performance.
Attempts are made to standardize the compression procedure by introducing pressure-based compression using a paddle equipped
with a real-time pressure indicator. We aimed to study the impact on compression practice when replacing the pressure-based
paddle with a conventional paddle without pressure indication in group experienced technicians.
Evaluation

Mammographic compression pressure was retrospectively obtained from mammographic images (VolparaAnalytics) and evaluated in
two datasets from the same radiology department with the same technician team. The first dataset (4 years, n=11.561
compressions) was collected when using a compression paddle equipped with a real-time pressure indicator aiming for a 10kPa
(75mmHg) compression pressure. The second dataset (3 months, n=1331 compressions) was collected 4 months after the
mammography system with pressure indicator was replaced by a system without pressure indicator. The average compression
pressure and variance significantly (P<0.001) increased from 11.23 ± 0.04 kPa to 11.60 ± 0.14 kPa (mean ± SEM) after removal of
the pressure indicator. The proportion of compressions in the pressure range 5-15 kPa decreased from 87.4% to 77.9%. The
proportion of high pressures (>15kPa) almost doubled (11.0% to 18.8%) and low pressures (<5kPa) more than doubled (1.6% to
3.3%).
Discussion
When removing the pressure indicator, the initially low variability is increasing rapidly, indicating that an indicator is needed to
remain high compression reproducibility. An increase in over- and under-compression can ultimately lead to decreased
mammographic performance.
SSE23-06 Development of Low Dose Digital Mammography Platform by Image Reconstruction Using Deep
Learning Algorithm: A Preliminary Study
Monday, Nov. 26 3:50PM - 4:00PM Room: S502AB

Participants
Su Min Ha, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Eunhee Kang, Daejon, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jong Chul Ye, PhD, Daejon, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE

To investigate whether low dose mammography can be reconstructed to standard dose mammography using the new deep learning
algorithm.
METHOD AND MATERIALS

14 specimens from 14 patients who underwent total mastectomy for primary breast cancer were included. Specimen mammograms
were obtained with standard routine dose and reduced sequential doses; 80% of routine dose, 60%, 40%, 20% and 10%. The
proposed de-noising method is designed based on semi-supervised learning with cycle consistency loss. Most of the mammography
has Automatic Exposure Control (AEC) system which chooses an appropriate current X-ray source. The routine dose and 20% dose
level images were selected as training dataset. Since the noise levels between two images are different and unavoidable slight
mismatch due to potential deformation between multiple acquisitions, we developed the semi-supervised learning using cyclic
consistency. We trained two generators (network G and F) and two discriminators (network Dx and Dy). Since we had 14 datasets,
we performed cross-validation. Last, image quality of reconstructed low dose image was compared with the standard full dose
image and was qualitatively rates as follows; 1= poor, 2= fair, 3= equal, 4=better.
RESULTS

As more radiation dose was decreased, noise was increased and contrast resolution was decreased accordingly. However, in the
reconstructed images, noise was decreased and contrast resolution was rather improved. Overall, when we evaluated the lesions
according to Breast imaging-reporting and data system lexicon, and with consideration of underlying breast parenchyma density,
the reduced dose of 20% cut-off of standard full dose showed no significant difference in image quality compared with standard
dose mammography.
CONCLUSION

The image quality of reconstructed low dose mammography using the new deep learning algorithm is comparable with standard dose
mammography until dose reduction cut-off 20% of standard full dose. Therefore, the radiation dose of mammography could be
considerably reduced using this deep learning algorithm.
CLINICAL RELEVANCE/APPLICATION
Image reconstruction using the new deep learning algorithm is effective in dose reduction of mammography, especially in young
women with high risk who are routinely examined with mammography for screening.
 VW 09

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Monday, Nov. 26 3:00PM - 3:30PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 60

The Clinical Benefits of Tomosynthesis Guided Breast Biopsy: Hologic Vendor Workshop
Monday, Nov. 26 3:00PM - 4:00PM Room: Booth 5524

Participants
Lars J. Grimm, MD, Durham, NC (Presenter) Editorial Advisory Board, Medscape, LLC; Educational program support, Hologic, Inc
Program Information
Clinical experience of the 3D prone biopsy system and the benefits to the patient followed by hands on demonstration of the
Affirm® Prone Biopsy System (Affirm® Prone Biopsy System)
Registration
https://hologicrsna.com
 VW 34

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Reading Sessions: Siemens
Healthineers Vendor Workshop
Monday, Nov. 26 3:50PM - 5:00PM Room: Booth 5530

Participants
Luis Pina, MD, PhD, San Sebastian, Spain (Presenter) Nothing to Disclose
Program Information
Learn about the value of wide-angle Digital Breast Tomosynthesis (DBT) and Contrast Enhanced Mammography (CEM) in the daily
routine from one of our most experienced clinical experts. The differences and respective advantages of the morphological (DBT)
and functional (CEM) breast imaging methods will be discussed. This, all with the flexible assistance of our multi-modality reading
solution syngo.via and the syngo® Breast Care applications.
 VW 10

3D ABUS: Hands-on Workshop: GE Vendor Workshop

Monday, Nov. 26 4:00PM - 5:00PM Room: Booth 8156

Program Information
Peer educator led workshop includes a live Invenia ABUS 2.0 scan station acquisition and hands-on review of clinical cases using
the Invenia ABUS Viewer. Learn about the importance of the coronal view and how 3D ABUS screening helps increase cancer
detection in women with dense breast tissue. Registration is required; adding this session to the RSNA calendar tool alone does not
secure your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 61

Advancements in Real-time Breast Specimen Imaging: Hologic Vendor Workshop

Monday, Nov. 26 4:30PM - 5:00PM Room: Booth 5524

Program Information
Learn how real-time breast specimen imaging can streamline workflow efficiency through specimen verification and automated post-
biopsy specimen handling with the Brevera® Breast Biopsy System. (Brevera® Breast Biopsy System)
Registration
https://hologicrsna.com
 ED001-TU

Breast Tuesday Case of the Day

Tuesday, Nov. 27 7:00AM - 11:59PM Room: Case of Day, Learning Center

AMA PRA Category 1 Credit ™: .50

Participants
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Susan O. Holley, MD, PhD, Raleigh, NC (Abstract Co-Author) Nothing to Disclose
Alexander B. Sevrukov, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Catherine A. Young, MD, JD, Saint Louis, MO (Abstract Co-Author) Research support, Hologic, Inc
Cheryl R. Herman, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Mai A. Elezaby, MD, Madison, WI (Abstract Co-Author) Research Grant, Exact Sciences Corporation
Lonie R. Salkowski, MD,PhD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Urvi A. Tailor, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Ramapriya Ganti, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1) Identify, characterize, and analyze abnormal findings on multimodality breast imaging studies. 2) Develop differential diagnostic
considerations based on the clinical information and imaging findings. 3) Recommend appropriate management for the patients
based on imaging findings.
 SPSH30

Hot Topic Session: Management of DCIS and Minimal Risk Lesions

Tuesday, Nov. 27 7:15AM - 8:15AM Room: E450B

BR
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

FDA Discussions may include off-label uses.

Participants
Linda Moy, MD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
SPSH30A LORIS Trial-Will it Address the Overtreatment of Low Grade DCIS?
Participants
Matthew G. Wallis, MD, Cambridge, United Kingdom (Presenter) Nothing to Disclose
For information about this presentation, contact:
Matthew.wallis@addenbrookes.nhs.uk
LEARNING OBJECTIVES
1) To learn about the risk of overdiagnosis and overtreatment. 2) To become familiar with the risk of avoiding surgery. 3) To
appreciate the prospective for the future.
ABSTRACT
Breast screening has led to an inexorable rise in the number of women living with a diagnosis of DCIS predominantly but not
exclusively as a result of the detection of micro-calcification. It is clear from long term follow up that conventional treatment fails
the 15 to 20% who develop invasive disease a few of whom then die from breast cancer. There is larger group that never progress
with in their life time which means we have either cured them or just over treated them. Traditional pathology and genetics suggest
that there is a low risk group that either never progress or if they do they develop 'low risk' invasive disease. I will describe the 4
international trials LORIS (The LOw RISk DCIS trial), LORD (LOw Risk DCIS) , COMET (Comparison of Operative versus Medical
Endocrine Therapy for Low Risk DCIS) and LORETTA (Low Risk Tamoxifen Treatment And surveillance ) . In particular commenting on
their differences. Using LORIS as my main example I will discuss how we have attempted to resolve the hurdles of setting up a 'no
treatment trial' and talk about some of the lessons learnt from successfully steering LORIS from feasibility to full trial. Anecdotally
patient views have been more entrenched than those of surgeons
SPSH30B Pathologic Interpretation of Borderline Breast Lesions
Participants
Laura C. Collins, MD, Boston, MA (Presenter) Nothing to Disclose
For information about this presentation, contact:
lcollins@bidmc.harvard.edu
LEARNING OBJECTIVES
1) To have a greater level of understanding of the nuances of interpreting atypical ductal proliferations. 2) To have a greater level
of understanding of the quantitative criteria for a diagnosis of low grade ductal carcinoma in situ. 3) Learners will have a greater
level of understanding of the limitations of providing definitive diagnosis of some atypical ductal proliferations on core needle biopsy
samples.
ABSTRACT
This presentation will cover the nuances of pathologic interpretation of borderline atypical ductal proliferations of the breast,
particularly in the setting of core needle biopsy samples. Features that favor a diagnosis of atypical ductal hyperplasia versus low
nuclear grade ductal carcinoma in situ will be discussed. Finally, the ramifications of definitive distinctions rendered on borderline
lesions will be reviewed.
SPSH30C High Risk DCIS and What are We Prepared To Do About it
Participants
Brian J. Czerniecki, MD, PhD, Tampa, FL (Presenter) Advisory Board, ImmunoRestoration
LEARNING OBJECTIVES
1. Understand that there are subgroups of women with DCIS in whom there is an increased risk of subsequent breast cancer deaths
and they include women under 40, African American females and those with hormone receptor negative breats cancer. 2.
Disseminated cancer cells may seed even from very early DCIS lesions 3. Surgery and Radiation do not appear to reduce the risk of
death from breast cancer in high risk groups
ABSTRACT
Controversy exists concerning whether all ductal carcinoma in situ (DCIS) lesions represent premalignant precursors destined to
become invasive breast cancer (IBC). There is also evidence that cancer cells can disseminate very early during the process of
tumorigenesis even prior to invasive breast cancer developing. There is a also evidence that certain groups of patients with DCIS
have increased risk of mortality from DCIS and they include those diagnosed under 40, African American females and those that are
hormone receptor negative. Local therapy with surgery and radiation does not decrease risk of mortality from DCIS. Many of these
high risk lesions have expression of HER2 which is associated with both invasion and dissemintation of cancer cells (DCC). Targeting
HER2 maybe benficial for eliminating both DCC and reducing risk of subsequent IBC. Dendritic cell (DC) vaccines administered in a
neoadjuvant trials to patients with HER2 expressing DCIS has resulted in complete regression of disease in about 30% of DCIS
patients. Vaccines drive anti-HER2 CD4 Th1 cells especially in sentinel nodes. Complete regression of DCIS is associated with
diminshed breast cancer events. Combinations of other immune therapies with DC vaccines may result in greater numbers of
complete responses. In summary, high grade and HER2 expressing DCIS in young women, African American females requires novel
approaches to reduce mortlaity and subsequent breast cancer events.
 RC315

Breast Series: MRI

Tuesday, Nov. 27 8:30AM - 12:00PM Room: Arie Crow n Theater

BR MR
AMA PRA Category 1 Credits ™: 3.50
ARRT Category A+ Credits: 4.00

FDA Discussions may include off-label uses.

Participants
Bonnie N. Joe, MD, PhD, San Francisco, CA (Moderator) Nothing to Disclose
Hiroyuki Abe, MD, Chicago, IL (Moderator) Nothing to Disclose
For information about this presentation, contact:
zuleyml@upmc.edu
habe@radiology.bsd.uchicago.edu
Sub-Events
RC315-01 Outcome Data: Does MRI Help?
Tuesday, Nov. 27 8:30AM - 8:50AM Room: Arie Crow n Theater

Participants
Francesco Sardanelli, MD, San Donato Milanese, Italy (Presenter) Speakers Bureau, Bracco Group; Advisory Board, Bracco Group;
Research Grant, Bayer AG; Advisory Board, General Electric Company; Reserach Grant, General Electric Company; Speakers Bureau,
Siemens AG; Reserach Grant, Real Imaging Ltd;
For information about this presentation, contact:
francesco.sardanelli@unimi.it
LEARNING OBJECTIVES
1) Know the current debate about the need of outcome data for an appropriate use of breast MRI in clinical practice. 2)
Understand why the acceptance of breast MRI has a large variability by other clinicians depending on indication, from screening of
BRCA1/2 or P53 mutated women to the preoperative setting. 3) Appraise the high complexity of the current debate on the evidence
in favor or against preoperative breast MRI. 4) Identify those applications where more research is needed for an increased use of
breast MRI, also considering the perspective of prognostic breast MRI.
RC315-02 Comparison of Diagnostic Performance of DBT and MRI Added to Mammography for Preoperative
Staging of Screening-Detected Breast Cancer: Which Method Is More Appropriate Depending On the
Mammographic Density?
Tuesday, Nov. 27 8:50AM - 9:00AM Room: Arie Crow n Theater

Participants
So Yeon Yang, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Ji Soo Choi, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Ko, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Boo-Kyung Han, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ko Woon Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Mi-Ri Kwon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
E-Ryung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Surin Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
soyeon.y6985@gmail.com
PURPOSE
To compare the diagnostic performance of digital breast tomosynthesis (DBT) and magnetic resonance imaging (MRI) added to
mammography for preoperative staging of screening-detected breast cancer depending on mammographic density.
METHOD AND MATERIALS
This retrospective study enrolled 281 patients with 332 screening-detected breast cancers recruited from Jan to Dec 2013. Three
radiologists independently reviewed three image sets of (mammography alone, DBT plus mammography and MRI plus mammography)
of the patients, and they recored final BI-RADS categories of detected lesions. BI-RADS categories 4-5 defined positive results and
BI-RADS category 1-3 defined negative results. Readers' sensitivities and positive predictive values (PPVs) were analyzed for each
reading mode. Readers' performances with the three reading modes were compared for dense breast (heterogeneously or extremely
dense, n=263) and non-dense breast (entirely fatty or scattered areas of fibroglandular density, n=120) groups, respectively.
RESULTS
In non-dense breast group, readers' sensitivities with DBT plus mammography (92.5-94.4%) were lower than MRI plus
mammography (96.3-98.1%), but higher than mammography alone (88.8-92.5%). Readers' PPVs with DBT plus mammography (97.1-
100%) were higher than those with MRI plus mammography (94.7-100%) and mammography alone (94.7-97.0%). However, there
was no statistically significant difference in both readers' sensitivities and PPVs between DBT plus mammography and MRI plus
mammography (p>0.05). In dense breast group, sensitivities with MRI plus mammography (93.3-98.2%) were significantly higher
than those with DBT plus mammography (87.6-92.0%) or mammography alone (84.9-87.6%) (p<0.05), but PPVs with MRI plus
mammography (92.1-97.5%) were lower than those with DBT plus mammography (96.1-97.6%) or mammography alone (96.1-
97.5%) without a statistical significance.
CONCLUSION
In non-dense breast group, diagnostic performances of DBT and MRI for preoperative staging of screening-detected breast cancer
were not significantly different when using as an adjunctive to mammography. In dense breast group, however, DBT had lower
sensitivitiy than MRI.
CLINICAL RELEVANCE/APPLICATION
In non-dense breast group, DBT plus mammography may provide similar diagnostic performance to MRI plus mammography for
preoperative staging of screening-detected breast cancer.
RC315-03 Pre-Chemotherapy Morphology and ADC Characteristics of Primary Breast Cancers Vary By
Hormone-Receptor and HER2 Subtype
Tuesday, Nov. 27 9:00AM - 9:10AM Room: Arie Crow n Theater

Participants
Bo La Yun, MD, Seongnam, Korea, Republic Of (Presenter) Nothing to Disclose
Wen Li, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Ella F. Jones, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Lisa Wilmes, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
David Newitt, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jessica Gibbs, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Nola M. Hylton, PhD, San Francisco, CA (Abstract Co-Author) Research support, General Electric Company
PURPOSE

To evaluate the hormone-receptor (HR) and HER2 subtype dependence of pre-treatment MRI morphology and apparent diffusion
coefficient (ADC) characteristics of primary breast cancers.
METHOD AND MATERIALS

A retrospective analysis of DCE-MRI, DW-MRI and T2WI was performed on pre-treatment MRI studies of 220 breast cancer patients
who were enrolled in a neoadjuvant breast cancer trial. DCE-MRI and T2WI were reviewed according to the BI-RADS lexicon, and
MRI morphologic pattern was categorized using a 1-5 scale for tumor containment. Extent of necrosis and presence of peritumoral
edema were also ranked. ADC values at 5, 15, 25, 50, 75 and 95 percentile were computed from the DW-MRI based on an ROI
encompassing the entire tumor volume. Fisher's exact test was used to compare the morphologic features and one-way ANOVA and
Scheffe post hoc test were used to compare ADC measurements among all breast cancer subtypes.
RESULTS

The triple negative (TN) subtype exhibited mass more frequently than non-mass enhancement (NME) (p=0.004), with masses
showing irregular versus spiculated margin (p=0.034). HR-/HER2+ subtype had NME more frequently than mass (p=0.027).
HR+/HER2+ showed heterogeneous enhancement rather than rim enhancement (p<0.001). There was no specific pattern observed
in NME among subtypes. In the MRI morphologic pattern, TN showed a well-defined pattern with more than 10% necrosis versus
other subtypes. The difference in ADC values at the lower 5 and 15 percentiles was found to be statistically significant between TN
vs. HR-/HER2+ (p=0.007 in 5 percentile and p=0.014 in 15 percentile), HR-/HER2+ vs. HR+/HER2- (p=0.002 in 5 percentile and
p=0.004 in 15 percentile), and HR+/HER2- vs. HR+/HER2+ (p=0.028 in 5 percentile and p=0.014 in 15 percentile).
CONCLUSION

The BI-RADS lexicon (lesion classification, internal enhancement pattern and margin of the mass), MR morphologic pattern, and the
amount of necrosis may be useful for distinguishing breast cancer subtypes. Among the variable measurements, the lower 5 or 15
percentiles of the ADC distributions showed potential to distinguish breast cancer subtypes.
CLINICAL RELEVANCE/APPLICATION

Adding the lower 5 or 15 percentile ADC with MR morphologic patterns may help refine MRI methods for distinguishing breast cancer
subtypes prior to neoadjuvant chemotherapy.
RC315-04 Pre-Operative Breast Magnetic Resonance Imaging: Relationship Between Magnetic Resonance-
Detected Additional Cancer and Survival Outcomes
Tuesday, Nov. 27 9:10AM - 9:20AM Room: Arie Crow n Theater

Participants
Eun Sook Ko, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Ko Woon Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ji Soo Choi, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE

To determine the additional cancer detection yield of pre-operative MRI in women with invasive breast cancer that was occult on
mammogrpahy and ultrasonography (US), to identify a subgroup of women who are likely to have additional cancer, and to
investigate whether the presence of MRI-detected additional cancer (MDAC) affects patients' long-term survival outcomes.
METHOD AND MATERIALS
The pre-operative MRI examinations of 1,843 women who had undergone surgery for invasive breast cancer were reviewed for the
presence of additional multifocal/multicentric /contralateral disease that was occult on mammgoraphy and US. Clinicopathological
findings and mammographic breast density were compared between patients with MDAC and those without. Logistic regression
analysis was conducted to find factors associated with MDACs. A Cox proportional hazards model was used to analyze the effects
of MDACs or other variables on disease-free survival (DFS) or overall survival (OS). Kaplan-Meier curves and log-rank tests were
used to analyze survival between the two groups.
RESULTS
Of 1,843 patients, 178 (9.7%) had an MDAC. Multivariate analysis showed that invasive lobular cancer (odds ratio: 1.151, 95%
confidence interval [CI]: 1.080, 1.239; P = 0.0002) and extensive intraductal component (odds ratio: 1.113, 95% CI: 1.080, 1.148;
P < 0.0001) were independently associated with a higher probability of MDAC. Kaplan-Meier curves did not show that MDACs
affected DFS (P = 0.343) or OS (P = 0.991).
CONCLUSION

MDACs had no significant impact on survival outcomes.

CLINICAL RELEVANCE/APPLICATION

No studies have focused on survival outcomes in MRI-detected additional cancers (MDACs) that were occult at mammography and
ultrasonography (US).
RC315-06 Long-Term Survival Outcomes in Invasive Lobular Carcinoma Patients with and Without Preoperative
MR Imaging: A Matched Cohort Study
Tuesday, Nov. 27 9:30AM - 9:40AM Room: Arie Crow n Theater

Participants
Su Min Ha, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE
To investigate and compare the effect of preoperative breast magnetic resonance (MR) imaging on recurrence-free survival (RFS)
and overall survival (OS) outcomes among patients with invasive lobular carcinoma (ILC).
METHOD AND MATERIALS
A total of 287 ILC patients (age range, 31-82 years; mean age, 49.8 years) between January 2005 and December 2012 were
included in the analysis. Of these patients, 120 (41.8%) had undergone preoperative breast MR imaging (MR group) and the
remaining 167 (58.2%) had not (no MR group). These two study groups were matched for 21 covariates in term of patient
demographics, tumor characteristics, and various clinical features. The RFS and OS outcomes were compared using Kaplan-Meier
estimates. MR effects were estimated after adjusting for significant potential confounders of specific outcomes in the multivariate
modeling.
RESULTS

In the matched cohort, no statistically significant association was observed between MR imaging and total recurrence (hazard ratio
[HR], 1.096; 95% CI: 0.497-2.416; P=0.821), loco-regional recurrence (HR, 1.204; 95% CI: 0.294-4.924; P=0.796), contralateral
breast recurrence (HR, 0.945; 95% CI: 0.147-6.061; P=0.952), or distant recurrence (HR, 1.020; 95% CI: 0.339-3.070; P=0.973).
MR imaging was associated with an improved OS with 51% reduction, but not significantly (HR, 0.485; 95% CI: 0.149-1.585;
P=0.231). Analysis with a multivariate Cox regression model indicated that MR imaging was not a significant independent factor for
better RFS (HR, 0.823; 95% CI: 0.409-1.658; P=0.586) or improved OS (HR, 0.478; 95% CI: 0.167-1.366; P=0.168).
CONCLUSION

Preoperative MR imaging is not a prognostic factor and produces no recurrence or survival outcome benefits in ILC patients.
RC315-07 Preoperative Breast MRI: Multicenter Prospective Study
Tuesday, Nov. 27 9:40AM - 9:50AM Room: Arie Crow n Theater

Participants
Fusun Taskin, MD, Istanbul, Turkey (Presenter) Nothing to Disclose
Nermin Tuncbilek, Edirne, Turkey (Abstract Co-Author) Nothing to Disclose
Gulden Acunas, MD, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Pinar Balci, MD, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Gul Esen, MD, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Burcin Tutar, MD, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Fahrettin Kilic, MD, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Aysenur Oktay, MD, Izmir, Turkey (Abstract Co-Author) Nothing to Disclose
Levent Celik, MD,BA, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
Erkin Aribal, Istanbul, Turkey (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
fusuntaskin@yahoo.com
PURPOSE

To investigate the effect of breast MRI for preoperative staging on clinical evaluation and treatment plan in women diagnosed with
breast cancer.
METHOD AND MATERIALS

In the prospective, multicenter study, the institutional ethics committee approval was obtained for all centers. Conventional
imaging (mammography and ultrasonography) findings, preoperative breast MRI findings, treatment plan and histopathology results
were evaluated in 432 consecutive breast cancer patients at nine centers. Cases that were scheduled to receive neoadjuvant
chemotherapy were excluded. The effect of preoperative breast MRI added to conventional breast imaging on clinical-radiological
evaluation and on surgical treatment plan was investigated. Chi-square and McNemar tests were used for statistical analysis.
RESULTS

Two-hundred thirty-four cases (54.2%) were premenopausal and 198 cases (45.8%) were postmenopausal. Cancer was detected in
134 women (31%) at the time of screening, and cancer was found in 298 (69%) women who had undergone diagnostic radiological
evaluation due to complaints or physical examination. Physical examination was positive in 248 (57%) women and negative in 184
(43%) women. 23 women had bilateral breast cancer. The frequencies of multifocal and multicentric tumor detection were 9-7%,
16-11%, 17-28% for MG, US and MRI, respectively. Breast-conserving surgery (BCS) was performed on a total of 210 cancers and
modified radical mastectomy (MRM) on 255 cancers. A total of 8 cases required re-excision surgery due to positive surgical margin.
MRI changed the surgical treatment plan in 14% of patients for whom BCS was planned based on conventional imaging. The
difference between the conventional imaging and MRI in the preoperative evaluation was considered statistically significant
(p=0.001).
CONCLUSION

Breast MRI added to conventional breast imaging in the preoperative evaluation of patients with breast cancer contributes to an
accurate treatment plan by lower need for re-excision surgery and providing accurate treatment of the 14% cases.
CLINICAL RELEVANCE/APPLICATION

This study showed that breast MRI contributes to the preoperative evaluation and treatment plan in women who were diagnosed
with breast cancer.
RC315-08 Ongoing Trials Update
Tuesday, Nov. 27 9:50AM - 10:10AM Room: Arie Crow n Theater

Participants
Christiane K. Kuhl, MD, Aachen, Germany (Presenter) Nothing to Disclose
For information about this presentation, contact:

ckuhl@ukaachen.de
LEARNING OBJECTIVES
To list the current studies published on the use of MRI for screening To list cancer detection rates and predictive values of
abbreviated MRI for screening in comparison to those of digital breast tomosynthesis and breast ultrasound.
RC315-09 Breast MRI-based Radiomics Nomogram for the Prediction of Recurrence in Patients with Triple-
negative Breast Cancer: A Nested Case-Control Matched Study
Tuesday, Nov. 27 10:10AM - 10:20AM Room: Arie Crow n Theater

Participants
Su Min Ha, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa Jung Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ki Chang Shin, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE
To develop a breast MRI-based radiomics nomogram including clinicopathologic factors for individualized prediction of local or distant
recurrences in patients with triple-negative breast cancers (TNBC).
METHOD AND MATERIALS
From 2006 to 2013, a total of 2604 patients were diagnosed as TNBC and 836 patients underwent preoperative breast MRI. Among
them, patients with recurrence and without recurrence were matched in terms of age, stage, and type of chemotherapy, and
developed 115 nested case-control pairs. Within the intratumor and peritumoral regions on early post-contrast T1-weighted images,
percent enhancement (PE) map, signal enhancement ratio (SER) map, and T2-weighted images, a total of 1029 quantitative MR
radiomic features, each referred to as a computer-extracted image phenotypes (CEIP), were calculated based on the
semiautomatically derived three-dimensional tumor segmentations. Elastic Net was used for feature selection and radiomics score
building. A radiomics nomogram was constructed from a multivariable logistic regression prediction model with the radiomics score
and independent pathologic predictors. We divided 115 case-control pairs into a training set (n=154) and a validation set (n=76),
and the internal validation for the validation set was performed.
RESULTS
The radiomics score, consisted of 20 selected CEIPs, was significantly associated with the prediction of recurrence (C-index of
0.867 for training set and 0.778 for validation set). Independent pathologic factors in the nomogram were lymphovascular invasion,
Ki-67 status, and lymph node ratio (C-index of 0.665 for training set and 0.668 for validation set). Radiomics nomogram showed
better prediction of recurrence (C-index of 0.879 for training set and 0.802 for validation set) due to incremental value of 0.214
and 0.134, respectively, by addition of radiomics score to the pathologic predictors.
CONCLUSION

Our results indicate that the radiomics nomogram which incorporates the MRI-based radiomics score and pathologic features, show
promise for the individualized prediction of local or distant recurrence in patients with TNBC.
CLINICAL RELEVANCE/APPLICATION

Nomogram using breast MRI-based radiomics score and pathologic predictors can facilitate the individualized prediction of
recurrence in patients with TNBC.
RC315-11 Advanced Sequences
Tuesday, Nov. 27 10:40AM - 11:00AM Room: Arie Crow n Theater

Participants
Nariya Cho, MD, PhD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
For information about this presentation, contact:
river7774@gmail.com
LEARNING OBJECTIVES

1) Review the standard dynamic contrast-enhanced MRI (DCE-MRI) and quantitative MRI using two compartment model. 2)
Understand the current status of abbreviated breast MRI. 3) Explore the clinical value of ultrafast DCE-MRI.
ABSTRACT

Breast dynamic contrast-enhanced (DCE)-MRI refers to MR imaging techniques with temporal resolution of 2 minutes or less to
assess the changes of contrast uptake and washout in tumors. Recent technological advances realize various combinations of
spatial and temporal resolution of breast MRI. Refined quantification (Ktrans, Ve, Kep) of exchange of contrast agent between
vascular space and interstitial space provide sophisticated hemodynamic information. Pre-contrast with only one post-contrast
image makes MRI screening more feasible by reducing time and cost while maintaining diagnostic performance. DCE-MRI with a 4 to
7-second temporal resolution during the first minute before a standard image acquisition shows the potential to improve lesion
conspicuity and characterization. This session will focus on the review of variations of breast DCE-MRI.
RC315-12 Agreement between Radiologist-Assigned Categories and Quantitative Measures of Background
Parenchymal Enhancement on Breast MRI
Tuesday, Nov. 27 11:00AM - 11:10AM Room: Arie Crow n Theater

Participants
Bethany L. Niell, MD,PhD, Tampa, FL (Presenter) Nothing to Disclose
Mahmoud Abdalah, PhD, Tampa, FL (Abstract Co-Author) Nothing to Disclose
Olya Stringfield, PhD, Tampa, FL (Abstract Co-Author) Nothing to Disclose
Malesa M. Pereira, MPH, Tampa, FL (Abstract Co-Author) Nothing to Disclose
Natarajan Raghunand, PhD, Tampa, FL (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
Bethany.niell@moffitt.org
PURPOSE

Because background parenchymal enhancement (BPE) on breast magnetic resonance imaging (MRI) reflects the volume and
intensity of contrast uptake, quantitative values of enhancement can be measured by 1) averaging the voxels of enhancement (PE
= percent enhancement) above a pre-defined threshold, 2) computing the total volume of FGT that enhances above the threshold
value (absolute volume of BPE), and 3) estimating the percentage of breast tissue that enhances above the threshold value
relative to the total breast volume (BPE%). We developed a semi-automated segmentation algorithm to extract these quantitative
measures of BPE. In this study, we investigated the agreement of computed measures of BPE with radiologist-assigned categories.
METHOD AND MATERIALS

In this IRB approved HIPAA compliant retrospective study, we identified 123 patients with breast MRI performed for screening
indications. As previously described, the breast segmentation algorithm co-registers pre- and post-contrast T1-weighted fat-
suppressed and non-fat-suppressed sequences. Active contours method merged chest components and non-fat voxels were
clustered using Otsu's method to identify fibroglandular tissue (FGT) voxels. Within the segmented FGT on the first post-contrast
phase, we computed median and inter-quartile ranges for absolute volume of BPE and BPE% using a PE=30% threshold. Student's t-
test evaluated BPE volume and BPE% by radiologist-assigned categories.
RESULTS

Using the previously described 30% threshold, median and inter-quartile ranges for the volume of BPE by radiologist-assigned
category were as follows (cm3): minimal (57.2, 24.4-100.1), mild (41.8, 30.7-65.4), moderate (70.6, 43.5-111.1), marked (67.1,
54.3-137.9). BPE% median and inter-quartile ranges were as follows (%): minimal (3.5, 1.7-5.5), mild (3.2, 1.7-4.6), moderate (4.7,
2.7-7.4), marked (5.7, 4.0-10.0). BPE volume and BPE% differed significantly between minimal/mild and moderate/marked
radiologist-assigned categories (p=0.030 and 0.004, respectively) (Figure: Box plot of BPE% by BPE category).
CONCLUSION
Quantified BPE volume and BPE% were significantly different between minimal/mild and moderate/marked radiologist-assigned
categories.
CLINICAL RELEVANCE/APPLICATION
Given the inter-reader variability in BPE categorical assessments, the development and validation of quantitative measures is a
necessary step towards incorporation of BPE into future risk prediction models.
RC315-13 Correlation between 3T Multi-Parametric MRI and Molecular Subtypes of Breast Cancer
Tuesday, Nov. 27 11:10AM - 11:20AM Room: Arie Crow n Theater

Participants
Stefania Montemezzi, MD, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Carlo Cavedon, DPhil, Verona, Italy (Presenter) Nothing to Disclose
Lucia Camera, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Maria Grazia Giri, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Alice Pozzetto, MD, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Maria Vittoria Bisighin, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Anna Calio, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Gabriele Meliado, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Francesca Caumo, MD, Padua, Italy (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

stefania.montemezzi@aovr.veneto.it
PURPOSE
To test whether 3T multi-parametric magnetic resonance imaging (mp-MRI) provides information related to molecular subtypes of
breast cancer.
METHOD AND MATERIALS

Women with mammographic or US findings of breast lesions (BI-RADS 4-5) underwent 3T mp-MRI (DCE, DWI and MR spectroscopy).
DCE-MRI was evaluated by classifying the wash-in/wash-out curve in three classes (I-III). DWI was used to calculate the mean
ADC value within a region of interest centered on the tumor. MR spectroscopy (MRS) was evaluated by means of the signal-to-
noise ratio (SNR) of the total choline peak (tCho). The histological type of breast cancer was assessed. Estrogen-receptor (ER),
progesterone-receptor (PgR), Ki-67 status and HER-2 expression, assessed by immunohistochemistry (IHC), were used to identify
four molecular subtypes: Luminal-A, Luminal-B, HER2-enriched and triple-negative tumors. Non-parametric tests (Kruskal-Wallis, k-
sample equality of medians, and Mann-Whitney) and logistic regression were performed to investigate correlations between mp-MRI
features (lesion volume, margins, ADC, type of DCE curve, and tCho SNR) and molecular subtypes.
RESULTS
483 patients (505 lesions) were included in the study. Volume was smaller in Luminal-B and larger in triple-negative tumors (non-
parametric tests, p<0.03 and p<0.004, respectively). A prevalence of irregular margins was observed in triple negative tumors
(p<0.01). The type of DCE curve was significantly different in Luminal-A (lack of type III curves compared to average, p<0.03).
ADC values were higher in Luminal-A (p<0.04 and p<0.016 in non-parametric tests and logistic regression, respectively). tCho SNR
was higher in triple-negative tumours (p<0.05 and p<0.01).
CONCLUSION
A significant correlation was found between some MRI features and molecular subtypes of breast tumors. The strongest correlations
were observed between Luminal A tumors and ADC, Luminal A tumors and DCE-MRI findings, Triple negative tumors and tCho SNR.
These results warrant further research to improve the prognostic value of multi-parametric MRI.
CLINICAL RELEVANCE/APPLICATION
Significant correlations were observed between multi-parametric MRI features and molecular subtypes of breast tumors. Further
research is needed to improve the prognostic value of mp-MRI.
RC315-14 Apparent Diffusion Coefficient Difference Value on Diffusion-Weighted Imaging: Association with
Distant Metastasis-Free Survival of Patients with Invasive Breast Cancer
Tuesday, Nov. 27 11:20AM - 11:30AM Room: Arie Crow n Theater

Participants
Jin You Kim, MD, Busan, Korea, Republic Of (Presenter) Nothing to Disclose
Jin Joo Kim, MD, Busan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Suk Kim, MD, Pusan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

youdosa@naver.com
PURPOSE
To investigate whether apparent diffusion coefficient (ADC) parameters on diffusion-weighted imaging (DWI) are associated with
distant metastasis (DM)-free survival in patients with invasive breast cancer.
METHOD AND MATERIALS

This retrospective study was approved by the institutional review board. The requirement to obtain informed consent was waived.
Between June 2013 and June 2014, 258 consecutive women (mean age, 50.9 years; age range, 23-85 years) with newly diagnosed
invasive breast cancer who underwent preoperative breast MR imaging with DWI were evaluated. All DWI were retrospectively
reviewed by two radiologists blinded to the clinical information. The mean, minimum, and maximum ADC values were measured by
manually placing regions of interest within the lesions and the ADC difference value (which is the difference between minimum and
maximum ADC) was calculated to evaluate intratumoral heterogeneity. Cox proportional hazards models were used to reveal the
associations between ADC parameters and DM-free survival after adjusting for clinicopathological factors.
RESULTS
In 25 (9.7%) patients, DM developed without prior locoregional recurrence at a mean follow-up of 48.7 months. The mean of ADC
difference value was significantly higher in patients with DM than in those without DM (0.781 × 10-3mm2/s vs. 0.620 × 10-
3mm2/s, P = .007). Kaplan-Meier survival analysis showed that patients with high ADC difference value (>0.793 × 10-3mm2/s) had
shorter DM-free survival times compared with those with low ADC difference value (<=0.793 × 10-3mm2/s) (log-rank test; P <
.001). Furthermore, multivariate Cox proportional hazards analysis showed that a high ADC difference value (>0.793 × 10-3mm2/s)
(hazard ratio [HR] = 3.448; 95% confidence interval [CI]: 1.567, 7.586; P = .002), presence of axillary node metastasis (HR =
5.101; 95% CI: 2.127, 12.234; P < .001), and estrogen receptor negativity (HR = 2.429; 95% CI: 1.104, 5.343; P = .027) were
associated with worse DM-free survival.
CONCLUSION

High ADC difference value on DWI was significantly associated with worse DM-free survival of patients with invasive breast cancer.
CLINICAL RELEVANCE/APPLICATION
Quantitative analysis of ADC difference value as a biomarker of intratumoral heterogeneity can be used to identify a subgroup of
breast cancer patients at higher risk of developing distant metastasis.
RC315-15 Bradiomics (Breast Radiomics) Can Improve Breast Cancer Detection: Preliminary Clinical Results
Using Multivariate Magnetic Resonance Tensor Modeling Fitting
Tuesday, Nov. 27 11:30AM - 11:40AM Room: Arie Crow n Theater

Participants
Anabel M. Scaranelo, MD, PhD, Toronto, ON (Presenter) Nothing to Disclose
Edna Furman-Haran, PhD, Rehovot, Israel (Abstract Co-Author) Nothing to Disclose
Hadassa Degani, PhD, Rehovot, Israel (Abstract Co-Author) Nothing to Disclose
Vivianne Freitas, MD, Toronto, ON (Abstract Co-Author) Nothing to Disclose
Dov Grobgeld, Rehovot, Israel (Abstract Co-Author) Nothing to Disclose
Karen Bodolai, Toronto, ON (Abstract Co-Author) Nothing to Disclose
Nancy Talbot, MSc, RT, Toronto, ON (Abstract Co-Author) Nothing to Disclose
Pavel Crystal, MD, Toronto, ON (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
anabel.scaranelo@uhn.ca
PURPOSE
To evaluate performance measurements of a radiomics model breast lesions extracted from 30 directions fitting of MR images
without gadolinium enhancement.
METHOD AND MATERIALS

Research ethics board approved this prospective study including data of 269 MR studies from patients of 3 institutions. All
consented women presented with clinically/imaging suspicious or a biopsy proven breast cancer and an indication for dynamic
contrast-enhanced (DCE) breast MRI. Before gadolinium injection, diffusion MR imaging (b values, 0, 800 sec/mm2) was performed
using a dedicated 3.0T scanner with 16-channel breast coil. A total of 7 readers independently assessed DCE where BPE, lesion size
and BIRADS category for each breast were recorded. Two readers blind to DCE results in consensus assessed the 11 features
extracted from pixel-by-pixel fitting modeling optimized to lambda-1 values. Histopathology was used as the gold standard.
Adequate statistical tests were used to compare the diagnostic values
RESULTS
There were 248 malignant and 37 benign lesions in 229 patients. 7 patients presented with bilateral cancers. The bradiomics feature
tensor model reduced false-positive results from 57 to 29 (specificity 88.9% [95% IC 0.843-0.923]) and diffusion imaging alone was
less sensitive 89.9% (95% CI 0.855-0.931) than the conventional reading of DCE that provided sensitivity of 95.1% (95% CI 0.916-
0.973) and specificity of 78.2% (95% CI 0.727-0.83) at the threshold including in situ disease. Diagnostic accuracy was 89.41%
(95% CI 0.8941-0.9190) for tensor modeling and 86.77% (95% CI 0.8358-0.8954) for DCE.
CONCLUSION
The bradiomics model based on diffusion tensor allowed for similar diagnostic accuracy of obtained using clinical set reading DCE.
This may translate to less recalls and improve clinical outcomes.
CLINICAL RELEVANCE/APPLICATION

The use of MR techniques that lead to high diagnostic accuracy without IV contrast may play a role in the clinical set.
RC315-16 Computer-Aided Diagnosis (CAD)-assessed Kinetic Features of Invasive Breast Cancers: Correlation
with Clinical-pathologic Prognostic Factors
Tuesday, Nov. 27 11:40AM - 11:50AM Room: Arie Crow n Theater

Participants
Sung Eun Song, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Kyu Ran Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Bo Kyoung Seo, MD, PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation;
Research Grant, Guerbet SA; Research Grant, Koninklijke Philips NV;
Ok Hee Woo, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Deuk Jae Sung, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
akaeuny@hanmail.net
PURPOSE

To investigate the association of kinetic features with clinical-pathologic factors in breast cancer patients using computer-aided
diagnosis (CAD).
METHOD AND MATERIALS
Between July 2016 and March 2017, 85 patients with invasive breast cancers (mean, 1.8cm; range, 0.8-4.8cm) who had undergone
preoperative 3.0T MR imaging and surgery were retrospectively enrolled. All MR image were processed using CAD, and kinetic
features of tumors were acquired: peak enhancement, angio-volume, early and delayed enhancement profiles. The relationships
between kinetic features and clinical-pathologic factors were assessed. Mann-Whitney test, Spearman's correlation test and binary
logistic regression analysis were used for statistical analysis.
RESULTS
In correlation tests, CAD-assessed peak enhancement and angio-volume were significantly correlated with histologic grade, Ki-67
index, and tumor size: r = 0.355 (P = .001), r = 0.330 (P = .002), and r = 0.231 (P = .033) for peak enhancement, r = 0.410 (P =
.005), r = 0.341 (P < .001), and r= 0.505 (P < .001) for angio-volume. Plateau compoment at delayed phase was significanly
correlated with Ki-67 index (r = 0.255 [P = .019]), but correlated coefficient between rapid component at early phase and Ki-67
index did not reach statistical significance (r = 0.202 [P = .063]). In binary logistic regression analysis, higher peak enhancement
was a significant independent predictor of higher histologic grade (odds radio [OR] = 1.004; 95% CI: 1.001,1.008; P = .024), larger
angio-colume was a predictor of larger tumor size (OR = 1.384; 95%CI: 1.141, 1.679; P = .001), higher plateau component was a
predictor of negative estrogen receptor status (OR = 0.928; 95%CI: 0.877, 0.982; P = .010), and both higher plateau component
and angio-volume were predictors of higher Ki-67 index (OR = 1.051; 95%CI: 1.011, 1094; P = .013 for plateau component. OR =
1.178; 95%CI:1.023;1.356; P = .023 for angio-volume).
CONCLUSION
Of the CAD-assessed preoperative breast MRI kinetic features, higher peak enhancement may predict higher histologic grade, larger
angio-volume may predict larger tumor size, higher plateau component may predict negative estrogen receptor status, and both
higher plateau component and angio-volume may predict higher Ki-67 index.
CLINICAL RELEVANCE/APPLICATION
CAD-assessed preoperative breast MRI kinetic features can be considered as a useful imaging biomarker reflecting clinical-
pathologic prognostic factors.
RC315-17 Correlation of MRI Texture Features With Tumor Infiltrating Lymphocytes and Pathologic Complete
Response in HER2 Positive and Triple Negative Subtypes of Breast Cancer
Tuesday, Nov. 27 11:50AM - 12:00PM Room: Arie Crow n Theater

Participants
Gaiane M. Rauch, MD, PhD, Houston, TX (Presenter) Nothing to Disclose
Hongtu Zhu, PhD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Heng Li, Houston, TX (Abstract Co-Author) Research funded, Varian Medical Systems, Inc
Beatriz E. Adrada, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Lumarie Santiago, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Wei T. Yang, MD, Houston, TX (Abstract Co-Author) Consultant, General Electric Company; Medical Advisory Board, Seno Medical
Instruments, Inc
Rosalind P. Candelaria, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Jessica W. Leung, MD, Houston, TX (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc;
Speakers Bureau, FUJIFILM Holdings Corporation
Mohamed Elbanan, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Alper H. Duran, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Ebru Unlu, Houston, TX (Abstract Co-Author) Nothing to Disclose
Minhua Wang, MD,PhD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Stacy Moulder, MD, Houston, TX (Abstract Co-Author) Research funded, AstraZeneca PLC; Research funded, F. Hoffmann-La Roche
Ltd; Research funded, Oncothyreon; Research funded, Novartis AG; Research funded, Merck KGaA
Yun Wu, Houston, TX (Abstract Co-Author) Nothing to Disclose
Elizabeth Mittendorf, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Alastair Thompson, Houston, TX (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
gmrauch@mdanderson.org
PURPOSE
To evaluate associations of quantitative MRI texture features and tumor infiltrating lymphocytes (TIL) levels in HER2+ and triple
negative (TN) subtypes of breast cancer (BC) receiving neoadjuvant chemotherapy (NAC), as potential prognostic non-invasive
imaging markers for pathologic complete response prediction (pCR).
METHOD AND MATERIALS
Retrospective review of BC patients who had MRI at staging, neoadjuvant chemotherapy and surgery from January 1, 2008 through
December 31, 2015 was performed. Demographic, imaging, and pathologic data including TIL levels were documented. Quantitative
MRI texture analysis was performed using 3 types of textural features (TF): local binary patterns (LBP), gray-level co-occurrence
matrix (GLCM), and threshold adjacency statistics (TAS). Associations between MRI quantitative TF, TIL levels, and pCR were
evaluated by Pearson correlation and logistic regression.
RESULTS
There were 50 HER2+ and 38 TN patients (median age 51 years, range 29-59) with pretreatment MRI and TIL status for analysis;
27 HER2+ patients and 15 TN patients had pCR at surgery. For HER 2+ patients 9 TF significantly correlated with pCR (p<0.05): f1
(angular 2nd moment), l3 (75 percentile), l4 (standard deviation), t1-t6 (adjacency 0-5). Four TF were significantly associated with
high TIL levels (p<0.05): texture l4 (standard deviation), t2 and t3 (adjacency 1 and 2). Additional 4 TF had weak association with
TIL (p<0.1): feature f8 (sum entropy), t1, t3 and t4 (adjacency 0, 3 and 4). Three TF were significantly associated with both, pCR
and TIL (p<0.05): texture l3 (75 percentile), l4 (standard deviation), t9 (adjacency 8). For TN patients 4 TF f2 (contrast), t1,t3
and t4 (adjacency 0 ,2,3) were significantly associated with pCR (p < 0.005). No TF were significantly associated with TIL levels
for TNBC, only t3 and t4 (adjacency 4 and 5) showed weak association with TIL levels (p<0.1).
CONCLUSION
Quantitative tumor MRI texture analysis in HER2+ BC showed 9 TF associated with pCR, 8 TF with TIL and 3 TF with both pCR and
TIL; for TNBC 4 TF were associated with pCR, and 2 TF weakly associated with TIL.
CLINICAL RELEVANCE/APPLICATION

Analysis of associations of MRI quantitative TF with pCR and TIL in HER2+ and TNBC may help to develop prognostic non-invasive
imaging markers for treatment response prediction.
 RC320

Role of MR Imaging in Cancer Staging and Treatment

Tuesday, Nov. 27 8:30AM - 10:00AM Room: S403B

BR GI MR RO
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Kathryn J. Fowler, MD, San Diego, CA (Moderator) Nothing to Disclose
LEARNING OBJECTIVES

1) Overview of MRI use in cancer staging, treatment delivery, and response assessment. -Understand the imaging approach to
staging rectal cancer -Review pertinent anatomy -Discuss reporting of stage -Understand response indicators and how to report
Sub-Events
RC320A Role of MR Imaging in GI Cancer Staging
Participants
Kathryn J. Fowler, MD, San Diego, CA (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Review basic MRI approach to staging gastrointestinal malignancies (pancreatic and rectal). 2) Understand the application of
MR/imaging features for assessing response to therapy.
RC320B MR-guided Radiotherapy for GI Cancers
Participants
Michael F. Bassetti, MD, Madison, WI (Presenter) Research Grant, Merck KGaA; Research Grant, AstraZeneca PLC;
LEARNING OBJECTIVES
1) Identify the clinical sites where MR-guided radiation may have the highest impact. 2) Understand the unique sources of
uncertainty of MR-guided radiation that differ from conventional LINAC radiation. 3) Identify the most common indications for on-
line MR -guided adaptive radiotherapy observed in clinical practice.
RC320C Role of MR Imaging in Breast Cancer Staging
Participants
Bethany L. Niell, MD,PhD, Tampa, FL (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Identify evidence-based indications for MR imaging in breast cancer staging. 2) Describe the frequencies of ipsilateral multifocal
or multicentric disease and contralateral breast cancer detected on breast MRI. 3) Explain potential pitfalls and limitations of breast
MRI performed for staging.
RC320D MR-guided Radiotherapy for Breast Cancer
Participants
Maria A. Thomas, MD, PhD, Saint Louis, MO (Presenter) Nothing to Disclose
Michael F. Bassetti, MD, Madison, WI (Presenter) Research Grant, Merck KGaA; Research Grant, AstraZeneca PLC;
LEARNING OBJECTIVES
1) Identify potential applications of MR-guided radiotherapy for breast cancer. 2) Describe the advantages and disadvantages of
MR-guided radiotherapy for breast cancer.
 VW 37

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Reading Sessions: Siemens
Healthineers Vendor Workshop
Tuesday, Nov. 27 10:15AM - 11:25AM Room: Booth 5530

Participants
Luis Pina, MD, PhD, San Sebastian, Spain (Presenter) Nothing to Disclose
Program Information
During this hands-on workshop you will learn to evaluate 2D mammography and 3D Breast Tomosynthesis. An expert tutor will lead
you through cases that will both fascinate and challenge you! All cases have been acquired with Siemens Mammomat Inspiration
and are displayed on our syngo® Breast Care workstations so you will become familiar with the quality of our HD Tomo images and
ease of use of our systems.
 SSG01

Breast Imaging (Ultrasound Screening and Diagnostic Indications)

Tuesday, Nov. 27 10:30AM - 12:00PM Room: S406A

BR US
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Donna M. Plecha, MD, Strongsville, OH (Moderator) Research Grant, Hologic, Inc
Jessica W. Leung, MD, Houston, TX (Moderator) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc; Speakers
Bureau, FUJIFILM Holdings Corporation
Sub-Events
SSG01-01 Performance of Screening Breast Ultrasound Over 5-Year Period
Tuesday, Nov. 27 10:30AM - 10:40AM Room: S406A

Participants
Stamatia V. Destounis, MD, Scottsville, NY (Presenter) Research Grant, Hologic, Inc; Research Grant, Delphinus Medical
Technologies, Inc
Amanda Santacroce, Rochester, NY (Abstract Co-Author) Nothing to Disclose
Andrea L. Arieno, BS, Rochester, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
sdestounis@ewbc.com
PURPOSE
To review and compare performance of screening ultrasound in women with dense breast tissue post adoption of state breast
density mandate.
METHOD AND MATERIALS
Through a retrospective chart review, data was collected on 23,878 screening ultrasound exams performed from 2013-2017 in
patients with heterogeneously dense or extremely dense breast tissue. We stratified the data by year to compare screening
ultrasound performance over time. Cancer detection rate (CDR), PPV and biopsy rate were calculated for US only findings
(mammographically normal). Data was collected on patient demographics, number of DBT exams performed, BI-RADS score,
pathology, tumor size, and lymph node status.
RESULTS
A total of 23,878 screening ultrasound exams were performed; 691 (2.9%) in 2013, 1700 (7.1%) in 2014, 4767 (20.0%) in 2015,
7389 (30.9%) in 2016, 9339 (39.1%) in 2017. Use of BI-RADS 1 and 2 remained stable; ranging from 95.0% in 2013 to 97.1% in
2017. The use of DBT increased in the population increased, from 18.7% in 2013 to 99.3% in 2017. Cancer detection rate increased
in the first 3 years, from 1.4/1000 (2013) to 3.6/1000 (2015) then decreased in 2016 to 2.4/1000, and again in 2017 to 2.2/1000.
Biopsy rate steadily decreased since 2013, from 2.9% to 1.0% in 2017. PPV for biopsy initially increased substantially (5% in 2013
to 20% in 2014) and then slightly declined; 18.8% in 2015, 16.2% in 2016, and increased again to 22.3% in 2017. The assignment
of BIRADS 3 decreased over time; from 2.5% of exams in 2013, to .76% in 2017.
CONCLUSION
Screening ultrasound continues to detect mammographically occult malignancy, though the rate can vary substantially, as we saw
an initial increase, followed by a decrease to 2.2/1000 in 2017. This decrease could be due to the increase in use of DBT, which
increases the visibility of malignancies on mammography. The biopsy rate consistently decreased, from 2.9% to 1.0%, as did
assignment of BI-RADS category 3. PPV varied over time, with the highest rate over the period in 2017 (22.3%), potentially
suggesting more appropriate use of biopsy in this population.
CLINICAL RELEVANCE/APPLICATION
There are 30 states with breast density legislation. As policies continue to change and center around individualized medicine,
understanding the performance and value of US will be helpful for facilities as they continue to adopt improved screening practices
and modalities for women with dense breast tissue.
SSG01-02 Non-Mass Lesion Detected by Breast US: Stratification of Cancer Risk for Clinical Management
Tuesday, Nov. 27 10:40AM - 10:50AM Room: S406A

Participants
Ko Woon Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ji Soo Choi, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Surin Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Mi-Ri Kwon, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Boo-Kyung Han, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Ko, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Sook Ko, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
So Yeon Yang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
E-Ryung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kowoon.park@samsung.com
PURPOSE
To develop a practical diagnostic prediction model by using imaging features of breast non-mass lesion (NML) detected by B-mode
ultrasonography (US).
METHOD AND MATERIALS
From Dec 2011 to March 2016, B-mode US was performed in 777 NMLs in consecutive 828 patients. This retrospective study
consecutively included 669 breast NMLs in 669 patients for which final diagnoses were established and mammographies obtained at
the time of US examination. For each lesion, radiologists assessed BI-RADS category. For developing a diagnostic prediction model
to estimate malignant risk of NMLs using a scoring, univariate and multivariate logistic regression analyses were performed to find
US or mammographic features associated with malignancy in a developmental dataset (n=460). A score for each significant imaging
feature was assigned and multiplied by the regression coefficient, and the risk score of malignancy for each NML was defined as the
sum of individual scores. Based on area under the receiver operating characteristic curve (AUC), diagnostic performance of the
prediction model was compared to the radiologist's BI-RADS classification. Later, we verified the usefulness of developed scoring
model by applying into the remaining validation dataset (n=209).
RESULTS

Among 669 NMLs, 354 (52.9%) were benign and 315 (47.1%) were malignant. In the developmental dataset, the following US
features, within or around the main lesion, showed significant association with malignancy: the presence of calcifications,
architectural distortion, posterior acoustic shadowing or abnormal ductal change, and absence of microcysts. The following
mammographic features also showed significant association with malignancy: the presence of calcifications or focal asymmetry. The
predictive model's AUC was higher than the radiologist's BI-RADS classification (0.952 vs. 0.930). In the validation dataset, AUC of
our prediction model was 0.961.
CONCLUSION

The prediction model using features of US and mammography may be useful in stratification of cancer risk of breast non-mass
lesions.
CLINICAL RELEVANCE/APPLICATION

Cancer risk stratification for breast non-mass lesions using features of US and mammography may be useful in managing breast
non-mass lesions detected by US.
SSG01-03 Decreasing Short-Term Follow-Up and Biopsies by Following BI-RADS 3 Lesions at 1 Year: A
Prospective Study
Tuesday, Nov. 27 10:50AM - 11:00AM Room: S406A

Participants
Richard G. Barr, MD, PhD, Campbell, OH (Presenter) Consultant, Siemens AG; Consultant, Koninklijke Philips NV; Research Grant,
Siemens AG; Research Grant, SuperSonic Imagine; Speakers Bureau, Koninklijke Philips NV; Research Grant, Bracco Group; Speakers
Bureau, Siemens AG; Consultant, Canon Medical Systems Corporation; Research Grant, Esaote SpA; Research Grant, BK Ultrasound;
Research Grant, Hitachi, Ltd
Carmine Tinelli, MD, MSC, Pavia, Italy (Abstract Co-Author) Nothing to Disclose
Annalisa DeSilvestri, PhD, Pavia, Italy (Abstract Co-Author) Nothing to Disclose
PURPOSE
Supplemental US is know to detect node negative cancers not identified on screening mammography. However the large number of
short-term follow-ups and low positive biopsy rate make this technique not cost effective. The majority of these are due to BI-
RADS 3 lesions (B3), with an incidence of cancer in < 1%. This prospective study evaluates the effect of following B3 detected on
supplemental ultrasound at 1 year.
METHOD AND MATERIALS

This HIPPA compliant, IRB approved with written informed consent study invited patients receiving a B1 or B2 screening
mammogram with density 3 or 4 breast of any risk to receive a free automated volume whole breast (ABVS) ultrasound. The ABVS
was performed on a Siemens S2000 using a 15cm L14-5 transducer. ABVS were read by radiologist with 20 year experience as BI-
RADS category 1, 2, 3, or 0. Category 0 patients received a hand held ultrasound (HH). Patients were followed for 2 years.
RESULTS
Of 23426 screening patients, 8542(36.5%) had density 3 or 4 and were asked to participate. 2257 (26.4%) agreed to participate
(50 yo mean age, range 31 to 90) (<10% high risk). The ABVS was interpreted as B1 in 1186 (52.5%), B2 in 591 (26.2%), B3 in 395
(17.5%) and B0 in 85 (3.8%). Of the 395 B3 patients, 310 had 1-year follow-up and were cancer free, 0%, 254 had 2-year follow-
up and were cancer free, 0%. Of the 85 B0 patients, (recall rate of 3.8%) on HH 8 (11.3%) were B1, 51 (71.8%) were B2, 2 (2.8%)
were B3, 6 (8.5%) were B4, and 4(5.6%) were B5. The B4 and 5 lesions were 2 B4A lesions were fibroadenomas,2 B4A were
fibrocystic change, 2 category 4C lesions were IDC, and 4 category 5 lesions were IDC. The biopsy rate was 0.4% (10/2257) with a
positive biopsy rate (PPV3) of 60.0% (6/10). The supplemental ultrasound detected 2.8/1000 additional cancers (4/1412). 5
cancers (0.28%, 5/1777) were detected in 3 yr f/u in B1 and B2 and 1 IDC (0.25%, 1/395) was identified in B3 in the contralateral
breast.
CONCLUSION
Following B3 at 1 year interval decreases the recall rate (8.8% (233/2637) ACRIN 6666 to 3.8% (54/1412)) (p<0.001) and increases
the PPV3 (8.9% (21/235)ACRIN 6666 to 60.0% (6/10))(p=0.001) without substantial node positive cancer misses. With these
improved screening characteristic, supplemental ultrasound could be cost effective.
CLINICAL RELEVANCE/APPLICATION

Following B3 at 1 year markedly reduces the recall rate and increases the PPV3 making supplemental screening more cost effective.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Richard G. Barr, MD, PhD - 2017 Honored Educator
SSG01-04 A Preliminary Study of Predicting Molecular Subtypes of Breast Cancer by the Radiomics Features of
Contrast-Enhanced Ultrasound
Tuesday, Nov. 27 11:00AM - 11:10AM Room: S406A

Participants
Lei Tang, Shanghai, China (Presenter) Nothing to Disclose
Man Chen, Shanghai, China (Abstract Co-Author) Nothing to Disclose
PURPOSE
The aim was to analyze the quantitative features of contrast-enhanced ultrasound(CEUS) imaging of breast cancer through
computer image extraction, and to explore the preoperative prognosis of the different molecular subtypes of breast cancer.
METHOD AND MATERIALS
CEUS images of 189 patients were collected from our hospital. Everyone signed informed consent before CEUS. Surgical pathology
and molecular typing results were obtained in all patients. Through mapping the borders of breast cancers on the images, the entire
internal area of the tumor was determined. The lesions were divided into four quadrants(Fig 1). The high perfusion within the lesion
area was defined as a high brightness area. The temporal and special features of the images in different areas were extracted by
the computer in a dynamic CEUS file and the time intensity curves(TIC) were drawn. From the TIC curve, the features were
extracted such as wash in rate(WiR), wash out rate(WoR), rise time(RT), base intensity(BI) and peak intensity(PI) and EI(=PI-BI),
so on. The parameters of the curve in each area were calculated in different molecular typing groups.
RESULTS
The patient' numbers of Luminal A, Luminal B, HER2+, and triple negative breast cancer (TNBC) of each molecular subtypes were
46, 75, 37, and 31, respectively. The extracted effective features included EI of the internal bright area, WoR of internal bright
area, WoR of the internal overall area, RT of the internal overall area, and so on. The cutoff value 1.566 of WoR in the internal
bright area might help to find Luminal A, with a specificity of 82.61%. When to find TNBC, the cutoff value of EI at the internal
bright area was 0.3494, the sensitivity was 64.52%, and the specificity was 68.99%. When distinguishing between Her2+ and
TNBC, WOR of the internal overall had a cutoff of 5.7496 and a sensitivity of 74.19%.
CONCLUSION

The radiomics features of contrast-enhanced ultrasound could contribute to preoperative prediction of breast cancer molecular
subtypes. Further research needed to be larger sample, multi-center expansion.
CLINICAL RELEVANCE/APPLICATION

The radiomics features of contrast-enhanced ultrasound could contribute to preoperative prediction of breast cancer molecular
subtypes, which may help to predict efficacy and to select treatment options.
SSG01-05 Sonographic Features of Radial Scars and Complex Sclerosing Lesions
Tuesday, Nov. 27 11:10AM - 11:20AM Room: S406A

Participants
Reni S. Butler, MD, Madison, CT (Presenter) Nothing to Disclose
Liva Andrejeva-Wright, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Jaime L. Geisel, MD, New Haven, CT (Abstract Co-Author) Consultant, QView Medical, Inc
Laura S. Sheiman, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Madhavi Raghu, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Regina J. Hooley, MD, New Haven, CT (Abstract Co-Author) Consultant, Hologic, Inc
Liane E. Philpotts, MD, New Haven, CT (Abstract Co-Author) Consultant, Hologic, Inc
For information about this presentation, contact:

reni.butler@yale.edu
PURPOSE

To assess the morphologic characteristics of radial scars (RS) and complex sclerosing lesions (CSL) when they are visualized with
ultrasound (US)
METHOD AND MATERIALS

A HIPAA-compliant, IRB-approved retrospective review of core needle biopsies (CNB) performed between 1/1/2007 and 12/31/2017
was performed and filtered for RS or CSL as the primary diagnosis. Patients with a concurrent diagnosis of malignancy or with only a
microscopic incidental RS were excluded. The method of detection, mammographic and sonographic features, histology at CNB and
at surgical excision, if performed, were recorded for all lesions visualized with US.
RESULTS

190 lesions with a CNB diagnosis of RS or CSL were identified. 57.4% (109/190) were visible on US. Most [75.2% (82/109)] US-
visible lesions were initially detected on screening mammography, followed by screening US [19.3% (21/109)], diagnostic US [2.8%
(3/109)], diagnostic mammography [1.8% (2/109)], and screening MRI [0.9% (1/109)]. Among US-visible lesions, 53.2% (58/109)
appeared as non-mass areas of abnormal echogenicity, 44.0% (48/109) as masses, 1.8% (2/109) as architectural distortion only,
and 0.9% (1/109) as dilated hypoechoic ducts. More lesions were anti-parallel [58.7% (64/109)] compared to parallel [41.3%
(45/109)]. While most were hypoechoic [64.2% (70/109)], others were isoechoic [14.7% (16/109)], hyperechoic [3.7% (4/109)],
or mixed [17.4% (19/109)]. Posterior acoustic features were most often none [58.7% (64/109)], followed by shadowing [28.4%
(31/109)] and enhancement [12.8% (14/109)]. Color Doppler images, available in 87 lesions, showed no vascular flow in 34.5%
(30/87), adjacent flow in 34.5% (30/87), and internal flow in 31.0% (27/87). Lesions presenting as architectural distortion on
mammography [50.4% (55/109)] most often appeared as non-mass areas of variable echogenicity on US [89.1% (49/55)] rather
than masses [10.9% (6/55)].
CONCLUSION

RS and CSL sonographic features are variable and include non-mass areas of abnormal echogenicity. While not strictly part of the
BI-RADS lexicon, subtle 'nonmass' findings may be the only US correlate in some mammographically-detected lesions, especially
those presenting as architectural distortion.
CLINICAL RELEVANCE/APPLICATION

As the incidence of RS and CSL increases with tomosynthesis utilization, understanding the variety of US appearances may increase
the likelihood of detecting an US correlate and facilitate biopsy.
SSG01-06 Shear-Wave Elastography of the Breast: Value of a Novel 5-Point Technical Quality Score
Tuesday, Nov. 27 11:20AM - 11:30AM Room: S406A

Participants
Masoud Baikpour, MD, Boston, MA (Presenter) Nothing to Disclose
Shinnhuey S. Chou, MD, Boston, MA (Abstract Co-Author) Researcher, Investigator, General Electric Company
Wei Zhang, MD, Allston, MA (Abstract Co-Author) Nothing to Disclose
Sarah Mercaldo, PhD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Anthony E. Samir, MD, Boston, MA (Abstract Co-Author) Consultant, Pfizer Inc; Consultant, General Electric Company; Consultant,
PAREXEL International Corporation; Research Grant, Koninklijke Philips NV; Research Grant, Siemens AG; Research Grant, Canon
Medical Systems Corporation; Research Grant, General Electric Company; Research Grant, Samsung Electronics Co, Ltd; Research
Grant, Analogic Corporation; Research support, SuperSonic Imagine; Research support, Hitachi, Ltd
Constance D. Lehman, MD,PhD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Medical Advisory
Board, General Electric Company
For information about this presentation, contact:

mbaikpour@mgh.harvard.edu
PURPOSE
To determine the value of a novel 5-point technical quality score on the diagnostic performance of shear-wave elastography (SWE)
of the breast.
METHOD AND MATERIALS
This IRB-approved HIPAA-compliant study included 110 consecutive women (mean age 55.1 + 15.3 years) with 122 breast lesions
who underwent SWE and ultrasound-guided biopsy from Oct. 2017 to Jan. 2018. We recorded the maximum (Emax), mean (Emean),
and standard deviation (Esd) elasticity measurements for each lesion. We defined five specific SWE technical quality parameters:
(1) lesion visibility on B-mode image panel, (2) red pattern (high stiffness) in the near field of the field-of-view (FOV), (3) size and
location of FOV box relative to lesion, (4) heterogeneity, vertical streaks, and absence of color in tissue surrounding the lesion, (5)
size and location of the region-of-interest circle on the lesion for elasticity measurements. Three blinded readers independently
assessed each SWE parameter as low (score=0) or high (score=1) quality. SWE total quality score < 3.3 was classified as low and
> 3.3 as high. Intra-class correlation coefficient (ICC) was used to measure inter-observer agreement for the total score and the
Hanley and McNeil's method was used to compare areas under the receiver operating characteristic curve (AUC) of SWE in low vs.
high-quality images.
RESULTS

Mean size of the 122 lesions was 13.9 + 10.4 mm; 64 (52%) were benign and 58 (48%) were malignant. Inter-observer agreement
was good among readers (ICC 0.805). AUCs were significantly improved in the high-quality group compared to the low-quality group
for Emean (0.858 vs. 0.631, p=0.009; AUC difference=0.227, 95%CI [0.056, 0.398]) and Esd (0.816 vs. 0.629, p=0.040; AUC
difference=0.187, 95%CI [0.008, 0.366]). AUC for Emax also increased to 0.861 in the high-quality group compared to 0.714 in the
low-quality group but without statistical significance (p=0.077; AUC difference=0.147, 95%CI [-0.015, 0.309]).
CONCLUSION

Incorporating our novel 5-point technical quality score can improve the diagnostic performance of SWE in differentiating malignant
from benign breast lesions.
CLINICAL RELEVANCE/APPLICATION

A simple 5-point technical quality metric provides guidance to users during real-time acquisition and improves diagnostic
performance of SWE in differentiating malignant from benign breast lesions.
SSG01-08 Can Mid-Treatment Ultrasound in Triple Negative Breast Cancer Patients Predict Residual Pathologic
Disease in the Axillary Nodes?
Tuesday, Nov. 27 11:40AM - 11:50AM Room: S406A

Participants
Rosalind P. Candelaria, MD, Houston, TX (Presenter) Nothing to Disclose
Beatriz E. Adrada, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Lumarie Santiago, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Deanna L. Lane, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Wei Wei, Houston, TX (Abstract Co-Author) Nothing to Disclose
Wei T. Yang, MD, Houston, TX (Abstract Co-Author) Consultant, General Electric Company; Medical Advisory Board, Seno Medical
Instruments, Inc
Monica L. Huang, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Elsa M. Arribas, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Gaiane M. Rauch, MD, PhD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Alastair Thompson, Houston, TX (Abstract Co-Author) Nothing to Disclose
Stacy Moulder, MD, Houston, TX (Abstract Co-Author) Research funded, AstraZeneca PLC; Research funded, F. Hoffmann-La Roche
Ltd; Research funded, Oncothyreon; Research funded, Novartis AG; Research funded, Merck KGaA
For information about this presentation, contact:
rcandelaria@mdanderson.org
PURPOSE

To determine if the number of abnormal lymph nodes visualized on mid-treatment ultrasound in triple negative breast cancer (TNBC)
patients who are undergoing neoadjuvant chemotherapy associates with residual nodal disease on surgical pathology
METHOD AND MATERIALS

As part of an on-going single-institution, clinical trial of stage I-III TNBC patients, the first 106 patients who underwent surgery
were included in this interim analysis. Mid-treatment was defined as the period following completion of four cycles of AC (Adriamycin
and cyclophosphamide) chemotherapy and before initiating either anticipated Taxol chemotherapy or an investigational therapy.
The number of abnormal nodes at mid-treatment was assessed and recorded by experienced, fellowship-trained breast radiologists.
These radiologists empirically categorized lymph nodes using a binary approach of sonographically abnormal versus normal.
Pathologic lymph node positive was defined as the presence of macrometastasis and micrometastasis in at least one axillary node
from sentinel lymph node biopsy and/or axillary lymph node dissection as stated in the surgical pathology reports. Wilcoxon rank
sum test and Fisher's exact test were used to determine statistical significance.
RESULTS

There were 26 of 106 patients (25%) who had residual nodal disease at surgery and 80 of 106 patients (75%) who had nodal
pathologic complete response. The median number of abnormal nodes at mid-treatment was 3 (range 0-16 nodes) for patients who
had residual nodal disease compared to 0 (range 0-12 nodes) for patients who had nodal pathologic complete response. TNBC
patients with residual nodal disease on surgical pathology had significantly more abnormal nodes at mid-treatment (p<0.0001). More
specifically, TNBC patients with at least 2 abnormal lymph nodes at mid-treatment ultrasound had a significantly higher chance of
being pathologic lymph node positive at surgery (p<0.0001).
CONCLUSION

There is a highly significant association between the number of abnormal lymph nodes identified at mid-treatment ultrasound and
the presence of residual metastatic axillary nodes at surgery in triple negative breast cancer patients.
CLINICAL RELEVANCE/APPLICATION

The number of abnormal lymph nodes at mid-treatment may be an independent predictor of residual disease in TNBC patients and
may assist in identifying chemoresistant patients who could benefit from investigational therapies.
SSG01-09 Is a BI-RADS 4 or 5 Assessment Reasonable on Screening Ultrasound?
Tuesday, Nov. 27 11:50AM - 12:00PM Room: S406A

Participants
Wendie A. Berg, MD, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
Golbahar Houshmand, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
David Gur, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Terri A. Gizienski, MD, Greenwood Village, CO (Abstract Co-Author) Nothing to Disclose
Denise M. Chough, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Marcela Bohm-Velez, MD, Pittsburgh, PA (Abstract Co-Author) Consultant, Koninklijke Philips NV; Research Grant, Delphinus Medical
Technologies, Inc
Christiane M. Hakim, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Marie A. Ganott, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Thomas S. Chang, MD, Pittsburgh, PA (Abstract Co-Author) Research Grant, Seno Medical Instruments, Inc; Research Grant,
Delphinus Medical Technologies, Inc; Consultant, iCAD, Inc
Michelle R. Straka, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Danielle Sharek, MD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Jamie Hartman, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
Amy E. Kelly, MD, Bridgeville, PA (Abstract Co-Author) Nothing to Disclose
Cathy S. Tyma, MD, Sewickley, PA (Abstract Co-Author) Nothing to Disclose
Kimberly S. Harnist, MD, Gibsonia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

bergwa@upmc.edu
bergwa@upmc.edu
PURPOSE
For participants recalled for further testing, to determine the utility of a BI-RADS final assessment in determining management
based on technologist-performed screening ultrasound.
METHOD AND MATERIALS
5689 women in three centers (1 academic and 2 private practice) were enrolled in an IRB-approved study of screening
tomosynthesis (DBT) and technologist-performed handheld screening ultrasound (US, with orthogonal views of each finding other
than simple cysts) and underwent 8151 screens (5689 yr1, 2462 yr2). Two experienced breast-imaging specialized radiologists
prospectively reviewed each DBT-US set independently and in opposing order. Whenever additional imaging was recommended prior
to the next annual screening, readers recorded a "final" assessment: BI-RADS 3, 4A, 4B, 4C or 5, together with recommendations,
which could include immediate additional imaging and/or possible biopsy, or six-month follow-up. When US was read first and still
resulted in a recall after integration with the DBT, we compared subsequent management and outcomes with the original BI-RADS
assessments. We excluded technical recalls. At least targeted prior sonograms were available for 980 participants in year 1.
RESULTS
24 women were ultimately diagnosed with cancer after US recall, median patient age 55 (range 40-75). A total of 455 (5.6%) US
exams prompted recall: 348/5689 (6.1%) in year 1 and 107/2462 (4.3%) in year 2. Of 209 women scored BI-RADS 3, 114 had
immediate additional evaluation and 95 had 6-month follow-up; 20 (9.6%) ultimately had biopsy with 2 (1.0%) found to have
cancer. Of 153 rated BI-RADS 4A on screening, 86 (56.2%) had biopsy and 2 (1.3%) proved malignant. Of 60 BI-RADS 4B, 40
(66.7%) had biopsy and 4 (6.7%) had cancer. Of 11 BI-RADS 4C, 11 (100%) had biopsy and 7 (63.6%) had cancer. Of 9 BI-RADS
5, 9 (100%) had biopsy and 9 (100%) were found to have cancer.
CONCLUSION
Malignancy and biopsy rates were lower than expected for a BI-RADS 4A or 4B assessment on technologist-performed screening
US, but were reasonable for a BI-RADS 4C or 5 assessment.
CLINICAL RELEVANCE/APPLICATION

Directly scheduling biopsy based on screening US was appropriate for BI-RADS 4C and 5 assessments; 56% of 4A and 67% of 4B
assessments resulted in biopsy after physician-targeted evaluation.
 VW 35

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Self-guided Reading
Sessions: Siemens Healthineers Vendor Workshop
Tuesday, Nov. 27 10:30AM - 5:00PM Room: Booth 5530

Program Information
You are invited to our self-guided reading sessions with syngo® Breast Care workstations configured especially to allow you to
work at your own place at a time that suits you! A series of breast tomosynthesis and contrast enhanced mammography cases
presented as problem cases with a solution enables you to develop and test your reading skills.
 VW 36

Automated Breast Volume Scanner (ABVS) Self-guided Reading Sessions: Siemens Healthineers Vendor
Workshop
Tuesday, Nov. 27 10:30AM - 5:00PM Room: Booth 5530

Program Information
With syngo® Ultrasound Breast Analysis (sUSBA) software, self-guided reading sessions with real clinical cases will enable you to
become familiar with the coronal plane while providing practical approaches to interpretation of 3D automated breast ultrasound.
 VW 62

Tomosynthesis Guided Prone Breast Biopsy Solutions in a Community Based Practice: Hologic Vendor
Workshop
Tuesday, Nov. 27 10:30AM - 11:30AM Room: Booth 5524

Participants
Harriet B. Borofsky, MD, San Mateo, CA (Presenter)
Program Information
Clinical benefits of tomosynthesis guided biopsy which includes a hands-on demonstration of the Affirm® Prone Biopsy System and
the Brevera® Breast Biopsy System (Affirm® Prone Biopsy System, Brevera® Breast Biopsy System and Viera™ Portable Breast
Ultrasound)
Registration
https://hologicrsna.com
 VW 38

A Practical Approach to Breast Magnetic Resonance Imaging (MRI) Interpretation: An Interactive Session:
Siemens Healthineers Vendor Workshop
Tuesday, Nov. 27 11:40AM - 12:50PM Room: Booth 5530

Participants
Susan Weinstein, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Program Information
This interactive session will include both didactic and hands-on case review at workstations equipped with syngo® MR Brevis. A
practical approach to breast MRI interpretation will be discussed as well as utilizing the available sequences and techniques to
improve interpretive skills.
 VW 63

Triaging Dense Breast Patients in Clinical Practice: Hologic Vendor Workshop

Tuesday, Nov. 27 12:00PM - 12:30PM Room: Booth 5524

Participants
Regina J. Hooley, MD, New Haven, CT (Presenter) Consultant, Hologic, Inc
Program Information
The Genius™ 3D Mammography™ exam is the only mammogram that is FDA approved as superior for women with dense breasts.
Attend this 30 minute session to learn how to triage women with dense breasts using 3D Mammography. (3D Mammography™
Technology)
Registration
https://hologicrsna.com
 BRS-TUA

Breast Tuesday Poster Discussions

Tuesday, Nov. 27 12:15PM - 12:45PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

FDA Discussions may include off-label uses.

Participants
Samantha L. Heller, MD, PhD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
BR241-SD- Evaluating the Timeliness of Abnormal Mammography Follow-up Based on Race at an Urban Safety
TUA1 Net Hospitalty
Station #1
Awards
Student Travel Stipend Award
Participants
Neeta Kannan, MD, San Francisco, CA (Presenter) Nothing to Disclose
Amie Y. Lee, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Kimberly M. Ray, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Bonnie N. Joe, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jessica H. Hayward, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
neeta.kannan@ucsf.edu
PURPOSE
To investigate racial and ethnic disparities in timeliness of abnormal mammography follow-up at an urban safety net hospital.
METHOD AND MATERIALS
All women who underwent screening or diagnostic mammography at our facility between 4/1/2016-4/1/2017 whose examinations
received actionable Breast Imaging Reporting and Data System (BI-RADS) assessments 0, 4 or 5 were identified. Self-reported
racial or ethnic group was recorded. Median follow-up times from abnormal screening to diagnostic mammogram and abnormal
diagnostic mammogram to biopsy were calculated. Normality of time to follow-up values was assessed by the Shapiro-Wilk test.
Median values were compared using the Wilcoxon test.
RESULTS
Seven hundred seventy-eight women were included in the study: 98 (12.6%) Caucasian, 275 (35.3%) Asian, 256 (32.9%) Hispanic,
and 100 (12.9%) African-American. Six-hundred sixty-two women were recalled for abnormal screening mammograms of whom 30
were lost to follow-up. The remaining 632 women had median follow-up time for an abnormal screening mammogram as follows: 15
days (95% CI 15.5, 17.5) for all women combined, 14 (10, 22.8) for Caucasian, 15 (10, 22) for Asian, 14 (9, 21.5) for Hispanic, and
15 (12, 28) for African-American women. Of the 280 women with diagnostic mammography BI-RADS assessment of 4 or 5, 13
declined biopsy and 12 were lost to follow-up. The remaining 255 patients underwent biopsy with a median time interval to biopsy in
days of 7.4 (95% CI 1.5, 14.5) for all women combined, 7.5 (3.5, 14.5) for Caucasian, 7.6 (2.5, 8.5) for Asian, 5.6 (1.4, 14.6) for
Hispanic, and 8.4 (2.4, 14.5) for African-American women. Relative to Caucasian women, there were no significant differences in
abnormal screening or diagnostic mammography follow-up intervals for any racial or ethnic group.
CONCLUSION
In contrast to previous studies, there were no significant delays in follow-up for abnormal screening and diagnostic mammograms
for minority women relative to Caucasian women at the studied urban safety-net hospital. Further research is needed to identify
factors that promote timely follow-up in the studied minority patient population.
CLINICAL RELEVANCE/APPLICATION
Delay in follow-up of abnormal mammograms can lead to adverse outcomes. Identifying where disparities occur can help address
barriers to care.
BR242-SD- Outcome Analysis of BI-RADS Category 3 Lesions in Young Women Imaged in Tertiary University
TUA2 Versus Safety-Net Hospital Setting: Is There an Impact of Healthcare Disparity?
Station #2
Participants
Jody C. Hayes, MD, Southlake, TX (Presenter) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Lena A. Omar, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Kanwal A. Merchant, MD, Dallas, TX (Abstract Co-Author) Stockholder, Sensogram Technologies; Spouse, Stockholder, Sensogram
Technologies
Yin Xi, PhD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
Basak E. Dogan, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
PURPOSE
To compare the follow up, image-guided biopsy and surgical excision patterns of women <30 with BI-RADS Category 3 lesions in a
safety-net hospital and an academic tertiary care center.
METHOD AND MATERIALS
In an IRB approved, HIPAA compliant study, consecutive women younger than 30 years who received sonographic (US) BI-RADS 3
assessment between January 2013 to December 2014 in breast imaging facilities of a safety-net (PMH) and a tertiary university
hospital (CUH) provided by the same fellowship-trained breast imaging group were retrospectively reviewed. We compared the
patient clinical presentation, imaging findings, lesion size and number, frequency and time interval of follow up, and biopsy and
excision status. Statistical analysis including chi-square and Cochran-Mantel-Haenszel (CMH) analysis for overall odds ratios were
performed.
RESULTS
One hundred and ninety-seven lesions in 157 women in PMH and 92 lesions in 70 patients in CUH had US BI-RADS 3 assessment.
Median patient age was the same in both groups (25 yrs). In PMH, 178 of 197 (90%) and in CUH 71 of 92 (77%) of lesions were
palpable (p= 0.2965). Mean lesion size was 1.6 cm (SD±0.8) in PMH, and 1.3 cm (SD±0.6) at CUH (p=0.2338). Patients were less
likely to complete all imaging follow up time points in PMH (n=29, 14.7%) compared to those in CUH (n=37, 52.8%) (p<0.05).While
the overall biopsy rates were similar (PMH, 23.5%, CUH, 19.1%), a higher rate of initial biopsies was observed in PMH (35 of 50
biopsies,70%) vs CUH (4 of 14, 29%) (p=0.004). The odds of undergoing a biopsy at initial presentation at PMH were significantly
higher [OR: 4.783 (95%CI 1.9-11.6)] compared to CUH [OR: 0.485, (95%CI 0.02-10.7)] (p= 0.0044). Biopsy was more likely to be
prompted by patient preference in PMH (39,70%) compared to CUH (22.2%)(p=0.005). Surgical excision rates were similar (8%PMH
vs 5%CUH) between the two groups. No malignancy was identified in either group.
CONCLUSION

There is a higher probability of incomplete follow up and initial needle biopsy for US BI-RADS Category 3 lesions identified in young
women in safety-net setting compared to university setting.
CLINICAL RELEVANCE/APPLICATION

Radiologists should consider disparities between population groups while recommending follow up for probably benign lesions in
women under 30, considering the low probability of malignancy in this age group.
BR243-SD- Pre-Treatment Prediction of Pathologic Complete Response to Neoadjuvant Chemotherapy in Node-
TUA3 Positive Breast Cancer Patients: A Breast MRI Radiomics Pilot Study
Station #3
Participants
Karen Drukker, PhD, Chicago, IL (Presenter) Royalties, Hologic, Inc
Christopher Doyle, MD , Chicago, IL (Abstract Co-Author) Nothing to Disclose
Alexandra V. Edwards, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc; Research Consultant,
Quantitative Insights, Inc
John Papaioannou, MSc, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc
Maryellen L. Giger, PhD, Chicago, IL (Abstract Co-Author) Stockholder, Hologic, Inc; Shareholder, Quantitative Insights, Inc;
Shareholder, QView Medical, Inc; Co-founder, Quantitative Insights, Inc; Royalties, Hologic, Inc; Royalties, General Electric
Company; Royalties, MEDIAN Technologies; Royalties, Riverain Technologies, LLC; Royalties, Mitsubishi Corporation; Royalties,
Canon Medical Systems Corporation
Kirti M. Kulkarni, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kdrukker@uchicago.edu
PURPOSE
Evaluate the ability of breast MRI radiomics to predict pathologic complete response of tumor and lymph nodes prior to neoadjuvant
chemotherapy (NAC) treatment in patients with invasive lymph node-positive breast cancer.
METHOD AND MATERIALS

Sixtyfive patients were included in this retrospective HIPAA compliant IRB approved study. Upon surgery, 15 showed post-NAC
complete pathologic response (pathologic TNM stage T0,N0,MX, 'complete-responders') and 50 showed incomplete response to NAC
('incomplete-responders'). Only pre-NAC MRIs underwent computer analysis, initialized by an expert breast radiologist indicating
index cancers and metastatic axillary sentinel lymph nodes (LNs) on DCE-MRI (T2 and T1 postcontrast subtraction) images.
Subsequent automated analysis included computer segmentation and extraction of 38 radiomic features describing (i) size, (ii)
shape, (iii) margin, (iv) kinetic curve, (v) contrast-enhancement texture, and (vi) variance kinetics. For cancers and LNs
separately, each radiomic feature was evaluated to determine whether a statistically significant difference between the complete-
responders and incomplete-responders was demonstrated (Mann-Whitney U-test). The area under the ROC curve was calculated
for the task of distinguishing between the two groups.
RESULTS

All radiomic features describing index cancers failed to show a statistically significant difference between complete-responders and
incomplete-responders (p>0.05). Four radiomic features describing pre-treatment metastatic LNs demonstrated statistically
significant differences between the two groups: effective diameter, sphericity, surface to volume ratio, and most enhancing nodal
volume. The most predictive feature was sphericity with area under the ROC curve of 0.75 (standard error 0.07) in the prediction of
pathologic response.
CONCLUSION

Radiomics for breast MRI shows promise in the pre-treatment prediction of pathologic response to neoadjuvant chemotherapy in
patients with lymph-node positive invasive breast cancer.
CLINICAL RELEVANCE/APPLICATION

The ability to predict which patients will demonstrate pathologic complete response before initiating neoadjuvant chemotherapy
could positively impact patient management by saving the cost of unnecessary chemotherapy and also the mortality and morbidity
associated with additional adjunct treatment in cases of incomplete pathologic response such as axillary dissection and radiation
therapy.
BR244-SD- Histological Whole-Slide Imaging for Invasive Breast Cancer: A Novel Technique to Obtain
TUA4 Quantitative Parameters Correlating with the Apparent Diffusion Coefficient
Station #4
Participants
Naoko Mori, MD, PhD, Sendai, Japan (Presenter) Nothing to Disclose
Chihiro Inoue, Sendai, Japan (Abstract Co-Author) Nothing to Disclose
Shunji Mugikura, MD, PhD, Sendai, Japan (Abstract Co-Author) Nothing to Disclose
Kei Takase, MD, PhD, Sendai, Japan (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
naokomori7127@gmail.com
PURPOSE
To evaluate whether parameters obtained by quantitative analysis of histological whole-slide imaging (WSI) correlates with the
apparent diffusion coefficient (ADC) in diffusion-weighted imaging (DWI) for invasive breast cancer.
METHOD AND MATERIALS
The Institutional Review Board approved this retrospective study and waived the requirement for informed consent. Between
September 2015 and July 2016, 46 consecutive patients with 46 invasive breast cancer lesions diagnosed by surgery underwent
preoperative breast magnetic resonance imaging (MRI), including DWI (b value: 50, 850s/mm2), followed by mastectomy or
lumpectomy without neoadjuvant therapy. Regions of interest (ROIs) covering as much of the tumor volume as possible were placed
manually on the maximum cross-section of the tumor on ADC maps, and the mean ADC values of the ROIs were recorded. For
histological analyses, the digital data of cytokeratin-immunostained thin-slice sections using the largest cross-sectional area of the
tumor were used for WSI. The cytoplasm, interstitium, and nucleus were identified and segmented by their brown, light blue, and
purple colors, respectively. The area ratios of cancer cells (the sum of the cytoplasm and nucleus) and interstitial space were
calculated by identifying the color of each component for the whole tumor area. For conventional cell counts, we randomly selected
five areas at ×200 field and manually counted cancer cells, and then the average counts of five areas were recorded as cell
counts. Then, all histological parameters were compared with the ADC using Pearson's correlation coefficient. A value of p < 0.05
was considered statistically significant.
RESULTS

The area ratios of the interstitial space were significantly positively correlated with the ADC (r=0.53; p=0.0001). The area ratios of
cancer cells and conventional cell counts were significantly negatively correlated with the ADC with a lesser degree. (r=-0.36 and -
0.32, respectively; p=0.012 and 0.027, respectively).
CONCLUSION

There is a significant positive correlation between the ADC value and area of the interstitium, as measured by WSI.
CLINICAL RELEVANCE/APPLICATION

There is a significant positive correlation between the ADC and the area of the interstitium, as measured by whole-slide imaging
(WSI). The ADC reflects the amount of interstitium in breast cancer tissue.
BR245-SD- Usefulness of CAD (Computer-Aided Detection) System for Screening Automatic Breast Ultrasound
TUA5
Station #5
Participants
Jeong Min Lee, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Bong Joo Kang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sung Hun Kim, Su Won, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Kim Myeongjong, Suwon, Korea, Republic Of (Presenter) Nothing to Disclose
PURPOSE
To evaluate diagnostic performance of screening automated breast ultrasound using computer aided detection (CAD) system and to
analyze the characteristics of CAD marks and the causes of false positive marks.
METHOD AND MATERIALS

A total of 846 women aged 40-49 years who underwent automated breast ultrasound for screening from January 2017 to December
2017 were included. We applied the CAD (QVCADTM) to all of the automated breast ultrasound examinations and evaluated its
diagnostic performance. Then, we analyzed the frequency, characteristics, and causes of false positive marks and tried to find out
a way to reduce the false positive marks. Furthermore, we analyzed whether adding CAD would shorten the reading time.
RESULTS

Out of a total of 846 patients, 534 CAD marks were displayed with an average CAD mark per person of 0.8 ± 1 (range 0-6).
Through screening automated breast ultrasound, five breast cancers were diagnosed. The sensitivity, specificity, PPV, NPV, and
accuracy of CAD were 60.0%, 59.0%, 0.9%, 99.6% and 59.0% for 846 patients, respectively, while those of 534 CAD marks were
60.0%, 48.3%, 0.6%, 99.6%, and 48.4%. Among 531 false positive marks, 459 false marks for pseudolesions were well identified;
the most common cause was marginal shadowing (209, 39.1%), then, Cooper's ligament shadowing (143, 26.8%), periareolar
shadowing (64, 12%), rib (37, 6.9 %), and skin lesion (6, 1.1%). The false marks for pseudolesions appeared in the upper portion
rather than mid to lower portion and in the outer portion rather than mid to inner portion. In the case of a negative study, it was
less time-consuming and easier to make a decision.
CONCLUSION

Adding CAD does not improve accuracy for screening automated breast ultrasound in this study, but adding it helps to reduce
reading time nonetheless for negative screening ultrasound.
CLINICAL RELEVANCE/APPLICATION

Using CAD system is helpful to reduce reading time for screening automated breast ultrasound, but does not improve accuracy.
BR246-SD- Analysis of Background Echotexture on Automated Breast Ultrasound: Correlation with
TUA6 Mammographic Density
Station #6
Participants
Eun Jung Choi, MD, PhD, Jeonju, Korea, Republic Of (Presenter) Nothing to Disclose
Ji Hyun Youk, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Bo Ram Kim, Jeonju, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

cejcej80@hanmail.net
PURPOSE

To analyze the background echotexture (BE) on automated breast ultrasound (ABUS) according to BI-RADS and modified BE
classification in correlation with mammographic breast density (BD).
METHOD AND MATERIALS

196 women (age, 31-72 yrs) who underwent ABUS and mammography were included. Their menopausal status, parity, and family
history of breast cancer were collected. After independently reviewing all ABUS and mammography images, three radiologists
assigned BE as homogeneous-fat, homogeneous-fibroglandular, or heterogeneous based on BI-RADS as well as homogeneous-fat,
homogeneous-fibroglandular, or hetergeneous-mild, moderate, and marked based on modified BE classification. BD was classified
into category A, B, C, or D based on BI-RADS. The interobserver agreement was measured by kappa statistics (κ). The association
among demographics, BE, and BD was analyzed by Spearman's correlation coefficient (ρ) and multiple linear regression.
RESULTS

The overall interobserver agreement for BE based on BI-RADS (κ=0.83), modified BE classification (κ=0.66), and BD (κ=0.84) was
substantial to nearly perfect. By consensus, 29 homogeneous-fat (14.8%), 144 homogeneous-fibroglandular (73.5%), and 23
heterogeneous (11.7%) were assigned based on BI-RADS, while 41 homogeneous fat (20.9%), 9 homogeneous-fibroglandular
(4.6%), or 46 hetergeneous-mild (23.5%), 65 moderate (33.2%), and 35 marked (17.9%) were assigned based on modified BE
classification. For BD, 21 category A (10.7%), 79 category B (40.3%), 63 category C (32.1%), and 33 category D (16.8%) were
assigned. The result of modified BE classification was positively correlated with BD (P<0.0001; overall, ρ=0.61; premenopausal,
ρ=0.42; postmenopausal, ρ=0.57). BE based on BI-RADS was significantly correlated with BD in overall (ρ=0.2, P=0.005) and
postmenopausal women (ρ=0.3, P=0.006). On multiple linear regression, BD and modified BE classification or parity and BE based on
BI-RADS was associated (P<0.0001).
CONCLUSION

Interobserver agreement of BE based on BI-RADS was nearly perfect, higher than modified BE classification. BE based on BI-RADS
and modified BE classification on ABUS had good correlation with BD.
CLINICAL RELEVANCE/APPLICATION

Automated breast ultrasound would enable the objective analysis of background echotexture throughout the whole breast and its
chronological change in correlation with mammographic density.
BR188-ED- Missed Breast Cancers: The Unconscious Bias in Breast Imaging
TUA7
Station #7
Awards
Cum Laude
Identified for RadioGraphics

Participants
Leslie Lamb, MD, Boston, MA (Presenter) Nothing to Disclose
Raman Verma, MD, Ottawa, ON (Abstract Co-Author) Nothing to Disclose
Jean M. Seely, MD, Ottawa, ON (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Medical errors are a substantial cause of morbidity and mortality and the third leading cause of death in the United States. Errors
resulting in missed breast cancer are the most prevalent condition precipitating medical malpractice lawsuits against all physicians.
In breast imaging, missed and interval breast cancers at screening mammography ranges from 1 to 35%. As such, it is important to
understand various cognitive processes during mammographic interpretation to improve awareness of unconscious biases and
decrease the number of missed breast cancers. It is particularly important to be aware of these biases when encountering the most
commonly missed and misinterpreted breast lesions. The purpose of this educational exhibit is to demonstrate various cognitive
processes that lead to unconscious biases through a pictorial review of missed breast cancers, and highlight strategies to reduce
the rates of these missed cancers.
TABLE OF CONTENTS/OUTLINE
The most commonly missed and misinterpreted lesions will be reviewed (stable, benign appearing and one-view masses, developing
asymmetries). These will help illustrate the common unconscious biases with appropriate minimization strategies in breast imaging
(anchoring, confirmation, satisfaction of search, hindsight, inattention, satisfaction of report and premature closing).
BR189-ED- Perceptive and Interpretive Pitfalls at Digital Breast Tomosynthesis (DBT) - Lessons from Continued
TUA8 Clinical Practice
Station #8
Awards
Certificate of Merit

Participants
Sushma Gaddam, MD, New York, NY (Presenter) Nothing to Disclose
Samantha L. Heller, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Yiming Gao, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

sushma.gaddam@nyumc.org
TEACHING POINTS

Digital breast tomosynthesis (DBT) is rapidly becoming the standard of care in mammographic screening, replacing conventional 2D
mammography. DBT technique continues to evolve, with synthetic 2D images (s2D) increasingly substituting for full field 2D
mammography, decreasing radiation dose. Because DBT is a relatively new technology, continued learning and improvement of
radiologist skills are essential. We will discuss perceptive and interpretive pitfalls of DBT interpretation, illustrate exceptions to the
rules in DBT diagnosis, and highlight teaching points via a case-based review.
TABLE OF CONTENTS/OUTLINE
1. Intro: a. 2D vs. DBT/2D vs. DBT/s2D b. Perceptive vs Interpretive errors2. Perceptive errors a. Satisfaction of search b.
Localization or triangulation errors on DBT c. Calcifications on synthetic 2D images d. DBT occult cancers e. Extremely dense breast
f. Invasive lobular cancers g. Technical issues - patient positioning, motion3. Interpretive errors a. Skin vs. superficial lesions b.
One-view-only findings c. Developing asymmetries d. Fat containing lesions e. Architectural distortions f. DBT-only architectural
distortions4. Case illustrations - highlight pearls and pitfalls
BR190-ED- Breast Imaging of Ductal Carcinoma in Situ: Dilemma Between Overtreatment and Underestimation
TUA9
Station #9
Participants
Hiroko Satake, MD, Nagoya, Japan (Presenter) Nothing to Disclose
Satoko Ishigaki, MD, Nagoya, Japan (Abstract Co-Author) Nothing to Disclose
Shinji Naganawa, MD, Nagoya, Japan (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Given the uncertainty in determining whether ductal carcinoma in situ (DCIS) is likely to progress, every DCIS is intensively treated
as invasive breast cancer. Overdiagnosis and overtreatment may potentially occur, especially for low-grade DCIS. On the other
hand, a pathologically positive or a close-to-margin status, due to preoperative underestimation, are risk factors for local DCIS
recurrence post-excision. Furthermore, we occasionally encounter preoperatively diagnosed DCIS that is then upgraded to invasive
cancer postoperatively. We review the image findings in DCIS and discuss how they can be used to resolve dilemmas between
overtreatment and underestimation.
TABLE OF CONTENTS/OUTLINE

1. Explain the controversies surrounding DCIS diagnosis and treatment. 2. Review the imaging features of DCIS on mammography,
ultrasonography, and MRI. 3. Review the usefulness and pitfalls of preoperative MRI for DCIS to estimate tumor extensions. 4.
Review the current literature and discuss the capabilities of diagnostic imaging, focusing on the following topics: · Differentiating
low-grade from high-grade DCIS · Preoperative predictions concerning occult invasion in DCIS · Identifying imaging risk factors for
local DCIS recurrence · Quantitative imaging analysis of DCIS
BR191-ED- Not Your Mother's Breast Cancer - Imaging and Review of Breast Cancer Under 30 and Other Mimics
TUA10
Station #10
Participants
Vincent G. Champion, MD, Westwood, MA (Presenter) Nothing to Disclose
Bonny Lee, MD, MS, brookline, MA (Abstract Co-Author) Nothing to Disclose
Eileen Delaney, MD, Worcester, MA (Abstract Co-Author) Nothing to Disclose
Amy K. Patel, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Evguenia J. Karimova, MD, Memphis, TN (Abstract Co-Author) Nothing to Disclose
Valerie J. Fein-Zachary, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Michael D. Fishman, MD, Boston, MA (Abstract Co-Author) Consultant, Zebra Medical Vision Ltd
Jordana Phillips, MD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Consultant, General Electric
Company
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

1. Review the role of imaging in the diagnosis of breast cancer and potential mimics for women under 30 years old. 2. Discuss the
unique challenges involved in the diagnosis and management of breast cancer in young patients.
TABLE OF CONTENTS/OUTLINE

1. Review the current state of breast cancer in young women. 2. Discuss risk factors related to development of breast cancer in
young women including socioeconomic status and race 3. Review unique pathologic features encountered in young women with
breast cancer as opposed to those in older women. 4. Review imaging protocols used in the diagnosis and management of breast
cancer in the young women. 5. Provide examples of mimickers of breast cancer encountered in young women. 6. Discuss
management considerations for the adolescent or young adults with breast cancer including risks of surgery, chemotherapy,
radiation, and hormonal therapy.
BR192-ED- The Bright Side of the Post-Surgical Breast: Benign Findings After Breast Manipulation
TUA11
Station #11
Participants
Pedro Henrique Hasimoto e Souza, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Juliana H. Catani, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Sofia R. Cartaxo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carla C. Caravatto, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
pedrohenriquehs@hotmail.com
TEACHING POINTS
Prepare the radiologist to face common and uncommon presentations of manipulation / post-surgical benign changes in breast
imaging; Exhibit clinical cases from our radiology department, with various types of manipulation resulting in breast benign imaging
findings; Discuss the effects of manipulation on breast cancer detection; Provide management recommendations.
TABLE OF CONTENTS/OUTLINE

- Brief discussion on post-treatment follow-up care in breast cancer; - Risk factors and incidence of recurrence; - Imaging features
(mammography, ultrasound and magnetic resonance imaging) of manipulation / post-surgical benign changes in breast imaging: 1.
Mastectomy and variants 2. Breast implants Expanders and silicone implants Double prosthesis Implants in transexual patients Intra
and extracapsular ruptures Industrial liquid silicone injection 3.Mastopexy Mesh Internal suture 4. Reduction mammoplasty Vertical
incision Inferior pedicle 5. Autologous reconstruction Autologous fat grafting Pedicled TRAM flap Dorsal flap 6. Miscellaneous Fat
necrosis Subcutaneous emphysema Radiotherapy actinic changes Keloid Tattoo Others
BR198-ED- Luminal B Subtype Breast Cancer: Radiogenomic Correlation
TUA12
Station #12
Participants
Bruna Mannato, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Decio Roveda Junior, PhD, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Caio Castro, MD, Santo Andre, Brazil (Abstract Co-Author) Nothing to Disclose
Mario S. Campos, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Gustavo M. Badan, MD,PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

brunamannato@hotmail.com
TEACHING POINTS
1. The main objective of this pictorial review is to show the most prevalent imaging patterns of that luminal B subtype at
mammography (MMG), ultrasound (US) and magnetic resonance (MR). 2. Series of cases of patients diagnosed with invasive breast
cancer, between January 2015 and March 2018, classified as luminal B, analyzing the most prevalent features in the different
imaging modalities (MMG, US and MR). 3. Recent literature review and comparison with the data found in our service.
TABLE OF CONTENTS/OUTLINE
• Introduction - The traditional clinicopathological model. - New classification Immunohistochemical • Luminal Subtype Breast -
Cancer Luminal B x Luminal A • Imaging features -MMG - US - MR • Radiogenomics correlation • Conclusion
BR009-EB- Quantitative MRI Radiomics in Breast Cancer: Tumor Perfusion and Heterogeneity
TUA
Hardcopy Backboard
Participants
Myoung-Ae Kwon, Ansan, Korea, Republic Of (Presenter) Nothing to Disclose
Bo Kyoung Seo, MD, PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Research Grant, Canon Medical Systems Corporation;
Research Grant, Guerbet SA; Research Grant, Koninklijke Philips NV;
Eun Kyung Park, MD,PhD, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Chang Sub Ko, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Young Ju Son, Ansan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Kyu Ran Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ok Hee Woo, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

kwon4715@gmail.com
TEACHING POINTS

Radiomics is defined as the use of automated post-processing and analysis of quantitative imaging characteristics that can be
extracted from medical images. Quantification of imaging features is important to link the biomarkers with biological processes and
clinical endpoints. Magnetic resonance imaging(MRI) is the most sensitive imaging modality in breast cancer and has been used for
staging before treatment and assessing the response to the chemotherapy. Perfusion and tumor heterogeneity are related with
cancer biological characteristics and prognosis. In this exhibit, we will illustrate how to quantitatively assess tumor perfusion and
heterogeneity on breast MRI and correlate radiological imaging parameters with prognostic biomarkers and clinical outcomes. In
addition, we will discuss the future potential of radiomics in oncology imaging.
TABLE OF CONTENTS/OUTLINE
1) Radiomics : Definition : Requirements for quantification of medical images 2) Assessment of tumor perfusion on breast MRI :
Imaging technique : Quantitative parameters 3) Assessment of tumor heterogeneity on breast MRI : Texture analysis mechanism :
Quantitative parameters 4) Correlation of quantitative parameters of tumor perfusion and heterogeneity with biomarkers and clinical
outcomes 5) Future potential of radiomics in oncology imaging
 LL21

Lunch and Learn: The Future of Women's Breast Health: Breakthrough Technologies in Breast Imaging,
Patient Experience and Clinical Research: Presented by Mammosphere (invite-only)
Tuesday, Nov. 27 12:30PM - 1:30PM Room: S404AB

Participants
Ethan Cohen, MD, Assistant Professor, Department of Diagnostic Radiology, Division of Breast Imaging, The University of Texas MD
Anderson Cancer Center; Heather Greenwood, MD, an Assistant Professor of Clinical Radiology in the Breast Imaging section at the
University of California, San Francisco; Kathryn Pearson Peyton, MD, Founder of Mammosphere; Cristin Gardner, Director of
Consumer Products & Markets, Mammosphere
Program Information
While Radiology continues to shift from a volume to value-based world, it's not a new concept for breast imagers who have been
practicing value-driven care since the inception of the Mammography Quality Standards Act (MQSA) in the early 90s. At the core of
these efforts, patient experience remains key to optimize care and reduce costs. This means offering patients the right tools that
drive engagement to achieve these desired outcomes. In this digital age, it's no longer a "nice to have." For mammography, access
to a patient's data - especially prior exams - is critical for a better experience and outcomes. As more patients live increasingly
mobile lifestyles, it can be challenging for both the provider and patient to gather the data needed at the time of the exam. This
can increase the rate of false-positive diagnoses, leading to unnecessary orders of additional expensive diagnostic tests and
biopsies. Patients also experience needless anxiety. With so many health and wellness platforms available, how can providers, along
with healthcare payers and employers, provide the right resources to women in the populations they serve to help them gather
their medical data while empowering them to take control of their health? This lunch and learn session will feature a panel of
clinical, academic and technology experts in this field exploring the following topics: The current state of breast health research
and its impact on patient behavior and technology. The best practices organizations can follow to incorporate breast health
resources into their population health strategy. How to effectively implement breast health technology to drive patient
engagement.
 LL23

Lunch and Learn: Real-World Deployment of Deep Learning for Breast Cancer Screening: Presented by
Kheiron Medical Technologies (invite-only)
Tuesday, Nov. 27 12:30PM - 1:30PM Room: S403B

Participants
Peter D. Kecskemethy, PhD , London, United Kingdom (Presenter) Stockholder, Kheiron Medical Technologies Ltd
Hugh Harvey, MBBS, London, United Kingdom (Presenter) Employee, Kheiron Medical
Christopher C. Austin, MBBCh,MSc, Seattle, WA (Presenter)
Program Information
Kheiron is at the cutting edge of deep learning technology for breast cancer screening, and leading the way for healthcare and
industry collaborations in real-world deployment of deep learning technologies. Please join us for this lunchtime session where we
will use our experiences of partnerships and real-world deployment to explore and discuss this exciting new space, including: - the
organizational benefits and challenges posed by the next generation of supportive technology - how radiologists can take a leading
role in the development and adoption of deep learning in radiology - discussion on the challenges faced by the industry, including
structural changes and operations of professional bodies, individual providers, and nationwide programs.
RSVP Link
https://www.kheironmed.com/rsna
 VW 11

3D ABUS: Hands-on Workshop: GE Vendor Workshop

Tuesday, Nov. 27 12:30PM - 1:30PM Room: Booth 8156

Program Information
Peer educator led workshop includes a live Invenia ABUS 2.0 scan station acquisition and hands-on review of clinical cases using
the Invenia ABUS Viewer. Learn about the importance of the coronal view and how 3D ABUS screening helps increase cancer
detection in women with dense breast tissue. Registration is required; adding this session to the RSNA calendar tool alone does not
secure your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 BRS-TUB

Breast Tuesday Poster Discussions

Tuesday, Nov. 27 12:45PM - 1:15PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

Participants
Samantha L. Heller, MD, PhD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
BR247-SD- Clinical Usefulness of Digital Breast Tomosynthesis (DBT) and Hybrid 18F-FDG PETMRI (PETMR) for
TUB1 Monitoring Neoadjuvant Chemotherapy (NAC) in Breast Cancer
Station #1
Participants
Nachiko Uchiyama, MD, Tokyo, Japan (Presenter) Nothing to Disclose
Hiroaki Kurihara, MD, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Takayuki Kinoshita, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Masayuki Yoshida, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Mari Kikuchi, MD, Chuou-ku, Japan (Abstract Co-Author) Nothing to Disclose
Kyoichi Otsuka, Toyko, Japan (Abstract Co-Author) Nothing to Disclose
Masahiko Kusumoto, MD, Chuo, Japan (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
nuchiyam@ncc.go.jp
PURPOSE
To compare the usefulness of DBT and PETMR for evaluation of loco-regional staging and treatment response of NAC with reference
to pathological findings.
METHOD AND MATERIALS
30 invasive breast cancers (IDC: n=27, ILC n=2, and Metaplastic Ca: n=1) among 28 cases with 29 breasts, that NAC was
preoperatively underwent, were enrolled. The average age was 49.71 ± 11.94 y. o. Clinical stages were IIA (n=6), IIB (n=9), IIIA
(n=4), and IIIC (n=10). MMG+DBT, US, and PETMR were obtained before and after NAC. Regarding PETMR, as whole-body scans
was obtained after intravenous injection of 18F-FDG followed by a resting period of 60 min in a spine position as early phase In
addition, breast MRI were conducted with a dedicated bilateral 8-channel breast radiofrequency coil in a prone position and
contrast enhanced four phases of dynamic studies were obtained. As well as breast MRI, breast PET images as late phase of 80
min. Regarding DBT, the images were acquired by MLO and CC views with the rotation angle of ± 25° and reconstructed into 2mm
thick slice having 1mm overlap with high in-plane resolution of 85μm. NAC response was classified in accordance with RECIST and
compared with pathological response in accordance with JBCS.
RESULTS
Before NAC, PETMR detected primary lesions with CEMR and with PET by 100% (30/30). Regarding LN metastasis, the diagnostic
accuracy was 96.6% (28/29) by CEMR and 93.1% (27/29) by PET. Among the pathological response of Grade 0-1a (n=5),
MMG+DBT demonstrated as SD (3/5: 60%) and PR (2/5: 40%). PETMR demonstrated as SD (2/5: 40%) and PR (3/5: 60%). Among
Grade 1b-2 lesions (n=17), MMG+DBT detected as PR (13/17: 76.5%) and CR (4/17: 23.5%). PETMR demonstrated as PR (14/17:
82.4%) and CR (3/17: 17.6%). Among Grade 3 (n=8), MMG+DBT demonstrated as CR (6/8: 75.0%) and PR (2/8: 25.0%). PETMR
demonstrated as CR (4/8: 50%) and PR (4/8: 50%). In addition, we evaluated estimated pathological responses with MMG+DBT and
PETMR to compare the actual pathological results.
CONCLUSION
Concurrent usage of MMG+DBT and PETMR can demonstrate promising results for loco-regional staging prior to NAC and can predict
more accurate treatment response after NAC.
CLINICAL RELEVANCE/APPLICATION
Concurrent usage of MMG+DBT and hybrid PETMR are useful among NAC cases of breast cancer, because they can provide detailed
analysis by morphological evaluation with DBT, by evaluation of vascularity with dynamic MR , and by metabolic evaluation with
PET.
BR248-SD- Changes in the Utilization of the BI-RADS Category 3 Assessment in Recalled Patients Screened with
TUB2 Digital Mammography Before and After the Implementation of Diagnostic Digital Breast
Tomosynthesis
Station #2
Participants
Tricia Stepanek, Cleveland, OH (Presenter) Nothing to Disclose
Niki M. Constantinou, MD, Westlake, OH (Abstract Co-Author) Nothing to Disclose
Holly N. Marshall, MD, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Ramya M. Pham, MD, Solon, OH (Abstract Co-Author) Nothing to Disclose
Cheryl L. Thompson, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Donna M. Plecha, MD, Strongsville, OH (Abstract Co-Author) Research Grant, Hologic, Inc
PURPOSE
To compare the utilization of the Breast Imaging Reporting and Data System (BI-RADS) category 3 assessment in patients recalled
from digital mammography (DM) screening before and after the implementation of digital breast tomosynthesis (DBT) in the
diagnostic setting.
METHOD AND MATERIALS
This was an IRB-approved and HIPAA-compliant retrospective review of 22980 DM screening exams. The diagnostic DM cohort was
limited to patients screened with DM between January 1, 2010 and August 31, 2011 (n = 11478) who were assigned a category 3
after diagnostic DM work-up without tomosynthesis. The diagnostic DM+DBT cohort was limited to patients screened with DM
between January 1, 2014 and June 30, 2015 (n = 11502) who were assigned a category 3 after diagnostic DM+DBT work-up.
Diagnostic ultrasound was performed at the discretion of the radiologist in both cohorts. Category 3 lesions were classified as
architectural distortion, asymmetry, calcification, mass, and other and followed for a minimum of 2 years.
RESULTS
The addition of DBT to diagnostic work-up after screening with DM resulted in a 61.9% reduction (22.6 women per 1000) in the
utilization of BI-RADS category 3 compared to diagnostic DM work-up alone (3.7% for DM vs. 1.4% for DM+DBT; p < 0.0001). There
was a statistically significant change in the distribution of category 3 findings with DM+DBT characterized by an increase in masses
(p = 0.0056) and a decrease in calcifications (p = 0.0024). There was no change in category 3 assessment for distortions (p >
0.99), asymmetries (p = 0.073), or other findings (p = 0.58). Diagnostic DM+DBT resulted in a two-fold increase in incidental
findings leading to a category 3 compared to DM alone (2.3% for DM vs. 7.0% for DM+DBT; p = 0.0053). The delayed cancer
detection rate was 1.4% (10 malignancies in 419 patients) for category 3 lesions diagnosed by DM and 0% (0 malignancies in 160
patients) for category 3 lesions diagnosed by DM+DBT (p = 0.069).
CONCLUSION

DBT in the diagnostic evaluation of patients recalled from DM screening decreased the number of patients assigned to short-term
follow-up by 22.6 women per 1000 (61.9%), despite increased rates of category 3 incidental findings with 0% delayed cancer
detection.
CLINICAL RELEVANCE/APPLICATION

The use of DBT for diagnostic work-up at recall from DM screening may reduce the utilization of the BI-RADS category 3
assessment, decreasing the number of women committed to short-term follow-up.
BR249-SD- Management of Atypical Breast Papillomas by US-Guided Vacuum-Assisted Removal: Long-Term
TUB3 Outcomes
Station #3
Participants
Jose Maria Oliver-Goldaracena, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Maria Jose Roca Navarro, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Vicenta Cordoba Chicote, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Diego Garrido Alonso, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Agustin Andres Mateo, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Cesar Oterino Serrano, MBBS, Madrid, Spain (Presenter) Nothing to Disclose
Miguel Bello Erias, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Alberto Jimenez, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Maria Jesus Garcia Sanchez, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
jmolivergoldaracena@gmail.com
PURPOSE
To review long-term outcomes of the percutaneous management of bening breast papillomas with atypia (AP) by US-Guided
Vacuum-Assisted Removal (US-VAR).
METHOD AND MATERIALS

In our Institution, probable intraductal papillomas diagnosed with US (benign intraductal mass within a dilated duct or cyst with
Color-Doppler signal or correlation on ductography) or papillomas diagnosed at US-CNB of less than 30mm are managed by US-VAR.
Between April 2010 and March 2018 , 149 probable intraductal papillomas and 62 papillomas (diagnosed at US-CNB :54 Benign
papillomas, BP and 8 atypical papillomas, AP) in 188 patients (pts) were removed with US-VAR. Histology showed benign papilloma
(BP) in 149 (168pts), atypical papilloma in 12(11pts) and papillary carcinoma in 9 (9pts). Conservative management was decided in
the 12 APs (mean size 7mm, range 5-26mm) of 11pts (mean age 61, range 35-68) because of appropiate radiologic-pathologic
correlation (pathology report showed atypical hyperplasia without suspected DCIS). Pathologic dischargue was present in 4APs
(4pts) with US findings that showed probable intraductal papillomas . The 8 asymptomatic APs(7pts) were previously diagnosed by
US-CNB as 4 BPs and 4 APs. The US findings of the 12 APs were classified BIRADS 3 (8), BIRADS 4A (2) and BIRADS 4B (2). All
patients underwent US follow-up at 1-2 months, 6-8 months, and 12-14 months after US-VAR and later annual US follow-up. When
a residual or recurrent suspicious papilloma was detected at US follow-up, re-excision by US-VAR was performed. Clinical, US
follow-up and pathologic outcomes were recorded.
RESULTS
US follow-up ranged between 84 and 19 months (mean 68 months). No recurrent lesions were detected . None were upgraded to
carcinoma at long term US follow-up. Nipple discharge disappeared in all 4 symptomatic patients.
CONCLUSION

US-VAR allows percutaneous long-term management of atypical papillomas with proper radiologic-pathologic correlation.
CLINICAL RELEVANCE/APPLICATION

US-VAR is appropriate for the long-term percutaneous management of atypical papillomas with proper radiologic-pathologic
correlation.
BR250-SD- Longitudinal Investigation of Tumor Heterogeneity in Breast DCE-MRI to Improve Early Response
TUB4 Assessment to Neoadjuvant Chemotherapy for Locally Advanced Breast Cancer: Results From the
ACRIN 6657/I-SPY-1 Trial
Station #4
Participants
Nariman Jahani, Philadelphia, PA (Presenter) Nothing to Disclose
Eric A. Cohen, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Meng-Kang Hsieh, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Susan Weinstein, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Christos Davatzikos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

nariman.jahani@uphs.upenn.edu
PURPOSE

Utilizing longitudinal dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) scans of locally advanced breast cancer
under neoadjuvant chemotherapy (NAC), we evaluate voxel-wise spatio-temporal changes in tumor heterogeneity to improve early
prediction of pathologic complete response (pCR) and recurrence free survival (RFS).
METHOD AND MATERIALS

We retrospectively analyzed 106 women from a subset of ACRIN 6657/I-SPY TRIAL with complete imaging and prognostic marker
data using DCE-MRI scans acquired before and after the first cycle of NAC. Utilizing a robust deformable image registration
technique, we identified corresponding voxels between pre- and post-treatment images and extracted two groups of voxel-wise
features to quantify longitudinal tumor heterogeneity: 1) voxel-wise tumor deformation including local volume ratio and anisotropic
deformation; and 2) parametric response maps of DCE-MRI kinetic variables (signal enhancement ratio, peak enhancement, wash-
in/wash-out slope). Using best-subset regression, features were added to a baseline model including established prognostic markers
and functional tumor volume to predict pCR and RFS. Multivariate logistic regression and Cox proportional hazard models were used
to assess pCR and RFS, respectively. Furthermore, conventional analysis where analogous metrics were averaged over the tumor
was conducted to compare to voxel-wise analysis. The performances of models were evaluated with area under curve (AUC) and c-
statistics for pCR and RFS analyses, respectively (flowchart in Fig.1).
RESULTS
Voxel-wise analysis improved prediction significantly for both pCR and RFS, with AUC=0.85 and c-statistics=0.79, compared to the
baseline model with AUC=0.76 and c-statistics=0.69, respectively. The aggregate model indicated similar performance to the
baseline model with AUC=0.76 and c-statistics=0.70, respectively.
CONCLUSION
Our results suggest that quantification of voxel-wise changes after NAC can extract markers revealing spatio-temporal tumor
heterogeneity that can significantly improve early tumor response assessment, while conventional aggregate tumor features may
not adequately capture such longitudinal changes to add significant new information for prognosis.
CLINICAL RELEVANCE/APPLICATION

Improvement of early treatment assessment using markers based on voxel-wise feature analysis may provide complementary new
information to better modify treatment plans and to optimize therapy.
BR251-SD- Development and Validation of a Deep Learning Model For More Accurate and Consistent Assessment
TUB5 of MRI Background Parenchymal Enhancement
Station #5
Participants
Benjamin Wang, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Tally Portnoi, Cambridge, MA (Abstract Co-Author) Nothing to Disclose
Regina Barzilay, PhD, Cambridge, MA (Abstract Co-Author) Nothing to Disclose
Brian N. Dontchos, MD, Boston, MA (Presenter) Nothing to Disclose
Dorothy A. Sippo, MD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company
Christine E. Edmonds, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Randy C. Miles, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Constance D. Lehman, MD,PhD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Medical Advisory
Board, General Electric Company
For information about this presentation, contact:

bdontchos@mgh.harvard.edu
PURPOSE

Increased MRI background parenchymal enhancement (BPE) is associated with breast cancer risk. However, radiologist BPE
assessment is variable. Our goal was to train a deep learning (DL) model to assess breast MRI BPE and to evaluate its performance
compared to expert radiologists.
METHOD AND MATERIALS
After IRB approval, consecutive breast MRI examinations from January 2011 to December 2014 with BPE assessment were identified
(n=4683). 3249, 696, and 885 exams were selected for the training, validation, and test data sets, respectively. A single
representative maximum intensity projection image from each examination and the radiologist's prospective assessment of BPE were
used to train a deep learning classifier (ResNet). From the test data set, a sample of 100 MRI exams were then randomly selected
and submitted for blinded review by 5 fellowship trained breast imaging radiologists for binary assessment of BPE (minimal versus
non-minimal). We estimated both our DL model agreement and our radiologist readers' agreement with the original radiologist's BPE
assessment using percent agreement with Wilson confidence intervals (CI) and with linear-weighted kappa statistics, compared
across 5,000 bootstrap samples to assess significance.
RESULTS

For the subset of 100 reader study exams, DL model agreement with the original radiologist's BPE assessment was 80.0% (95% CI
of 71.1, 86.7) for binary assessment. The 5 radiologist readers' agreement with the original radiologist's BPE assessment for the
same 100 cases was 69.8% (95% CI 65.4, 73.5) for binary assessment (p=0.02). Compared to the original radiologist's BPE
assessment, our DL model showed moderate agreement (K=0.55, 95% CI 0.38-0.74) on this subset compared to weak agreement
by the 5 radiologist readers' consensus assessment (K=0.43, 95% CI 0.26-0.61).
CONCLUSION

A deep learning model can more accurately and reliably assess BPE based on maximal intensity projection images than a human
reader.
CLINICAL RELEVANCE/APPLICATION

A deep learning model that accurately and consistently predicts BPE could reduce variability in radiologist BPE assessment and may
also serve as a reliable tool for breast cancer risk prediction.
BR252-SD- Diagnosis of Triple Negative Breast Cancer Using Machine Learning Methods of Quantitative
TUB6 Computerized Ultrasound Features
Station #6
Participants
Tong Wu, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Laith R. Sultan, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Theodore Cary, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Jiawei Tian, Harbin, China (Abstract Co-Author) Nothing to Disclose
Chandra Sehgal, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

lsutlan@pennmedicine.upenn.edu
PURPOSE
Triple negative (TN) breast cancers are known for aggressive biological characteristics and poor clinical outcomes. In the current-
state-of-the-art much of the technological advances have been aimed at differentiating malignant and benign tumors where all the
subtypes are grouped together. In this study, we go beyond this discrimination of the solid breast masses by applying computer
methods with ultrasound imaging to differentiate TN and non-triple negative (NTN) subtypes.
METHOD AND MATERIALS

140 surgically confirmed breast cancer were classified into triple negative (TN) and non-triple negative (NTN) subtypes based on
the expression of ER, PR, HER2. Nine quantitative grayscale features describing margin and shape characteristics of the lesion, and
three tumor vascularity features describing the magnitude of vascularity were extracted from a manually drawn region of interest
on grayscale and color Doppler images. The features that showed difference (P<0.05) were used with logistic regression and leave-
one-out cross validation to train and test the differentiation of TN and NTN masses. Diagnostic performance was measured by the
area under ROC (AUC) and sensitivity and specificity measured at the Youdons index.
RESULTS

Twenty-five of the 140 cases were found to be TN. Of the twelve grayscale and Doppler features, eight showed statistical
difference (P< 0.002) for the TN and NTN. AUC of the statistically significant GS and Doppler features when used alone was 0.850
and 0.657, respectively. The AUC increased to 0.882 when all the significant GS and CD features were used. The improvement by
inclusion of Doppler features was significant (P <0.0001). Sensitivity and specificity of combined grayscale and Doppler was 78.26%
and 85.47%, respectively. Consideration of patient age in the analysis did not improve discrimination of TN and NTN.
CONCLUSION

The analysis of breast ultrasound by machine learning can achieve high level of differentiation between the TN and NTN subtypes
that is comparable to the diagnostic performance by standard visual assessments of the images.
CLINICAL RELEVANCE/APPLICATION

TN breast cancers are high grade and aggressive with shorter survival time, higher metastasis and recurrence. This study proposes
a quantitative sonographic approach for improving the TN diagnosis
BR253-SD- Comparison of Automated Breast Ultrasound (ABUS) QVCAD Standalone Performance on Somo-v
TUB7 and Invenia Cases
Station #7
Participants
Yulei Jiang, PhD, Chicago, IL (Presenter) Research Grant, QView Medical, Inc; Research Consultant, QView Medical, Inc; Research
Consultant, Quantitative Insights, Inc; Research Consultant, RadOnc eLearning Center, LLC
Gene Pennello, PhD, Silver Spring, MD (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

yjiang@uchicago.edu
PURPOSE

To evaluate QVCAD standalone performance on somo•v and Invenia cases, and to infer likely effect of QVCAD on Invenia cases
from its effect on somo•v cases.
METHOD AND MATERIALS

Evaluation was made on 164 somo•v (31 with cancer) and 145 Invenia (25 with cancer) cases, all of BI-RADS density C or D. Area
under ROC curve (AUC), per-patient sensitivity and specificity, false-positives per volume, and FROC curves were compared. Non-
inferiority on Invenia vs. somo•v cases was tested with Bonferroni correction at a family-wise a=0.05 level for non-inferiority
margins of -0.05 for AUC, -5% for sensitivity, -5% for specificity, and +0.05 for false positives per volume. A previous observer
study of somo•v cases evaluated QVCAD effect on reader performance.
RESULTS
Somo•v and Invenia results were comparable. AUC was 0.73±0.05 (AUC ± standard error) for somo•v vs. 0.79±0.07 for Invenia. For
QVCAD computer marks, sensitivity was 71.0% (22/31) [95% CI: 53.4%, 84.8%] vs. 84.0% (21/25) [65.8%, 94.7%], specificity
was 49.6% (66/133) [41.2%, 58.0%] vs. 55.0% (66/120) [46.0%, 63.7%], and false-positive rate was 0.11 (101/891) [0.09, 0.14]
per volume vs. 0.15 (107/697) [0.12, 0.19] per volume, respectively. For QVCAD computer-enhanced dark areas, sensitivity was
96.8% (30/31) [85.1%, 99.8%] vs. 100.0% (25/25) [88.7%, 100.0%], specificity was 6.0% (8/133) [2.8%, 11.1%] vs. 7.5%
(9/120) [3.7%, 13.3%], and false-positive rate was 0.67 (596/891) [0.60, 0.74] per volume vs. 0.75 (525/697) [0.65, 0.85] per
volume, respectively. Non-inferiority tests with Bonferroni correction did not show statistical significance. FROC curves were similar
or apparently higher for Invenia than somo•v (Fig.). Previous observer study showed concurrent read of somo•v screening cases
with QVCAD reduced reading time and produced non-inferior diagnostic accuracy compared with no QVCAD.
CONCLUSION

QVCAD standalone performance is comparable on somo•v and Invenia cases. Its benefit for reducing reading time and producing
non-inferior reader performance can be expected on Invenia cases.
CLINICAL RELEVANCE/APPLICATION

This study provides a performance benchmark for clinical use of QVCAD on the current Invenia ABUS system and a frame of
reference to the results of a previous observer study done on somo•v cases.
BR193-ED- Freezing Instead of Excising: Cryoablation for Breast Cancer - How It's Done
TUB8
Station #8
Awards
Magna Cum Laude
Participants
Linda DeMello, MD, Providence, RI (Presenter) Nothing to Disclose
Robert C. Ward, MD, Providence, RI (Abstract Co-Author) Nothing to Disclose
Martha B. Mainiero, MD, Providence, RI (Abstract Co-Author) Nothing to Disclose
Ana P. Lourenco, MD, Providence, RI (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

ldemello2@lifespan.org
TEACHING POINTS

1. To illustrate the cryoablation techniques utilized in early stage breast cancer, highlighting the importance of avoiding potential
complications and pitfalls. 2. To present a series of cryoablation cases from our breast center, including follow-up imaging. 3. To
recognize the role of cryoablation in treating early stage cancers in patients who are not surgical candidates or prefer an
alternative treatment to surgery.
TABLE OF CONTENTS/OUTLINE

1. Introduction to Cryoablation 2. Techniques and Equipment 3. Presentation of at least 5 cases from our breast center a. Each
with a different dilemma with regard to clinical context and technique b. Detail potential pitfalls and complications along with
strategies to avoid them 4. Patient experience and outcomes 5. Future applications of cryoablation and summary
BR194-ED- Getting Acquainted with the Man(l)y Facets of the Lesser Known Male Breast: Lessons Learned From
TUB9 a 10-year Institutional Look-Back
Station #9
Awards
Certificate of Merit

Participants
Nayanatara Swamy, MD , Little Rock, AR (Presenter) Nothing to Disclose
Gwendolyn M. Bryant-Smith, MD, Little Rock, AR (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

nswamy@uams.edu
TEACHING POINTS

1. The fundamental anatomical differences between the male and female breast helps us understand the vast difference in
epidemiology of male and female breast diseases, including breast cancer. Accordingly, breast imaging guidelines are different in
men and women. 2.We used our institutional experience over the past decade to understand our regional male breast epidemiology.
a. The most common finding in male diagnostic mammograms is gynecomastia. Hence, familiarity with the wide spectrum of
gynecomastia appearances on mammography is crucial. b. Male patients are less likely to return for short term follow-up imaging. A
change in our approach to male patients is necessary to improve their compliance.
TABLE OF CONTENTS/OUTLINE

1. Differences between male and female breast: - Anatomical differences. - The transgender breast. - How do male breast diseases
differ from female breast diseases? - How does male breast cancer differ from female breast cancer? - Difference in screening
guidelines for men and women. - Difference in ACR appropriateness criteria for an indeterminate palpable breast mass in males
compared to females. - Is Imaging utilization different in male versus female patients? 2. Our Institutional 10 year results. 3.
Lessons learned. 4. Conclusion.
BR195-ED- The 123's of Breast Cancer Staging
TUB10
Station #10
Participants
Tirth V. Patel, MD, Chapel Hill, NC (Presenter) Nothing to Disclose
Marie Vogel, BS, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
Sheila S. Lee, MD, Durham, NC (Abstract Co-Author) Nothing to Disclose
Sheryl G. Jordan, MD, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
tirth.patel@unchealth.unc.edu
TEACHING POINTS

The purposes of this exhibit are to: 1. Educate radiologists on the American Joint Committee on Cancer (AJCC) eighth edition
staging manual, the newly published (2018) prognostic stage for breast cancer 2. Review relevant breast and locoregional lymph
node anatomy 3. Present breast cancer survival statistics by stage 4. Identify key imaging findings for radiologists to assure
accurate breast cancer staging
TABLE OF CONTENTS/OUTLINE
1. Introduction with relevant anatomy and survival statistics on breast cancer 2. Outline the AJCC 8th edition breast cancer
staging system and define the TNM (Tumor, Nodes, Metastasis) system 3. Case-based illustration of each breast cancer stage 4.
Short video animation teaching tool highlighting key teaching points
BR196-ED- The Future Breast Imager: Understanding the Clinician's Perspective
TUB11
Station #11
Awards
Certificate of Merit
Participants
Hyung Won Choi, MD, Los Angeles, CA (Presenter) Nothing to Disclose
Irene S. Tsai, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Bo Li, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
As there is increasing advocacy for a multidisciplinary approach to provide higher quality of patient care, it is becoming more
pertinent for the radiologist to have an adequate understanding of current breast cancer treatment. The ideal breast radiologist will
be able to identify salient imaging findings that change clinical management. After reviewing this presentation, participants will be
familiar with: 1. Indications for surgical, radiation, and systemic treatment of breast cancer 2. Imaging findings that change the
surgeon, radiation oncologist, or oncologist's treatment plan.
TABLE OF CONTENTS/OUTLINE
1. Surgical management A. Indications for mastectomy versus lumpectomy with example cases B. Types of mastectomy: Radical,
modified radical, simple, and nipple sparing mastectomy C. Indications for full axillary dissection versus sentinel lymph node biopsy 2.
Radiation treatment A. Indications for radiation treatment with example cases B. Types of radiation treatment: Whole breast versus
partial breast irradiation 3. Systemic therapy A. Indications for neoadjuvant and adjuvant chemotherapy with example cases B.
Different types of systemic therapy - chemotherapy, hormonal blockade, and biologic therapies
BR197-ED- The Dark Side of the Post-Surgical Breast: Recurrence in Breast Cancer Follow-Up Care
TUB12
Station #12
Participants
Pedro Henrique Hasimoto e Souza, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tatiana C. Tucunduva, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana H. Catani, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Barbara H. Bresciani, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
pedrohenriquehs@hotmail.com
TEACHING POINTS
Prepare the radiologist to face common and uncommon presentations of recurrence in breast imaging; Exhibit clinical cases from our
radiology department, with breast malignant imaging findings in the post-surgical follow-up care, showing its pathologic correlation;
Discuss the effects of manipulation on breast cancer detection; Provide management recommendations.
TABLE OF CONTENTS/OUTLINE

Brief discussion on post-treatment follow-up care in breast cancer; Risk factors and incidence of recurrence; Imaging features
(mammography, ultrasound, computerized tomography and magnetic resonance imaging) of recurrence in post-treatment breast
imaging, with their pathologic correlation; Describe alarming signs that should raise suspicion for malignancy; Review differential
diagnosis for recurrent breast cancer; Management recommendations.
 VW 64

Hologic's Biopsy Solutions Portfolio - A Clinical Perspective: Hologic Vendor Workshop

Tuesday, Nov. 27 1:00PM - 2:00PM Room: Booth 5524

Participants
Harriet B. Borofsky, MD, San Mateo, CA (Presenter)
Program Information
Experience Hologic's portfolio of breast biopsy products to meet your full biopsy suite needs under the guidance of an experienced
radiologists and clinical education specialists. Affirm® Breast Biopsy Guidance System, Affirm® Prone Biopsy System, Brevera®
Breast Biopsy System and Viera™ Portable Breast Ultrasound)
Registration
https://hologicrsna.com
 VW 39

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Tuesday, Nov. 27 1:05PM - 2:15PM Room: Booth 5530

Participants
Ingolf Karst, MD, Chicago, IL (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 ML36

Machine Learning Theater: How AI Can Improve Diagnostic Performance and Reduce Reading Time in Breast
Tomosynthesis: Presented by iCAD
Tuesday, Nov. 27 1:30PM - 1:50PM Room: Machine Learning Show case North Hall

Participants
Senthil Periaswamy, PhD, Nashua, NH (Presenter) Vice President, iCAD, Inc
Program Information
iCAD will share how its innovative breast health AI solution, built on the latest deep learning technology, improves breast cancer
detection, reduces recalls and improves reading efficiency for digital breast tomosynthesis.
 VW 12

Are We Ready For Automated Breast Ultrasound (ABUS) Coronal View?: GE Vendor Workshop
Tuesday, Nov. 27 2:00PM - 2:30PM Room: Booth 8156

Participants
Simone Schiaffino, MD, Bogliasco, Italy (Presenter) Nothing to Disclose
Program Information
In this presentation, Dr. Schiaffino will discuss his study of 188 patients with dense breast, aimed to evaluate the value of
automated breast ultrasound (ABUS) reconstructed coronal view, comparing diagnostic performance and reading times to the
complete multiplanar assessment. Registration is required; adding this session to the RSNA calendar tool alone does not secure
your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 40

AI-based Mammography Reading: Hands-on session: Siemens Healthineers Vendor Workshop

Tuesday, Nov. 27 2:30PM - 3:40PM Room: Booth 5530

Participants
Nico Karssemeijer, PhD, Nijmegen, Netherlands (Presenter) Director and Shareholder, ScreenPoint Medical BV Shareholder, Volpara
Health Technologies Limited Consultant, QView Medical, Inc Shareholder, QView Medical, Inc
Program Information
You will learn about the use of the AI-supported Transpara™* system from Screenpoint in combination with syngo® Breast Care to
support 2D or 3D mammography reading. Transpara™ provides detection and decision support together with an overall exam based
score for prioritization of reading. You will experience a case review session to explore the use of artificial intelligence in
mammography reading. *Transpara™ is pending 510(k) clearance, and is not yet commercially available in the United States
 VW 65

Improving Tomosynthesis Read-times, While Maintaining Clinical Performance: Hologic Vendor Workshop
Tuesday, Nov. 27 2:30PM - 3:30PM Room: Booth 5524

Program Information
Discussion and case review focuses on Hologic's new innovations to improve workflow efficiency, without sacrificing clinical
outcomes. (3D Mammography™ Technology, Clarity HD, Smart Mapping, SecurView® Workstations)
Registration
https://hologicrsna.com
 SSJ01

Breast Imaging (Monitoring Response to Treatment)

Tuesday, Nov. 27 3:00PM - 4:00PM Room: E450A

BQ BR MR
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Wendy B. Demartini, MD, Stanford, CA (Moderator) Nothing to Disclose
Constance D. Lehman, MD,PhD, Boston, MA (Moderator) Research Grant, General Electric Company; Medical Advisory Board,
General Electric Company
Sub-Events
SSJ01-01 Complete Response on MR Imaging After Neoadjuvant Chemotherapy in Breast Cancer Patients:
Factors of Radiologic-Pathologic Discordance
Tuesday, Nov. 27 3:00PM - 3:10PM Room: E450A

Participants
Jinyoung Chang, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
PURPOSE
Although MR imaging may show radiologic complete response (rCR) after neoadjuvant chemotherapy (NAC), there may be
discordance between the MR imaging and pathology. The purpose of this study was to evaluate the radiologic and clinicopathologic
factors of discordance between rCR and pathologic complete response (pCR) after NAC in breast cancer patients, and to determine
whether there are differences among molecular subtypes.
METHOD AND MATERIALS
Our institutional review board approved this retrospective study. We evaluated 209 consecutive patients who showed rCR in MR
imaging after NAC between January 2013 and December 2015. All patients had mammography, ultrasound and MR imaging before
and after completion of NAC prior to definitive surgery. rCR was diagnosed when no enhancement or faint enhancement was shown
in the previous lesion site equal to that of the background normal breast tissue. pCR was defined as the complete absence of both
invasive cancer and ductal carcinoma in situ in the breast on the surgical histopathological examination. Clinicopathologic and initial
radiologic findings were assessed and factors affecting the radiologic-pathologic discordance were analyzed.
RESULTS
One hundred eight patients (51.7%) showed pCR and 101 (48.3%) had residual lesion on surgical histopathology. The false negative
findings were significantly more frequent in luminal A and B subtype (67.3%, 68/101), with radiologic findings such as larger tumor
size (p = 0.048) in mammography, irregular shape (p = 0.021), high proportion of persistent component (p = 0.008), and low
proportion of washout component (p = 0.001). On multivariate analysis of radiologic findings in all patients to predict residual lesion,
calcification in mammography (p = 0.037), multifocal lesion (p = 0.004), and nonmass enhancement in MR (p = 0.023) were
significantly associated with residual lesion.
CONCLUSION
Luminal subtype has a significant high false negative rate who achieved rCR after NAC. Patients with calcification in mammography,
multifocal lesion, and nonmass enhancement in initial MR imaging are significantly associated with residual lesion.
CLINICAL RELEVANCE/APPLICATION
Although MR imaging showed rCR after NAC, multifocal breast cancer with calcification and nonmass enhancement, residual lesions
should be considered and may impact surgical planning.
SSJ01-02 How Should a Radiologist Diagnose a Complete Imaging Response on Breast MRI after Neoadjuvant
Chemotherapy?
Tuesday, Nov. 27 3:10PM - 3:20PM Room: E450A

Participants
Shreena Shah, MD, New York, NY (Presenter) Nothing to Disclose
Mary C. Hughes, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Natsuko Onishi, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Junting Zheng, New York, NY (Abstract Co-Author) Nothing to Disclose
Marinela Capanu, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth J. Sutton, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

shb9167@nyp.org
PURPOSE
There is no standard definition regarding what constitutes a complete imaging response (CIR) on breast MRI post-neoadjuvant
chemotherapy (NAC). The purpose is to evaluate the accuracy of a CIR defined as no residual enhancement in the tumor bed post-
NAC in diagnosing a pathologic complete response (pCR).
METHOD AND MATERIALS
The IRB approved this HIPAA-compliant retrospective study and waived informed consent. Women were identified with biopsy
proven invasive breast cancer who underwent: a) NAC, b) pre and post-NAC MRI and c) surgery between 2014-2016. CIR was
defined as no residual enhancement on any of the three post-contrast phases on post-NAC MRI. Two breast radiologists (R1 and
R2) independently reviewed all pre and post-NAC MRI blinded to pathology. pCR was defined as no residual invasive or in situ
carcinoma. Measures of accuracy, including sensitivity, specificity, negative predictive value (NPV) and positive predictive value
(PPV) were estimated using no enhancement on MRI to diagnose pCR. Kappa statistic was used to assess agreement between
readers.
RESULTS
275 women were included with 280 breast cancers (n=5, 1.8% had bilateral cancers). Of the 280 breast cancers, 256 (91.4%)
were invasive ductal carcinoma, 11 (3.9%) were invasive lobular carcinoma and 13 (4.7%) were other invasive carcinoma. 74
(26.4%) had a pCR and 186 (66.4%) had no pCR. The two readers had substantial agreement on enhancement (kappa=0.627,
87.5% concordant readings). Sensitivity was 40.5%(R1)/50.0%(R2), specificity was 87.9% (R1)/86.9%(R2), PPV was 54.5%
(R1)/57.8% and NPV was 80.4%(R1)/82.9(R2). These measures of accuracy were not significantly different between different
subtypes defined as: ERPR+HER2-(n=99), ERPR+HER2+(n=61), ERPR+HER2equiv (n=1), ERPR-HER2+(n=44) and ERPR-HER2- (n=75).
CONCLUSION
CIR defined as no residual enhancement in the tumor bed on post-NAC breast MRI is not sensitive in diagnosing a pCR. Residual
enhancement can be seen with a pCR and a better definition of what constitutes a CIR is needed; our results suggest that it should
include some degree of low-level tumor bed enhancement.
CLINICAL RELEVANCE/APPLICATION

Complete imaging response defined as no residual enhancement in the tumor bed is not an accurate assessment of pathologic
complete response, which radiologists need to be aware of as management decisions based upon post-NAC MRI interpretation can
impact the decision to perform breast conservation versus mastectomy.
SSJ01-03 Predicting Axillary Response to Neoadjuvant Chemotherapy Using Breast MRI and US Based Model in
Patients with Clinically Node-Positive Breast Cancer
Tuesday, Nov. 27 3:20PM - 3:30PM Room: E450A

Participants
Jung Min Chang, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Research Grant, General Electric Company
Rihyeon Kim, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Soo-Yeon Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Nariya Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Kyung Moon, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

rihyeon.kim@gmail.com
PURPOSE

The axillary LN pathologic complete response (pCR) is increasingly common after neoadjuvant chemotherapy (NAC) in breast cancer
patients, for whom axillary LN dissection may be spared. The aim of our study was to develop a clinical model including MRI and US
to better predict axillary pCR after NAC in clinically node-positive breast cancer patients.
METHOD AND MATERIALS
An IRB-approved retrospective review was performed for 172 consecutive clinically node-positive breast cancer patients who were
treated with NAC and following surgery at our institution from 2016 to 2017. All the patients underwent breast MRI and US, before
and after completion of NAC to evaluate the primary tumor extent and axillary LN status. The multivariate logistic regression
analysis determined independent predictors of axillary pCR. With combination of those variables we developed a predictive model to
increase axillary pCR rate.
RESULTS

Among 172 patients the overall axillary pCR rate was 59%, with 75% (46/61) and 42% (30/71) in cN1 and cN2 patients,
respectively. Pretreatment characteristics of clinical N stage, axillary lymphadenopathy and tumor subtype were correlated with
axillary pCR (all, p<0.05). Post-NAC imaging features including mean tumor size, the size change rate on MRI and the axillary LN
cortical thickness on US were predictors of axillary pCR (all, p<0.05). From the multivariate logistic regression analysis, independent
variables were clinical N stage, tumor subtype, tumor size change rate on MRI and axillary lymphadenopathy on post-NAC US, with
an AUC of 0.884 (all, p<0.05). The axillary pCR rate of cN1 and non-luminal A model was 88% (38/43) and addition of absent axillary
lymphadenopathy improved pCR rate to 94% (33/35). In cN2 patients, the axillary pCR rate of non-luminal A and no axillary
lymphadenopathy model increased from 70% (24/34) to 91% (10/11) with additional condition of more than 80% size decrease on
MRI.
CONCLUSION

Clinical N stage, tumor subtype and MRI/US findings were predictors of axillary LN pCR. Our predictive model including clinico-
pathologic features and image findings could achieve high axillary pCR rate, which may guide LN dissection planning after NAC in
clinically node-positive breast cancer patients.
CLINICAL RELEVANCE/APPLICATION
Our imaging based predictive model can improve risk stratification which may avoid unnecessary axillary LN dissection in clinically
node-positive breast cancer patients after NAC.
SSJ01-04 Texture Analysis of Lymph Node MRI Characteristics Improves Prediction of Progression-Free
Survival Following Chemotherapy in Breast Cancer
Tuesday, Nov. 27 3:30PM - 3:40PM Room: E450A

Participants
Renee Cattell, BA, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Haifang Li, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
James J. Kang, BA, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Pauline B. Huang, Douglaston, NY (Abstract Co-Author) Nothing to Disclose
Thomas Ren, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Jules A. Cohen, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Paul R. Fisher, MD, East Setauket, NY (Abstract Co-Author) Research Grant, Siemens AG;
Cliff Bernstein, MD, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Sean Clouston, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Tim Duong, PhD, Stony Brook, NY (Presenter) Nothing to Disclose
PURPOSE
Pathological complete response and reduction in lesion volume (LV) after chemotherapy do not guarantee progression-free survival
(PFS). About 1/3 of patients with pathological complete response had relapse at 5 years in our cohort. Breast cancer metastasizes
through the lymphatic system. Axillary lymph node (LN) is clinically evaluated by palpation and using ultrasound. This study
investigated the efficacy the texture analysis of the LN MRI in predicting PFS.
METHOD AND MATERIALS

Data were obtained from I-SPY 1 trial of breast cancer patients undergoing chemotherapy. We extracted 48 textures for pre- (T1),
early- (T2), during (T3) and post-chemo (T4) time points for a subset of patients (N=41) in whom LN MRI were usable. Area under
the receiver-operating curve (AUC) was calculated, with PFS at 5-year as a reference variable. Multivariable generalized linear
modeling was used to estimate model fit and assign a risk score at each time point based on top-ranking features.
RESULTS

For early (T1) in treatment, LV along with LN histogram skewness and grey-level run length matrix long-run emphasis were top
predictors of PFS (AUC=0.68). LV became less predictive in later time points. LN features surpassed LV in post-chemo (T4) time
point; the top features were neighborhood grey-level matrix coarseness and grey-level zone length matrix zone length non-
uniformity. When changes in texture features over time were analyzed, LN features outperformed LV. Top early (T2-T1) predictors
were change in compacity, gray-level non-uniformity, and volume of the LN. Top late (T4-T1) predictors were change in zone
length non-uniformity and coarseness. The combination of risk scores across all time points resulted in a model with an AUC of 0.89
(sensitivity:91%, specificity:95%, p= 0.004).
CONCLUSION
Texture analysis of lymph node MRI improves prediction of PFS. The multivariable risk prediction model identified key characteristics
and could provide strong predictors of relapse.
CLINICAL RELEVANCE/APPLICATION

Texture analysis of axillary lymph node MRI has the potential to accurately predict 5-year progression free survival.
SSJ01-05 Breast 3D Magnetic Resonance Fingerprinting Relaxometry: Utility in Measuring Early Response to
Neo-Adjuvant Chemotherapy in Breast Cancer
Tuesday, Nov. 27 3:40PM - 3:50PM Room: E450A

Participants
Ananya Panda, MD,MBBS, Rochester, MN (Presenter) Support, Siemens AG
Yong Chen, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
Verena Obmann, MD, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Satyam Ghodasara, MD, Cleveland, OH (Abstract Co-Author) Research support, Siemens AG
Marcie Stopchinski, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Paula Silverman, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Megan Miller, Clevleand, OH (Abstract Co-Author) Nothing to Disclose
Jill Dietz, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Robert Shenk, Cleveland, OH (Abstract Co-Author) Nothing to Disclose
Katherine Wright, Cleveland, OH (Abstract Co-Author) Research support, Siemens AG
Nicole Seiberlich, PhD, Cleveland, OH (Abstract Co-Author) Research Grant, Siemens AG
Mark A. Griswold, PhD, Cleveland, OH (Abstract Co-Author) Research support, Siemens AG Royalties, Siemens AG Royalties, General
Electric Company Royalties, Bruker Corporation Contract, Siemens AG
Donna M. Plecha, MD, Strongsville, OH (Abstract Co-Author) Research Grant, Hologic, Inc
Vikas Gulani, MD, PhD, Cleveland, OH (Abstract Co-Author) Research support, Siemens AG; Licensed Technology, Siemens
Healthineers - both myself and my spouse. MR Fingerprinting, on which we are both inventors, has been licensed by Siemens.
For information about this presentation, contact:
drananyapanda@gmail.com
PURPOSE

To assess (1) Repeatability of breast 3D MRF-based relaxometry and (2) preliminarily assess utility in measuring early response to
neo-adjuvant chemotherapy in breast cancer
METHOD AND MATERIALS

In this IRB approved pilot study,12 healthy pre-menopausal volunteers were scanned with 3D MRF twice within 7-15 days for
repeatability analysis.In 5 volunteers,same-day test-retest scans were also performed with repositioning after a 10-minute scan
interval.For breast cancer evaluation,13 women with biopsy-proven invasive ductal carcinoma underwent baseline 3D MRF scans
before chemotherapy.So far,7 women have undergone repeat 3D MRF 7-10 days after first cycle of chemotherapy.All scans were
performed at 3T(Siemens Verio) using 8-channel breast coil.A 3D FISP-based MRF sequence with fat suppression was used.Overall
acquisition time was 5.5min.For repeatability analysis,ROIs were drawn on a 3D MRF partition containing the largest area of normal
breast tissue by one radiologist(8 years radiology experience) in both breasts.In breast cancer patients,ROIs were drawn on
partitions showing tumor and the opposite normal breast.Mean T1 & T2 were estimated for all scans.Treatment response was based
on either final pathology staging after surgery or RECIST criteria and T1 & T2 changes were compared in responders & non-
responders
RESULTS
Same-visit test-retest within-subject coefficient of variation(wCV) was <5% for T1 and <6.5% for T2.Two-visit wCV was <6% for
T1 and <5% for T2,establishing a measurable effect size.In breast cancer,tumor T1 & T2 were longer than normal
breast(mean±SD,Tumor T1:1175±203 ms,Normal T1:846±388 ms,Tumor T2:72±13 ms,Normal T2:48±12 ms,p=0.030 for T1,p<0.001
for T2).After one cycle of chemotherapy,responders(n=4) showed greater decrease in T1 & T2(p=0.026 for T1,p=0.033 for T2)
while patients with stable/progressive disease(n=3) showed negligible changes in T1 & T2
CONCLUSION

Breast 3D MRF relaxometry was shown to be repeatable.Baseline tumor T1 & T2 were longer than normal breast.Treatment
responders showed larger change in T1 & T2 compared to non-responders.These preliminary results suggest that breast 3D MRF
may be useful in quantitatively predicting early response to chemotherapy
CLINICAL RELEVANCE/APPLICATION

Breast 3D MRF relaxometry is a repeatable technique for longitudinal studies in patients.Significant decrease in tumor T1 and T2
after one cycle of chemotherapy may predict treatment response
SSJ01-06 Are the Baseline Imaging Characteristics of Breast Cancer Associated With Reponses to Neoadjuvant
Chemotherapy?
Tuesday, Nov. 27 3:50PM - 4:00PM Room: E450A

Participants
Andrew Evans, MRCP, FRCR, Dundee, United Kingdom (Presenter) Nothing to Disclose
Colin Purdie, MBChB,PhD, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
Sarah Savaridas, FRCR,MBChB, Hexham, United Kingdom (Abstract Co-Author) Nothing to Disclose
Yee Ting Sim, MBBCh, FRCR, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
Jane Macaskill, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
Sarah J. Vinnicombe, MRCP, FRCR, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

a.z.evans@dundee.ac.uk
PURPOSE
Predictors of response to neoadjuvant chemotherapy (NACT) within subgroups are poor and not commonly used. We aimed to
establish if baseline mammographic (MGM) and ultrasound (US) features of breast cancer and breast cancer subtypes are
associated with rates of pathological complete response (pCR) after NACT.
METHOD AND MATERIALS

The baseline imaging of 210 consecutive women who underwent NACT were reviewed by an experienced breast radiologist blinded
to outcomes. The MGM features assessed were the presence or absence of spiculation and calcification. The US features
documented were echogenicity, distal effect, the presence of a circumscribed border and the US lesion size. The relationships
between baseline imaging features and pCR after NACT were documented. The findings according to sub-group (HER2+ve, triple
negative and ER+ve HER2-ve) were also assessed. The significance of differences were analysed using the chi-square test and ROC
curves.
RESULTS
Of the 210 patients, 46(22%) had a pCR while 164 did not. The cohort consisted of women with 82 triple negative (21 pCR's), 75
HER2+ve (21 pCR's) and 53 ER+ve HER2-ve (4 pCR's) cancers. For the complete cohort the features significantly associated with
pCR were the absence of MGM spiculation (4 of 58 (7%) vs. 42 of 151(28%), p=0.001), the absence of distal shadowing on US (8
of 73(11%) vs. 38 of 136(28%), p=0.015) and small US size (AUC 0.62, P=0.02). In HER2+ve patients all the above associations
were confirmed and in addition the presence of a circumscribed border on US was associated with high pCR rates (7 of 14(50%) vs.
14 of 60(23%), p=0.04). None of the above associations held true for triple negative cancers. The number of pCR's in the ER+HER2-
ve group was to small to allow analysis.
CONCLUSION
The baseline imaging features of HER2+ve breast cancer are strongly associated with the chance of achieving a pCR following
NACT. This is not so for triple negative cancers.
CLINICAL RELEVANCE/APPLICATION
This data could be helpful when discussing NACT and associated surgery with patients who have HER+ve breast cancer.
 SSJ02

Breast Imaging (US Advanced Applications)

Tuesday, Nov. 27 3:00PM - 4:00PM Room: E353C

BR MR MI US
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Jung Min Chang, MD, Seoul, Korea, Republic Of (Moderator) Research Grant, General Electric Company
Wei T. Yang, MD, Houston, TX (Moderator) Consultant, General Electric Company; Medical Advisory Board, Seno Medical
Instruments, Inc
Sub-Events
SSJ02-01 Can Baseline Ultrasound and Mammographic Features Help Predict Metastasis Free Survival in
Patients Receiving Neoadjuvant Chemotherapy?
Tuesday, Nov. 27 3:00PM - 3:10PM Room: E353C

Participants
Sarah Savaridas, FRCR,MBChB, Hexham, United Kingdom (Presenter) Nothing to Disclose
Sarah J. Vinnicombe, MRCP, FRCR, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
Colin Purdie, MBChB,PhD, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
Andrew Evans, MRCP, FRCR, Dundee, United Kingdom (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
s.savaridas@nhs.net
PURPOSE
To determine if baseline mammographic and ultrasound features in patients receiving neoadjuvant chemotherapy (NACT) are
associated with metastasis free survival (MFS).
METHOD AND MATERIALS
Informed consent for the study was obtained from consecutive women receiving NACT. All participants were metastasis free at
diagnosis. Baseline images were retrospectively reviewed by a breast radiologist blinded to outcomes. Ultrasound (US) features
documented included echo pattern, posterior effect, circumscribed margins, skin changes (invasion or thickening), lesion size and
shear wave stiffness (mean). Mammogram (MMG) features documented were the presence or absence of spiculation and
microcalcification. Tumour immunophenotype was assessed on core biopsy. Statistical significance was assessed using chi-square
and ROC analysis.
RESULTS
134 breast patients were included; 52 had triple negative cancer, 48 were HER2+ve and 34 had ER+ve, HER2-ve disease. During
the follow-up period (mean 4.3yrs) 41 women (31%) developed metastases. Across the whole cohort, skin involvement on baseline
US was the only feature associated with metastasis development. MFS was 56% (23/41) vs 75% (70/93) for those with and
without skin involvement respectively, p=0.03. US lesion size was not associated with MFS.In the HER2 positive subgroup
mammographic calcification was associated with poorer MFS (12/24 (50%) vs 20/23 (87%), p=0.006). The presence of posterior
shadowing on US was also associated with poorer MFS (11/22 (50%) vs 21/26 (81%), p=0.02) in this subgroup.No baseline imaging
features were shown to be associated with MFS for triple negative and ER+ve HER-ve cancers.
CONCLUSION
We have demonstrated that baseline imaging characteristics are associated with MFS in patients treated with NACT particularly in
the HER2 +ve subgroup. This prognostic information, which is available prior to treatment could aid patient treatment selection and
counselling.
CLINICAL RELEVANCE/APPLICATION
Baseline imaging characteristics are associated with MFS in patients treated with NACT. This prognostic information, available prior
to treatment, may aid patient treatment selection and counselling.
SSJ02-02 Combination of Different Types of Elastography in Downgrading Ultrasound Breast Imaging-
Reporting and Data System Category 4a Breast Lesions
Tuesday, Nov. 27 3:10PM - 3:20PM Room: E353C

Participants
Xueyi Zheng, Guangzhou, China (Presenter) Nothing to Disclose
Yini Huang, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Yun Wang, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Fei Li, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jing Han, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jianwei Wang, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jianhua Zhou, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

zhengxy1@sysucc.org.cn
PURPOSE
When using single type of elastography to downgrade ultrasound (US) Breast Imaging-Reporting and Data System (BI-RADS)
category 4a lesions, some cancers would be missed. This study purposed to determine whether combination of different types of
elastography could improve the accuracy of elastography aided downgrading BI-RADS category 4a lesions and reduce unnecessary
biopsies.
METHOD AND MATERIALS
For this prospective institutional review board-approved study, verbal informed consent was obtained from all patients. From
January 2016 to February 2018, 329 consecutive women with 347 US BI-RADS category 4a breast lesions were enrolled in the
study. These lesions, prior to biopsy, were subject to conventional US supplemented by elastography assessments, including strain
elastography of elasticity imaging (EI), virtual touch tissue imaging (VTI) and Virtual Touch IQ (VTIQ). The diagnostic performances
were calculated for BI-RADS category, EI, VTI and VTIQ, and the combination among EI, VTI and VTIQ (combined EI and VTI
[EI+VTI], combined EI and VTIQ [EI+VTIQ] and combined VTI and VTIQ [VTI+VTIQ]).
RESULTS
Pathologically, 313 lesions (90.2%) were benign and 34 (9.8%) were malignant. The cut-off values were EI score>3, VTI score>3
and shear wave speed (SWS) on VTIQ>3.29 m/s, respectively. For EI, VTI and VTIQ alone, the specificity were significantly higher
than that of BI-RADS (P <0.001), while the sensitivity were significantly lower than that of BI-RADS (76.5%, 70.6%, 67.6% vs.
100%, respectively, P <0.05). Among the combinations of different types of elastography, EI+VTI yielded the highest AUROC of
0.800 and negative predictive value of 99.5%. The sensitivity of EI+VTI was significantly increased as compared with single type
elastography (P <0.05). There was no significant difference in the sensitivity between EI+VTI and BI-RADS (97.1% vs 100%, P
=1.000) while the specificity was significantly higher than that of BI-RADS (P <0.001). When using EI+VTI to downgrade the
lesions, 57.1% of the lesions would be downgraded and 99.5% of these lesions were benign.
CONCLUSION

Combinations of EI and VTI to downgrade BI-RADS category 4a lesions increased the sensitivity and reduced the misdiagnosis of
breast cancers.
CLINICAL RELEVANCE/APPLICATION

Combination of different types of elastography provides a high sensitive way to downgrade BI-RADS category 4a lesions, potentially
applied in clinical practice without increasing misdiagnosed cancers.
SSJ02-03 Correlation Between Apparent Diffusion Coefficient Values and Ultrasound Elasticity kPa Values in
Breast Cancers
Tuesday, Nov. 27 3:20PM - 3:30PM Room: E353C

Participants
Seon Hyeong Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jihee Park, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Inyoung Youn, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Yoonjung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Juhee Moon, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Shin Ho Kook, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
seonhyeong.choi@samsung.com
PURPOSE
DWI reflects tumor cellularity and integrity of cell membranes. Breast ultrasound elastography is a method of imaging tissue
stiffness and the shear-wave elastography (SWE) allows measurement of the propagation speed of shear waves within the tissue
to locally quantify its stiffness in kilopascals (kPa) or meters per second (m/sec). This study was performed to evaluate the
correlation between breast apparent diffusion coefficient (ADC) values and ultrasound Shear wave elastography kPa values in
biopsy proven breast cancers.
METHOD AND MATERIALS

From January 2016 to November 2017, 121 patients who have breast cancer confirmed by US-guided biopsy underwent both pre-
operative breast diffusion MRI and breast SWE. Among these patients we included only who underwent examination by same
operator to reduce inter-operator variability. Finally, this study included 94 breast cancer patients. The investigated factors
included the ADC values, mean kPa values (Emean), maximum kPa values (Emax), pathology, size of tumor, associated
calcifications, ER/PR/HER2 status, molecular subtypes, Ki-67 index, mammographic density and BI-RADS US assessment category.
The results were analyzed using the statistical software SPSS for Windows (version 24). The correlation analysis was used to study
correlation between ADC values, Emean/Emax, size and Ki-67. And independent samples of t-test and ANOVA including post hoc
test were performed to evaluate for the above variables.
RESULTS

There was no correlation between ADC values and Emean (p=0.791)/Emax (p=0.634)[Fig1]. However, Emean (p=0.001), Emax
(p<0.001), and Ki-67 index value (p=0.010) were significantly correlated with the size of tumor. High Ki-67 index group showed
statistically significant lower ADC values (p=0.034) and higher Emean (p=0.064)/Emax value (p=0.065) without statistical
insignificance. The Emean (p=0.002)/Emax (p=0.001) were correlation with T stage but ADC (p=0.813) was not. Howevre, the
Eman (p=0.000)/Emax (p=0.000) and ADC (p=0.017) were also correlated with US BIRADS categories.
CONCLUSION
The ADC values and Emean/Emax in breast cancers were nor correlated each other. However, the size of tumor, Ki67 index, BI-
RADS assessment category affected the ADC and Emean/Emax values independently.
CLINICAL RELEVANCE/APPLICATION

The size of breast cancer, Ki67 index, BI-RADS assessment category affected the ADC and Emean/Emax values independently.
SSJ02-04 Breast Cancer Staging: Combined Digital Breast Tomosynthesis and Automated Breast Ultrasound
versus Magnetic Resonance Imaging
Tuesday, Nov. 27 3:30PM - 3:40PM Room: E353C

Participants
Rossano Girometti, MD, Udine, Italy (Presenter) Nothing to Disclose
Ludmila Tomkova, MD, Udine, Italy (Abstract Co-Author) Nothing to Disclose
Lorenzo Cereser, MD, Udine, Italy (Abstract Co-Author) Nothing to Disclose
Chiara Zuiani, MD, Udine, Italy (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
rgirometti@sirm.org
PURPOSE
To investigate whether combined Digital breast tomosynthesis and Automated breast volume scanner (DBT-ABVS) are comparable
to Magnetic resonance imaging (MRI) in staging breast cancer.
METHOD AND MATERIALS

We retrospectively included seventy-three patients with histologically proven breast cancer who underwent preoperative DBT,
ABVS and 1.5T MRI in the period July 2015-July 2016. Two radiologists in consensus recorded the number, site and BI-RADS
category of breast findings during two independent reading strategies, i.e. DBT-ABVS vs. MRI. Using histology or 1-year follow up
as the standard of reference, we calculated the accuracy for cancer of both imaging strategies. Bland-Altman analysis was used to
evaluate the agreement between MRI vs. DBT or ABVS in cancer size assessment.
RESULTS
Patients showed a total of 160 lesions (108 malignant and 52 benign). Malignant lesions were monofocal, multifocal, multicentric
and biltateral in 53, 15, 4 and 1 cases, respectively. Diagnostic accuracy of DBT-ABVS vs. MRI was comparable for all cancers
(90.0% vs. 93.8%, respectively), though DBT-ABVS showed lower sensitivity and positive predictive values for additional disease
(76.5% vs. 91.7%, and 78.8% vs 93.4%, respectively). Compared to MRI, ABVS+DBT missed 6 lesions, including two invasive
cancers and one extensive intravascular invasion associated to ductal carcinoma in situ. Bland-Altman analysis showed ABVS to
agree with MRI at a higher extent than DBT in assessing cancer size.
CONCLUSION

DBT-ABVS is less accurate than MRI in staging additional disease.

CLINICAL RELEVANCE/APPLICATION
Though less performing than MRI, DBT-ABVS showed acceptable diagnostic accuracy in staging breast cancer. This strategy might
be used if MRI is unavailable or unfeasible.
SSJ02-05 Quantitative Elastic Heterogeneity as a Potential Noninvasive Marker of Lymphovascular Invasion in
Breast Cancer
Tuesday, Nov. 27 3:40PM - 3:50PM Room: E353C

Participants
Yini Huang, MD, Guangzhou, China (Presenter) Nothing to Disclose
Fei Li, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Chuan Peng, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Xueyi Zheng, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Yun Wang, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jing Han, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jianwei Wang, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Qing Li, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Jianhua Zhou, MD, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
huangyn@sysucc.org.cn
PURPOSE

To evaluate the association between quantitative elastic heterogeneity (EH) and lymphovascular invasion (LVI) in breast cancers.
METHOD AND MATERIALS
This retrospective study consisted of 97 patients with breast cancers that had undergone shear wave elastography (SWE) with
virtual touch tissue imaging quantification (VTIQ) between August 2015 and August 2017. Three region of interests (ROIs) were
placed over the highest stiffness and the lowest stiffness areas of each lesion to measure shear wave velocity (SWV) and EH was
determined as the difference between the averaged highest SWV and lowest SWV. Classical prognostic factors including lesion size,
histopathological type and grade, subtype [luminal A, luminal B (HER2-), luminal B (HER2+), HER2 enriched and basal-like], and
axillary lymph node (LN) status were reviewed and their correlation with EH values were stratified by the presence or absence of
LVI. The diagnostic performance of EH in distinguishing LVI or not were analyzed in lesions smaller than 2 cm.
RESULTS

Tumors with LVI showed significantly higher EH values when compared to tumors without LVI (adjusted P < 0.001), regardless of
the tumor size, histological grade and type, and LN status. Lymphovascular invasion (adjusted P < 0.001), large tumor size
(adjusted P = 0.011) and lymph node involvement (adjusted P = 0.046) showed statistically positive association with high EH
values. In breast cancers smaller than 2 cm, tumors with LVI (4.31±1.16 m/s) showed significantly higher EH values when compared
to tumors without LVI (2.99±1.18 m/s) (adjusted P < 0.001). Using EH higher than 3.66 m/s to suggest LVI, the area under the
receiver operating characteristic curve was 0.796, and the sensitivity, specificity, positive predictive value and negative predictive
value were 78 % (14/18), 75 % (39/52), 52 % (14/27) and 91 % (39/43), respectively.
CONCLUSION
EH could be served as a potential marker to assess LVI status on preoperative imaging, especially for breast cancer less than 2 cm
in size.
CLINICAL RELEVANCE/APPLICATION

Quantitative elastic heterogeneity of breast cancer can be used as a non-invasive marker for preoperative evaluation of LVI,
thereby guiding clinical regulation and predicting prognosis.
SSJ02-06 Multiparametric Quantitative Ultrasound of the Breast Can Improve the Diagnostic Performance of the
Radiologist
Tuesday, Nov. 27 3:50PM - 4:00PM Room: E353C

Participants
Panagiotis Kapetas, Vienna, Austria (Presenter) Nothing to Disclose
Paola Clauser, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Ramona Woitek, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Georg J. Wengert, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Mathias Lazar, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Maria Bernathova, MD, Wien, Austria (Abstract Co-Author) Nothing to Disclose
Katja Pinker-Domenig, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
Thomas H. Helbich, MD, Vienna, Austria (Abstract Co-Author) Research Grant, Medicor, Inc Research Grant, Siemens AG Research
Grant, C. R. Bard, Inc
Pascal A. Baltzer, MD, Vienna, Austria (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

panagiotis.kapetas@meduniwien.ac.at
PURPOSE

To evaluate quantitative multiparametric ultrasound (mpUS) of the breast using B-mode US, elastography, Doppler and contrast
enhanced US (CEUS) in different combinations of 2, 3 or 4 parameters for the differentiation of benign and malignant lesions and
investigate a possible variation according to the degree of experience of the examiner.
METHOD AND MATERIALS

124 patients, each with one biopsy-proven, sonographically evident breast lesion were included in this prospective, IRB-approved
study. Each lesion was examined with B-mode US, elastography (Virtual Touch IQ-VTIQ), Doppler US and CEUS. Different
quantitative parameters were recorded for each modality: Shear Wave Velocity (SWV) for VTIQ, Pulsatility (PI) and Resistive Index
(RI) for Doppler US. For CEUS, 11 different parameters were calculated using a dedicated software. 4 readers (2 experienced breast
radiologists and 2 radiology residents) independently evaluated B-mode images of each lesion and assigned a BI-RADS score to it.
Using ROC curve analysis, the quantitative parameter with the best diagnostic performance for each modality was chosen and cut-
off values were calculated. Using these, all quantitative results were dichotomized. The BI-RADS scores of all readers were then
combined with the quantitative parameters. Descriptive statistics were used to evaluate the diagnostic performance of mpUS.
Histology served as the reference standard.
RESULTS
59 lesions were benign and 65 malignant. SWV, RI and mean transit time showed the highest diagnostic performance. MpUS with
three parameters (B-mode, VTIQ and CEUS) showed the highest diagnostic performance irrespective of the experience level of the
readers (averaged AUC 0.812 vs. 0.683 for B-mode US, p-value 0.0001), while the combination of B mode, VTIQ and Doppler US the
second best (averaged AUC 0.789, p-value 0.0001). All other combinations (with 2, 3 or 4 parameters) showed a lower AUC. MpUS
with B-mode, VTIQ and CEUS was able to significantly reduce the number of false positive biopsy recommendations (p<0.0001).
CONCLUSION

Quantitative breast mpUS with three parameters (B-mode US, VTIQ elastography and CEUS) significantly improves the diagnostic
performance of B-mode US alone, irrespective of the experience level of the examiner.
CLINICAL RELEVANCE/APPLICATION

MpUS of the breast offers quantitative parameters that may be used as imaging biomarkers for the differentiation of benign from
malignant breast lesions.
 SSJ17

Science Session with Keynote: Nuclear Medicine (Chest/Breast Oncology Nuclear Imaging)
Tuesday, Nov. 27 3:00PM - 4:00PM Room: S504CD

BR CH CT MR NM OI
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Peter S. Conti, MD, PhD, Los Angeles, CA (Moderator) Nothing to Disclose
Andrew C. Homb, MD, Rochester, MN (Moderator) Nothing to Disclose
Sub-Events
SSJ17-01 Nuclear Medicine Keynote Speaker: Radiomics in Lung Cancer
Tuesday, Nov. 27 3:00PM - 3:10PM Room: S504CD

Participants
Lawrence H. Schwartz, MD, New York, NY (Presenter) Committee member, Celgene Corporation Committee member, Novartis AG
Committee member, ICON plc Committee member, BioClinica, Inc
SSJ17-02 An Updated and Validated PET/CT Volumetric Prognostic Index for Non-Small Cell Lung Cancer
Tuesday, Nov. 27 3:10PM - 3:20PM Room: S504CD

Participants
Joshua H. Finkle, MD, Chicago, IL (Presenter) Nothing to Disclose
Bill C. Penney, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Yonglin Pu, MD, PhD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
PURPOSE
Whole-body metabolic tumor volume (MTVWB) and TNM staging are independent prognostic factors for overall survival (OS) in non-
small cell lung cancer (NSCLC). We aimed to update and validate the PET/CT volumetric prognostic index (PVP index) using the new
8th edition TNM staging system to evaluate its prognostic power versus TNM staging and MTVWB alone.
METHOD AND MATERIALS
This study was a retrospective analysis of 949 non-small cell lung cancer (NSCLC) patients diagnosed between 2004 and 2014.
Clinical TNM stage, MTVWB, age and gender, tumor histology type at the initial staging PET/CT exam, as well as treatment history
and long-term survival data were obtained. Patients were randomly assigned to modeling or validation group. Univariate and
multivariate Cox regression analyses were performed to compare PVP index, TNM stage, and MTVWB in the validation group.
RESULTS
The updated PVP index included the 3 variables TNM stage, and MTVWB and age. Univariate Cox models showed significant
association of PVP index with overall survival (OS) in patients with NSCLC (with Hazard ratio HR= 2.88 in the validation group,
p<0.001). The C-statistic of the PVP index (C-statistic = 0.71 in the validation group) was significantly greater than that of 8th
edition TNM staging (C-statistic = 0.68, p=0.029 ), MTVWB (C-statistic = 0.68, p=0.001), and patient age (C-statistic = 0.53,
p<0.001). Multivariate Cox regression analyses demonstrated significant association of PVP index with OS (with HR= 2.80, p<0.001)
after adjusting patient's gender and tumor histology.
CONCLUSION
The updated PVP index provides a quantitative risk assessment for NSCLC patients using 8th edition TNM staging, MTVWB, and age.
The index provides a simple and practical way for the care team to incorporate the independent prognostic value of both the TNM
stage and MTVWB. This approach can further improve the accuracy of overall survival prognosis.
CLINICAL RELEVANCE/APPLICATION
The PVP index combines the prognostic power of the TNM stage, whole-body metabolic tumor volume and age, offering prognostic
accuracy superior to whole-body metabolic tumor volume or TNM stage alone.
SSJ17-03 Prospective Comparison of 18F-FDG PET/MRI and 18F-FDG PET/CT for Thoracic Staging of Non-
Small Cell Lung Cancer
Tuesday, Nov. 27 3:20PM - 3:30PM Room: S504CD

Participants
Lino Sawicki, MD, Dusseldorf, Germany (Abstract Co-Author) Nothing to Disclose
Julian Kirchner, Dusseldorf, Germany (Presenter) Nothing to Disclose
Benedikt M. Schaarschmidt, MD, Essen, Germany (Abstract Co-Author) Stockholder, Bayer AG; Stockholder, General Electric
Company; Stockholder, Siemens AG; Stockholder, Teva Pharmaceutical Industries Ltd
Ken Herrmann, Essen, Germany (Abstract Co-Author) Co-founder, SurgicEye GmbH; Stockholder, SurgicEye GmbH; Consultant, Sofie
Biosciences; Consultant, Ipsen SA; Consultant, Siemens AG; Research Grant, Advanced Accelerator Applications SA; Research
Grant, Ipsen SA
Gerald Antoch, MD, Duesseldorf, Germany (Abstract Co-Author) Nothing to Disclose
Philipp Heusch, MD, Duesseldorf, Germany (Abstract Co-Author) Nothing to Disclose
PURPOSE
To compare the diagnostic performance of 18F-FDG PET/MRI and 18F-FDG PET/CT for primary and locoregional lymph node staging
in non-small cell lung cancer (NSCLC).
METHOD AND MATERIALS
In this prospective study a total of 84 patients (51 men, 33 women, mean age 62.5 ± 9.1 years) with histopathologically confirmed
NSCLC underwent 18F-FDG PET/CT followed by 18F-FDG PET/MRI in a single injection protocol. Two readers independently
assessed T and N staging in separate sessions according to the seventh edition of the American Joint Committee on Cancer staging
manual for 18FFDG PET/CT and 18F-FDG PET/MRI, respectively. Histopathology as reference standard was available for N staging in
all 84 patients and for T staging in 39 patients. Differences in staging accuracy were assessed by McNemars chi2 test. The
maximum standardized uptake value (SUVmax) and longitudinal diameters of primary tumors were correlated using Pearson's
coefficients.
RESULTS
T stage was categorized concordantly in 18F-FDG PET/MRI and 18F-FDG PET/CT in 38 of 39 (97.4%) patients. Herein, 18F-FDG
PET/CT and 18F-FDG PET/MRI correctly determined the T-stage in 92.3% and 89.7% of patients, respectively. N-stage was
categorized concordantly in 83 of 84 patients (98.8%). 18F-FDG PET/CT correctly determined the N stage in 78 of 84 patients
(92.9%), while 18F-FDG PET/MRI correctly determined the N stage in 77 of 84 patients (91.7%). Differences between 18F-FDG
PET/CT and 18F-FDG PET/MRI in T and N staging accuracy were not statistically significant (p > 0.5, each). Tumor size and
SUVmax measurements derived from both imaging modalities exhibited excellent correlation (r=0.963 and r=0.901, respectively).
CONCLUSION
18F-FDG PET/MRI and 18F-FDG PET/CT showed an equivalently high diagnostic performance for T and N staging in patients
suffering from NSCLC.
CLINICAL RELEVANCE/APPLICATION
PET/MRI as a dose-saving alternative to PET/CT proved coequal to the current gold standard for thoracic staging of NSCLC. Thus,
clinicians might use PET/MRI instead of PET/CT for this purpose. However, considering the longer examination times and higher
expenses of PET/MRI, a general recommendation in favor of PET/MRI cannot be drawn from this study.
SSJ17-04 The Relationship Between PET/CT Imaging Features and Pathological Types and Gene Mutations of
Primary Lung Cancer: A Study of 213 Untreated Lung Cancer Patients with Bone Metastases
Tuesday, Nov. 27 3:30PM - 3:40PM Room: S504CD

Participants
Xiaomeng Li, MD, Beijing, China (Presenter) Nothing to Disclose
Ning Wu, MD, Beijing, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
dora1327@163.com
PURPOSE
To evaluate the relationship between 18F-FDG PET/CT image characteristics and pathological types and gene mutations of primary
lung cancer in untreated lung cancer patients with bone metastases.
METHOD AND MATERIALS
A total of 213 untreated lung cancer patients with bone metastases were enrolled in this study. All patients underwent 18F-FDG
PET/CT examination, pathological and gene mutation examination of primary lung cancer. Spearman's correlation test was performed
to evaluate the association between primary tumors and bone metastases. Single factor analysis of variance was used to compare
groups.
RESULTS

(1)A total of 213 cases were evaluated. The mean SUVmax of primary lung cancer was 7.9±4.7; that of bone metastases was
8.2±4.3. The SUVmax of primary lesions had a significantly positive correlation with the SUVmax of bone metastases (r = 0.622; p =
0.000). Osteolytic metastasis was the most common type. (2)The SUVmax of primary lung lesions with different pathological types
were statistically different (all P = 0.000): squamous cell carcinoma > small cell carcinoma > adenocarcinoma. Their SUVmax were
11.7±3.7, 9.3±3.1, and 6.7±4.6, respectively. (3)In non-small cell lung cancer (NSCLC), the gene mutation rates of epidermal
growth factor receptor (EGFR), K-ras and anaplastic lymphoma kinase (ALK) were 35.7%, 10.1% and 3.8%, respectively. There was
no statistical difference in SUVmax of primary lung cancer between gene mutation type and wild type (P>0.05).
CONCLUSION

The SUVmax of primary lung lesions with different pathological types were statistically different. Squamous cell carcinoma was the
highest, and adenocarcinoma was the lowest. The SUVmax of primary lung cancer had a significantly positive correlation with the
SUVmax of bone metastases. In NSCLC, the mutation rate of EGFR is the highest. There was no statistical difference in SUVmax of
primary lung cancer between gene mutation type and wild type.
CLINICAL RELEVANCE/APPLICATION
The SUVmax of primary lung cancer is suggestive of its pathological type. But the SUVmax of primary lung cancer is not helpful to
predict the gene mutations in NSCLC.
SSJ17-05 Local and Whole-Body Staging in Patients with Primary Breast Cancer: A Comparison of One-Step to
Two-Step-Staging-Algorithms Utilizing PET/MRI
Tuesday, Nov. 27 3:40PM - 3:50PM Room: S504CD

Participants
Lale Umutlu, MD, Essen, Germany (Abstract Co-Author) Consultant, Bayer AG
Johannes Grueneisen, Essen, Germany (Abstract Co-Author) Nothing to Disclose
Mark Oehmigen, Essen, Germany (Abstract Co-Author) Nothing to Disclose
Harald H. Quick, PhD, Essen, Germany (Abstract Co-Author) Nothing to Disclose
Ann-Kathrin Bittner, Essen, Germany (Abstract Co-Author) Nothing to Disclose
Philipp Heusch, MD, Duesseldorf, Germany (Abstract Co-Author) Nothing to Disclose
Ken Herrmann, Essen, Germany (Abstract Co-Author) Co-founder, SurgicEye GmbH; Stockholder, SurgicEye GmbH; Consultant, Sofie
Biosciences; Consultant, Ipsen SA; Consultant, Siemens AG; Research Grant, Advanced Accelerator Applications SA; Research
Grant, Ipsen SA
Christian Buchbender, Duesseldorf, Germany (Abstract Co-Author) Nothing to Disclose
Julian Kirchner, Dusseldorf, Germany (Presenter) Nothing to Disclose
For information about this presentation, contact:

Lale.Umutlu@uk-essen.de
PURPOSE

To compare the diagnostic value of a one-step to a two-step staging algorithm for local and whole-body staging utilizing 18F-FDG
PET/MRI in breast cancer patients.
METHOD AND MATERIALS

A total of 38 patients (37 females and one male, mean age 57 ± 10 years; range 31-78 years) with newly diagnosed,
histopathologically proven breast cancer were prospectively enrolled in this trial. All PET/MRI examinations were assessed for local
tumor burden and metastatic spread in two separate reading sessions: (1) One-step algorithm comprising supine whole-body 18F-
FDG PET/MRI, (2) Two-step algorithm comprising a dedicated prone 18F-FDG breast PET/MRI and supine whole-body 18F-FDG
PET/MRI.
RESULTS

On a patient based analysis the two-step algorithm correctly identified 37 out of 38 patients with breast carcinoma (97%), while 5
patients were missed by the one-step 18F-FDG PET/MRI algorithm (33/38; 87% correct identification; p=0.37). On a lesion-based
analysis 56 breast cancer lesions were detected in the two-step algorithm and 44 breast cancer lesions could be correctly
identified in the one-step 18F-FDG PET/MRI (79%), resulting in statistically significant differences between the two algorithms
(p=0.0015). For axillary lymph node evaluation sensitivity, specificity and accuracy was 93%, 95 % and 94%, respectively.
Furthermore, distant metastases could be detected in 7 patients with both modalities.
CONCLUSION
The results demonstrate the necessity and superiority of a two-step 18F-FDG PET/MRI algorithm, comprising dedicated prone
breast imaging and supine whole-body imaging, when compared to the one-step algorithm for local and whole-body staging in
breast cancer patients.
CLINICAL RELEVANCE/APPLICATION
Two-step 18F-FDG PET/MRI comprising dedicated breast and whole-body imaging enables high-quality local and whole-body staging
in patients with breast cancer.
SSJ17-06 Multimodal Radiomic Imaging: Comparison of PET and MRI-pCM Heterogeneity in Breast Cancer
Tuesday, Nov. 27 3:50PM - 4:00PM Room: S504CD

Participants
Bert-Ram Sah, MD, London, United Kingdom (Presenter) Nothing to Disclose
Marta Bogowicz, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Christian Leissing, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Stephanie Tanadini-Lang, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Patrick Veit-Haibach, MD, Zurich, Switzerland (Abstract Co-Author) Research Grant, Bayer AG Resaarch Grant, F. Hoffmann-La
Roche Ltd Research Grant, General Electric Company
PURPOSE
This study investigated the value of pre-treatment F-18-Fluorodeoxyglucose (FDG)-positron-emission-tomography (PET) radiomics
in comparison to T1-weighted-post-contrast-magnetic-resonance-imaging (MRI-pCM) radiomics in patients with breast cancer.
METHOD AND MATERIALS
Following IRB approval and informed consent, a total of 30 patients with histologically proven breast cancer were prospectively
recruited. Patients were injected 225+/-55 MBq FDG intravenously. PET and MRI-pCM were acquired on the same machine. 154
radiomic features of first, second, and higher order were extracted from the primary tumor. Dimensionality of features was reduced
with a Principal Component Analysis. The relationship of selected features to staging and histological parameters was determined.
Association of features between the different modalities was compared (Spearman "ρ").
RESULTS

Selected radiomic features of PET showed moderate correlation to T-stage (-0.52 < ρ < 0.54) and weak correlation to N-Stage (-
0.35 < ρ < 0.38). Selected radiomic features of MRI-pCM showed moderate correlation to T-stage (-0.64 < ρ < 0.57) and to N-
stage (-0.52 < ρ < 0.54). Correlation of radiomic features of both modalities to hormone receptor status is shown in Table 1.
Comparison between PET and MRI-pCM showed moderate to strong correlation for the comparison of all radiomic features (-0.66 <
ρ < 0.68) (Figure 1), whereas the correlation for the comparison of a respective radiomic parameter was only weak to moderate
(0.22 < ρ < 0.56) (1st diagonal in Figure 1).
CONCLUSION

Radiomics in a multimodality approach might be a complementary tool for non-invasive pre-therapeutic characterization of breast
cancer.
CLINICAL RELEVANCE/APPLICATION

Combining radiomic features from different imaging modalities may help in non-invasive specification of breast cancer.
 VW 13

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Tuesday, Nov. 27 3:00PM - 3:30PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 14

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Tuesday, Nov. 27 3:30PM - 4:00PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 41

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
Tuesday, Nov. 27 3:50PM - 5:00PM Room: Booth 5530

Participants
Ingolf Karst, MD, Chicago, IL (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 VW 66

Breast Density Assessment: Beyond Radiologist Visual Assessment: Hologic Vendor Workshop
Tuesday, Nov. 27 4:00PM - 4:30PM Room: Booth 5524

Participants
Regina J. Hooley, MD, New Haven, CT (Presenter) Consultant, Hologic, Inc
Program Information
A brief session for radiologists interested in learning more about implementing Breast Density Assessment Software. Including
decreasing inter/intra-reader variability and increasing confidence in selecting appropriate patients that may need additional
screening. (Quantra™)
Registration
https://hologicrsna.com
 RC415

The Neoadjuvant Patient

Tuesday, Nov. 27 4:30PM - 6:00PM Room: E353B

BR MR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Jessica W. Leung, MD, Houston, TX (Moderator) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc; Speakers
Bureau, FUJIFILM Holdings Corporation
For information about this presentation, contact:
JWLeung@MDAnderson.org
zuleyml@upmc.edu
LEARNING OBJECTIVES
1) To discuss three clinically significant areas involving care of the breast cancer patient undergoing neoadjuvant therapy. 2) To
apply in everyday clinical practice the principles and conclusions learned.
Sub-Events
RC415A Ongoing Trials
Participants
Jessica W. Leung, MD, Houston, TX (Presenter) Scientific Advisory Board, Hologic, Inc; Speakers Bureau, Hologic, Inc; Speakers
Bureau, FUJIFILM Holdings Corporation
For information about this presentation, contact:
JWLeung@MDAnderson.org
LEARNING OBJECTIVES
1) To learn the design of some of the ongoing clinical trials involving care of the breast cancer patient receiving neoadjuvant
therapy. 2) To describe the imaging components of these trials.
RC415B Evaluation of the Axilla
Participants
Steven P. Poplack, MD, Saint Louis, MO (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Identify the key US criteria that are predictive of axillary lymph node metastases. 2) Appraise the accuracy of axillary US. 3)
Describe the role of axillary US in the surgical management of the axilla after neoadjuvant treatment.
RC415C Role of MR
Participants
Eric L. Rosen, MD, Denver, CO (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Describe the role of Breast MRI in identifying candidates for Neoadjuvant Chemotherapy. 2) Discuss published data regarding the
ability of Breast MRI to assess response to Neoadjuvant Chemotherapy. 3) Identify and review both the predictive and prognostic
ability of Breast MRI in patients receiving Neoadjuvant Chemotherapy. 4) Identify advances in MRI likely to enhance its already
established role in evaluating breast cancer patients receiving neoadjuvant Chemotherapy.
 ED001-W E

Breast Wednesday Case of the Day

W ednesday, Nov. 28 7:00AM - 11:59PM Room: Case of Day, Learning Center

AMA PRA Category 1 Credit ™: .50

Participants
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Susan O. Holley, MD, PhD, Raleigh, NC (Abstract Co-Author) Nothing to Disclose
Alexander B. Sevrukov, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Catherine A. Young, MD, JD, Saint Louis, MO (Abstract Co-Author) Research support, Hologic, Inc
Cheryl R. Herman, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Mai A. Elezaby, MD, Madison, WI (Abstract Co-Author) Research Grant, Exact Sciences Corporation
Lonie R. Salkowski, MD,PhD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Urvi A. Tailor, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Ramapriya Ganti, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1) Identify, characterize, and analyze abnormal findings on multimodality breast imaging studies. 2) Develop differential diagnostic
considerations based on the clinical information and imaging findings. 3) Recommend appropriate management for the patients
based on imaging findings.
 RC515

BI-RADS Interactive Challenge (Interactive Game)

W ednesday, Nov. 28 8:30AM - 10:00AM Room: N227B

BR MR US
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Jay R. Parikh, MD, Houston, TX (Moderator) Nothing to Disclose
GENERAL INFORMATION
This interactive session will use RSNA Diagnosis Live™. Please bring your charged mobile wireless device (phone, tablet or laptop) to
participate.
Sub-Events
RC515A Mammography
Participants
Cindy S. Lee, MD, Garden City, NY (Presenter) Nothing to Disclose
For information about this presentation, contact:
Cindy.Lee3@nyumc.org
LEARNING OBJECTIVES
1) To define BIRADS terms used to describe mammography findings and recognize common appearances of malignant and benign
disease.
ABSTRACT
This case-based session will review a variety of imaging findings at mammography, appropriate use of BI-RADS descriptors and
categories, as well as highlight potential pitfalls and strategies to avoid them.
Active Handout:Cindy S. Lee
http://abstract.rsna.org/uploads/2018/18000473/RSNA 2018 DBT Handout RC515A.pdf
RC515B Case-based Session on Ultrasound
Participants
Ana P. Lourenco, MD, Providence, RI (Presenter) Nothing to Disclose
For information about this presentation, contact:
alourenco@lifespan.org
LEARNING OBJECTIVES
1) Apply appropriate BI-RADS descriptors and categories to a variety of benign and malignant lesions on ultrasound. 2) Describe
how certain imaging features may predict pathology.
ABSTRACT
This case-based session will review a variety imaging findings at US, appropriate use of BI-RADS descriptors and categories, as well
as highlight potential pitfalls and strategies to avoid them.
Active Handout:Ana P. Lourenco
http://abstract.rsna.org/uploads/2018/18000474/Lourenco.RSNA.US RC515B.pdf
RC515C Case-based Session on MRI
Participants
Elizabeth S. McDonald, MD, Philadelphia, PA (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Define BIRADS terms used to describe MRI findings and recognize common appearances of malignant and benign disease.
RC515D Audit and Outcomes Monitoring
Participants
Jay R. Parikh, MD, Houston, TX (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Describe the basic clinically relevant mammography medical outcomes audit based on BIRADS. 2) Strategies to monitor outcomes
and provide radiologist feedback.
ABSTRACT
Bi-RADS enables annual the basic clnically relevant medical outcomes audit, which can be used to monitor outcomes and provide
radioloigsts feedback.
 VW 44

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Reading Sessions: Siemens
Healthineers Vendor Workshop
W ednesday, Nov. 28 10:15AM - 11:25AM Room: Booth 5530

Participants
Luis Pina, MD, PhD, San Sebastian, Spain (Presenter) Nothing to Disclose
Program Information
Learn about the value of wide-angle Digital Breast Tomosynthesis (DBT) and Contrast Enhanced Mammography (CEM) in the daily
routine from one of our most experienced clinical experts. The differences and respective advantages of the morphological (DBT)
and functional (CEM) breast imaging methods will be discussed. This, all with the flexible assistance of our multi-modality reading
solution syngo® via and the syngo® Breast Care applications.
 SSK01

Breast Imaging (Digital Breast Tomosynthesis: Screening and Diagnostic Indications)

W ednesday, Nov. 28 10:30AM - 12:00PM Room: E451A

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Sarah M. Friedewald, MD, Chicago, IL (Moderator) Consultant, Hologic, Inc; Research Grant, Hologic, Inc;
Susan Weinstein, MD, Philadelphia, PA (Moderator) Nothing to Disclose
Sub-Events
SSK01-01 Blinded Observer Study: "Virtual" Full-Dose (VFD) Digital Breast Tomosynthesis (DBT) Images
Derived from Reduced-Dose Acquisitions versus Clinical Full-Dose DBT Images
Participants
Junchi Liu, MS, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Syed Ammar Qadir, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Amin Zarshenas, MSc, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Limin Yang, MD, PhD, Iowa City, IA (Abstract Co-Author) Nothing to Disclose
Laurie L. Fajardo, MD, MBA, Park City, UT (Abstract Co-Author) Consultant, Hologic, Inc; Consultant, Siemens AG; Consultant,
FUJIFILM Holdings Corporation;
Kenji Suzuki, PhD, Chicago, IL (Presenter) Royalties, General Electric Company; Royalties, Hologic, Inc; Royalties, MEDIAN
Technologies; Royalties, Riverain Technologies, LLC; Royalties, Canon Medical Systems Corporation; Royalties, Mitsubishi
Corporation; Royalties, AlgoMedica, Inc
For information about this presentation, contact:
jliu118@hawk.iit.edu
PURPOSE
We developed a deep-learning-based "virtual" higher-dose (VHD) technology for radiation dose reduction in DBT. The purpose of our
study was to compare the image quality of our VFD images generated from half-dose acquisitions to that of real clinical full-dose
images in DBT.
METHOD AND MATERIALS

Our deep-learning-based VHD technology employed our original patched-based neural network convolutional deep learning to
convert lower-dose (LD) to higher-dose (HD) tomosynthesis images. To evaluate our VHD technology, we collected half-dose (50%
of the standard dose: 32±14 mAs at 33±5 kVp) and full-dose (100% of the standard dose: 68±23 mAs at 33±5 kvp) images of 51
clinical screening cases with a DBT system (Selenia Dimensions, Hologic, Inc, Bedford, MA) at University of Iowa Hospitals & Clinics.
We applied our VHD technology to the 51 cases to convert half-dose images to VFD images. We invited 35 breast radiologists to
participate in our observer rating study to rate and distinguish blinded VFD and real full-dose DBT images of 10 of the 51 cases. A
VFD image and its corresponding real full-dose image were shown on two clinical LCD monitors (EIZO RadiForce GX540) in a blinded
manner. Radiologists were asked to rate the image quality on a 0-to-100 scale and to provide their choices as to which one was
better in image quality.
RESULTS
Among the 35 breast radiologists, 21 (60%) radiologists either preferred our VFD DBT images over the real full-dose images or could
not distinguish between the two in our observer rating study. The mean scores of the image quality of our VFD images and the real
full-dose images were 83.2±3.2 and 84.0±3.0, respectively. The difference in image quality between VFD and real-full dose images
was not statistically significant (p=0.37).
CONCLUSION
Our blinded observer study with 35 breast radiologists demonstrated that our deep-learning-based VFD images generated from half-
dose acquisitions were equivalent to real full-dose DBT images. Thus, our VHD technology achieved 50% dose reduction without
sacrificing the image quality.
CLINICAL RELEVANCE/APPLICATION
Substantial radiation dose reduction with the observer-study-proven VHD technology would benefit patients by reducing the
lifetime risk of radiation-induced cancer from DBT screening.
SSK01-02 Comparison of DM/Tomosynthesis and Synthesized DM/Tomosynthesis False Negative Cancers in a
Population-Based Breast Cancer-Screening Program
W ednesday, Nov. 28 10:40AM - 10:50AM Room: E451A

Awards
Student Travel Stipend Award
Participants
Samantha P. Zuckerman, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Rebecca Hubbard, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. McDonald, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Susan Weinstein, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Marie Synnestvedt, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
For information about this presentation, contact:
samantha.zuckerman@uphs.upenn.edu
PURPOSE
Synthesized 2D imaging (s2D) is replacing 2D digital mammography (DM) in digital breast tomosynthesis (DBT) screening to reduce
radiation dose. However, there have been reports of decreased rates of detection of in situ carcinomas with s2D/DBT screening.
The purpose of this study is to compare screening outcomes as well as false negative rates and lesion types in DM/DBT versus
s2D/DBT screening.
METHOD AND MATERIALS
Recall rate percentage (RR), cancer detection (CDR) and false negative (FN) rates per 1000 screened, false negative cancer
subtype (invasive versus in situ) and the method of detection of the false negative cancer (symptomatic versus by another imaging
modality) were compared for 37,184 women screened with DM/DBT from 10/3/2011-1/6/2015 and 37,996 women screened with
s2D/DBT from 1/7/2015-1/6/2018. Differences were compared using chi-squared tests at the standard a=0.05 significance level
with Yates correction. All statistical tests were two-sided.
RESULTS

RR decreased with s2D/DBT versus DM/DBT screening - 6.9% versus 8.9% (p<0.001). CDR for DM/DBT and s2D/DBT screening were
not statistically different (6.0/1000 vs. 5.5/1000, p=0.37). However, FN rate doubled for s2D/DBT (0.84/1000, n=32) versus
DM/DBT screening (0.40/1000, n=15), p=0.02. While not statistically significant, there was a trend of more asymptomatic FN
cancers detected by other modalities (i.e., MR, US, CT) in the s2D cohort than in the DM/DBT cohort (13/32 (41%) versus 5/15
(33%), p=0.11) and a trend of higher proportion of DCIS in the s2D cohort than in the DM cohort (9/32 (28%) versus 3/15 (20%),
p=0.16).
CONCLUSION
s2D/DBT maintains CDR with the benefit of decreased recall rates. However, the FN rate increased with s2D/DBT compared to
DM/DBT with both in situ and invasive FN cancers increased in the s2D cohort. While not statistically significant, a greater
proportion of s2D FN cancers were detected by other modalities in asymptomatic patients. The increase in false negative rates is
multifactorial and may be intrinsically related to the new technology and/or a learning curve in implementing the s2D technology.
CLINICAL RELEVANCE/APPLICATION

The replacement of DM with s2D in a large DBT breast cancer screening program maintained cancer detection rates and decreased
recall rates, but increased false negative rates.
SSK01-03 Integration of Digital Breast Tomosynthesis into Breast Cancer Screening Practices in the United
States: A Comparative Modeling Analysis
W ednesday, Nov. 28 10:50AM - 11:00AM Room: E451A

Participants
Kathryn Lowry, MD, Seattle, WA (Presenter) Nothing to Disclose
Natasha K. Stout, PhD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Oguzhan Alagoz, PhD, Madison, WI (Abstract Co-Author) Consultant, Renaissance Rx
Elizabeth S. Burnside, MD,MPH, Madison, WI (Abstract Co-Author) Dr. Burnside has a research grant from Hologic
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
Karla Kerlikowske, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Anne Marie McCarthy, Boston, MA (Abstract Co-Author) Nothing to Disclose
Diana Miglioretti, PhD, Seattle, WA (Abstract Co-Author) Scientific Advisory Board, Hologic, Inc
Brian L. Sprague, PhD, Burlington, VT (Abstract Co-Author) Nothing to Disclose
Anna N. Tosteson, Lebanon, NH (Abstract Co-Author) Nothing to Disclose
Martin J. Yaffe, PhD, Toronto, ON (Abstract Co-Author) Research collaboration, General Electric Company; Shareholder, Volpara
Health Technologies Limited; Co-founder, Mammographic Physics Inc; Research Consultant, BHR Pharma LLC
Clyde Schechter, Bronx, NY (Abstract Co-Author) Nothing to Disclose
Amy Trentham-Dietz, Madison, WI (Abstract Co-Author) Nothing to Disclose
PURPOSE

To project long-term outcomes and cost-effectiveness of transitioning from breast cancer screening using digital mammography
(DM) to digital mammography with digital breast tomosynthesis (DBT) in the United States.
METHOD AND MATERIALS
Two established breast cancer models were used to simulate two scenarios: complete transition from DM to DBT for all women ages
40+ between 2011-2020 versus continued use of DM alone. Screening utilization was based on observed dissemination patterns
from national surveillance datasets. We assumed current screening and treatment patterns continued, and women were followed for
their remaining lifetimes. DM and DBT performance was based on screening data from the NCI's Population Based Research
Optimizing Screening through Personalized Regimen (PROSPR) consortium from 2011-2014. Costs and quality of life weights were
based on US national averages and published literature. Outcomes included life-years (LY), quality-adjusted life-years (QALYs),
breast cancer deaths, false positive exams (FP), costs (2017 US$) and incremental cost-effectiveness ratios (ICER). Analyses were
performed from the payer perspective. Results were summarized within and across models.
RESULTS
Transition to DBT had the greatest impact on FP screening mammograms, which reduced FP exams by 278-288/1,000 women. The
small difference in test sensitivity observed in the PROSPR data translated to minimal differences in breast cancer deaths and LY
gains, with a reduction in deaths from 0 to 0.03/1,000 women and LY gains from -1 to 0.05 years/1,000 women. Total costs
increased by $5.64-$5.66 million, with ICERs of $193,634-$217,532/QALY with DBT relative to DM. In sensitivity analyses, ICERs
were sensitive to both the test performance and costs of screening DBT. For example, ICERs decreased to $141,043/QALY with 2%
absolute improvement in DBT sensitivity; ICERs decreased to $163,092/QALY and $59,872 with $20 and $40 reductions in cost of
DBT exams, respectively.
CONCLUSION
The transition from DM to DBT for routine breast cancer screening in the U.S. reduces FP results, but substantially increases costs
assuming current estimates of performance and reimbursement rates.
CLINICAL RELEVANCE/APPLICATION
Digital breast tomosynthesis for routine breast cancer screening in the U.S. reduces false positive exams, but substantially
increases costs based on current estimates of performance and reimbursement rates.
SSK01-04 Interval Cancers at Digital Breast Tomosynthesis (DBT) and Full-Field Digital Mammography (FFDM)
in a Hybrid Imaging Environment
W ednesday, Nov. 28 11:00AM - 11:10AM Room: E451A

Participants
Gunjan M. Senapati, MD, Boston, MA (Presenter) Nothing to Disclose
Aijia Wang, Boston, MA (Abstract Co-Author) Nothing to Disclose
Pragya A. Dang, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Sona A. Chikarmane, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Ivan Ip, MD, MPH, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Ronilda Lacson, MD, PhD, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Ramin Khorasani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
gsenapati@bwh.harvard.edu
PURPOSE
To review imaging features and histopathology of interval cancers (IC) on screening full-field digital mammography (FFDM) and
screening digital breast tomosynthesis (DBT) in a hybrid imaging environment.
METHOD AND MATERIALS

This HIPAA-compliant IRB approved retrospective review of consecutive screening exams (9,828 DBT and 41,713 FFDM exams) from
October 2012-September 2014 identified interval cancers by cross reference to the institutional cancer registry. Interval cancer
was defined as cancer detected within 365 days of a negative screening exam. During the study period a hybrid imaging
environment for screening existed, using both FFDM and DBT. Three breast radiologists reviewed prior mammograms of all IC in
consensus. Cancers were classified as missed (actionable), minimal signs (non-actionable), or true negative. Mammographic lesion
features and breast density were described. Electronic medical record review of patient demographics and histopathology was
performed. Percentages were compared using Fisher's exact test.
RESULTS

There were 34 interval cancers (20 FFDM, 14 DBT). IC were considered missed, actionable in 2/20 (10%) FFDM and 2/14 (14%)
DBT; minimal signs, non-actionable in 3/20 (15%) FFDM and 0/14 (0%) DBT, and true negative in 15/20 (75%) FFDM and 12/14
(86%) DBT (p-value = 0.4061). Of the 5 cancers visible (2 missed, 3 minimal signs) on prior FFDM, 3 were asymmetries, 1
calcifications, and 1 architectural distortion. Both cancers visible (both missed) on prior DBT were spiculated masses. Most IC on
both FFDM and DBT were moderate-to-high grade invasive carcinoma (n=29): 17/20 (85%) FFDM and 12/14 (86%) DBT (p-value =
0.9999). 1 case of pure intermediate grade DCIS, presenting as nipple discharge, was missed on FFDM; the remaining four, 2 FFDM
and 2 DBT, were grade 1 invasive carcinoma. Most IC were in Category 3 and 4 dense breasts (n=26): 15/20 (75%) FFDM and
11/14 (79%) DBT (p-value = 0.9999).
CONCLUSION

In this hybrid FFDM/DBT screening environment, most IC are mammographically occult at prior imaging, occur in dense breast
tissue, and are moderate to high grade invasive cancers.
CLINICAL RELEVANCE/APPLICATION

Because FFDM and DBT techniques both rely on lesion morphology for detection, IC on each technique have similarities, occurring in
dense tissue and with moderate to high grade invasive histopathology.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
SSK01-05 Evidence Supporting Digital Breast Tomosynthesis as Primary Mammographic Screening Tool:
Sustained Improved Outcomes over 7 Consecutive Years
W ednesday, Nov. 28 11:10AM - 11:20AM Room: E451A

Participants
Liane E. Philpotts, MD, New Haven, CT (Presenter) Consultant, Hologic, Inc
Tamara Y. Carroll, MD, Seattle, WA (Abstract Co-Author) Nothing to Disclose
Madhavi Raghu, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
liane.philpotts@yale.edu
PURPOSE

Digital breast tomosynthesis (DBT) has shown promise in multiple individual, multi-institutional and population based practices to
improve screening metrics, particularly reduce recall rates and improve cancer detection. The sustainability of the early results has
not yet been well demonstrated. Long term results are needed to demonstrate if the technology is the preferred screening tool
compared with 2D full field digital mammography (FFDM). The purpose of this study was to assess screening metrics with DBT over
a 7 year period.
METHOD AND MATERIALS

DBT screening (Dimensions, Hologic, Inc, Bedford, MA) was offered to all women free of charge at a dedicated academic breast
center starting in August 2011 and 3 out-patient satellite offices that obtained DBT units in the following years. Over a 7 year
period, 124,669 screening DBT exams were performed. Screening metrics were obtained from the breast imaging electronic
database (PenRad, MN) and assessed by one-year intervals starting in August 2011. Recall rate (RR), Cancer detection rate (CRD),
positive predictive value of screening recall (PPV1) and of biopsy (PPV3) were assessed. Comparison with historic 2D rates (8/1/08
though 7/31/11) were performed.
RESULTS

The RR was significantly reduced over 2D and showed a decreasing trend for each consecutive year: 7.9%, 8.8%, 7.8%, 7.5%,
6.9%, 6.7%, 6.2% (2D = 11.4) (p<0.0001). The CDR (per 1000) showed a stable trend that was statistically significantly improved
over 2D: 5.8, 5.2, 5.4, 5.6, 6.6, 5.6, 5.1 (2D = 3.8)(p<.0001). The PPV1 showed a sustained significant improvement over 2D: 7.2,
5.8, 7.0, 7.5, 9.5, 8.4, 8.2 (2D=3.3) (p<.0001). The PPV3 also showed a striking significant increase over 2D and an upward trend
over consecutive DBT years: 35%, 31%, 36%, 37%, 47%, 42%, 44% (2D=29%) (p<0.05).
CONCLUSION

Screening metrics with DBT over 7 years were sustainably significantly improved over 2D rates and further demonstrate favorable
trends of improvement over time. This may reflect learning curve and/or increasing availability of prior comparison tomosynthesis
exams.
CLINICAL RELEVANCE/APPLICATION

The sustained use of DBT demonstrates that there are fewer false positive screening recalls and biopsies which is essential for
shifting the harms and benefits of screening.
SSK01-06 Impact of Using Digital Breast Tomosynthesis in Diagnostic Mammography
W ednesday, Nov. 28 11:20AM - 11:30AM Room: E451A

Participants
Emily Ambinder, MD, MSc, Baltimore, MD (Presenter) Nothing to Disclose
Lisa A. Mullen, MD, Cockeysville, MD (Abstract Co-Author) Nothing to Disclose
Delaram Shakoor, MD, Baltimore, MD (Abstract Co-Author) Nothing to Disclose
Kelly Myers, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Eniola T. Falomo, MD, Baltimore, MD (Abstract Co-Author) Nothing to Disclose
Susan C. Harvey, MD, Lutherville, MD (Abstract Co-Author) Consultant, Hologic, Inc Consultant, IBM Corporation
For information about this presentation, contact:

emcinto8@jhmi.edu
PURPOSE

The use of digital breast tomosynthesis (DBT) in the screening setting has been shown to decrease recall rate and improve cancer
detection. This study evaluates the impact of using DBT in the diagnostic setting at a large academic institution.
METHOD AND MATERIALS

All diagnostic mammograms performed from 7/1/2013 to 8/24/2017 were reviewed. Diagnostic mammograms performed to further
evaluate calcifications were excluded, as spot magnification views, rather than DBT views, are typically used for diagnostic
evaluation in these cases. Studies were divided into two groups: those performed with at least one DBT view and those performed
with only full field digital mammography (FFDM). We compared the frequency of a negative/benign assessment (BI-RADS 1 or 2), a
probably benign assessment (BI-RADS 3), and a suspicious or highly suspicious assessment (BI-RADS 4 or 5) between the two
groups. We also compared positive predictive value 2 and cancer detection rate between the two groups. The Chi-squared test
was used for statistical analysis.
RESULTS
12,324 studies were included in the DBT group and 11,775 studies were included in the FFDM group. There was a significantly
higher percentage of BI-RADS 1 or 2 assessments (77.8% vs. 74.9%, p<0.001) and a significantly lower percentage of BI-RADS 3
assessments (10.6% vs. 12.9%, p<0.001) in the DBT group compared to the FFDM group. There was no significant change in the
percentage of BI-RADS 4 or 5 assessments (biopsy recommendations) between the two groups (12.1% vs. 11.6%, p=0.20). Both
PPV2 and CDR were significantly higher for studies performed with DBT compared to FFDM (PPV2: 36.1 vs. 26.6, p<0.001; CDR:
41.8 vs. 32.3, p<0.001).
CONCLUSION

In our study, using DBT in the diagnostic setting led to more studies being assessed as normal with less frequent need for short-
term follow-up, implying decreased associated cost of follow-up and decreased stress and anxiety for patients. While the number of
biopsy recommendations was similar between the groups, PPV2 and CDR increased when DBT views were included, suggesting
improved accuracy of biopsy recommendations.
CLINICAL RELEVANCE/APPLICATION

Routine use of digital breast tomosynthesis for diagnostic mammography may result in more confident assessments, and could lead
to resource savings and improved patient-centered care.
SSK01-07 Architectural Distortion (AD) on Digital Breast Tomosynthesis (DBT): Outcomes, Histopathology, and
Predictive Features of Malignancy
W ednesday, Nov. 28 11:30AM - 11:40AM Room: E451A

Participants
Sona A. Chikarmane, MD, Boston, MA (Presenter) Nothing to Disclose
Christine M. Denison, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Gunjan M. Senapati, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Ronilda Lacson, MD, PhD, Brookline, MA (Abstract Co-Author) Nothing to Disclose
Ramin Khorasani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

cdenison@bwh.harvard.edu
PURPOSE
1) To determine histopathologic outcomes for architectural distortion (AD) detected on tomoshynthesis (3D) mammography, (2) to
investigate imaging features predictive of malignancy.
METHOD AND MATERIALS

A HIPAA-compliant IRB approved retrospective review of a mammography screening database between 12/2012-5/2015 identified
297 consecutive screening mammograms classified as BI-RADS 0 for AD. All breast imaging available for cases classified as BI-RADS
4 and 5 at diagnostic work-up were reviewed in consensus by 3 breast radiologists. Imaging features and biopsy accuracy were
assessed. Medical records were reviewed for patient demographics, histopathology and follow-up imaging. Chi-square tests were
performed with <0.05 significance.
RESULTS
Of 297 BI-RADS category 0 screening detected AD on DBT, 45 (15.2%) were subsequently categorized as BI-RADS category 4-5.
Cases were excluded if the finding was a mass rather than AD alone (n=3) or if no pathology results were available (n=2). Within
the study population of 40 cases, 20/40 (50%) were malignant (18/20 [90%] invasive, 2/20 [10%] ductal carcinoma in situ), 3/40
(7.5%) atypical ductal hyperplasia, and 17/40 (42.5%) benign. 31/40 (77.5%) cases of AD were visible on 3D only and 9/40
(22.5%) were seen on 2D and 3D. Malignancy was found in 15/31 (48.4%) 3D only AD and 5/9 (56%) AD visible on 2D and 3D
(p=1.00). While presence of a sonographic correlate did not increase likelihood of malignancy (US correlate in 33 cases [17 benign,
16 malignant] and no US correlate in 7 cases [3 benign, 4 malignant] [p=1.00]), the specific finding of a mass was more likely
malignant than non-mass findings (17 masses [4 benign, 13 malignant] and 16 non-mass [13 benign, 3 malignant] [p=0.0016].
Diagnostic MRI was performed in 7/40 (17.5%) cases, of which 3/7 (1 malignant, 2 benign) had a correlate for AD and 4/7 (0
malignant) did not.
CONCLUSION

The majority of AD cases in this series were seen on 3D only, although risk of malignancy for 3D only visible compared to 2D plus 3D
visible was similar. AD is more likely to represent invasive disease than in situ and is more likely malignant if a sonographic mass is
present. Given the high malignancy rate, negative US or breast MRI should not obviate the need for biopsy of mammographic
detected AD.
CLINICAL RELEVANCE/APPLICATION

Given the high malignancy rate associated with AD detected on 3D mammography, lack of US or MRI correlate should not obviate
biopsy.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
SSK01-08 Evaluating the Clinical Performance of Stationary 3D Mammography
W ednesday, Nov. 28 11:40AM - 11:50AM Room: E451A

Awards
Student Travel Stipend Award
Participants
Connor Puett, Chapel Hill, NC (Presenter) Nothing to Disclose
Christy Inscoe, MS, BS, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
Beilin Jia, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
Connie E. Kim, MD, Durham, NC (Abstract Co-Author) Spouse, Consultant, ClarVista Medical, Inc Spouse, Royalties, Leica
Biosystems Nussloch GmbH Spouse, Intellectual property, Leica Biosystems Nussloch GmbH
Sora C. Yoon, MD, Durham, NC (Abstract Co-Author) Nothing to Disclose
Ruth Walsh, MD, Durham, NC (Abstract Co-Author) Nothing to Disclose
Suk Jung Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Cherie M. Kuzmiak, DO, Chapel Hill, NC (Abstract Co-Author) Research Grant, Delphinus Medical Technologies, Inc
Donglin Zeng, PhD, Chapel Hill, NC (Abstract Co-Author) Nothing to Disclose
Jianping Lu, Chapel Hill, NC (Abstract Co-Author) Consultant, Xintek Inc; Consultant, XinVivo Inc; Consultant, XinRay Systems Inc
Otto Zhou, PhD, Chapel Hill, NC (Abstract Co-Author) Board of Directors, XinRay Systems Inc
Yueh Z. Lee, MD,PhD, Chapel Hill, NC (Abstract Co-Author) License agreement, XinRay Systems Inc
For information about this presentation, contact:

connor_puett@med.unc.edu
PURPOSE
The purpose of this study was to assess the clinical performance of stationary 3D mammography, which is a new approach to
digital breast tomosynthesis that uses a fixed array of carbon nanotube-based x-ray sources to acquire the projection views.
METHOD AND MATERIALS
Women with a suspicious abnormality (BIRADS 4 lesion) identified by screening digital mammography were recruited for the study,
which involved imaging by stationary 3D mammography prior to biopsy. Pre-clinical testing has demonstrated that stationary 3D
mammography offers a higher spatial resolution than commercially-available 3D mammography devices, since it solves the problem
of source blur. In this study, the stationary 3D mammography device collected 15 projections over an angular span of 28°. Four
radiologists were asked to evaluate the likelihood of malignancy and rate breast density (BIRADS 1-4) when interpreting the
standard mammogram and stationary 3D mammography images. Using pathology as ground truth, reader performance was quantified
as the area under the receiver operating characteristic curve (AUC), while multivariate analysis with a fitted linear mixed-effect
model was used to relate breast density to reader performance.
RESULTS
43 women [average age: 56.7 (35 to 83) years] provided a lesion-enhanced image set, in which malignancy was found to be
present in 28% of participants by pathology. On average, readers were more accurate identifying malignancy when interpreting the
stationary 3D mammography images compared to the standard mammograms, as demonstrated by the significantly higher
(p<0.0001) mean AUC for stationary 3D mammography. This higher accuracy was present and statistically-significant across the full
range of breast densities.
CONCLUSION
In this first-in-human study, readers were more likely to identify malignancy correctly when interpreting stationary 3D
mammography images compared to the standard digital mammogram. Given these encouraging findings, as well as the results of
pre-clinical testing, future trials are being designed to compare the performance of stationary 3D mammography to commercially-
available 3D mammography devices in the clinic.
CLINICAL RELEVANCE/APPLICATION
Stationary 3D mammography may prove to be a valuable clinical tool, as readers were more accurate identifying malignancy when
interpreting its images compared to the standard digital mammogram.
SSK01-09 Molecular Characterization of Breast Cancers: Could It Change Potential Overdiagnosis Analysis?
W ednesday, Nov. 28 11:50AM - 12:00PM Room: E451A

Participants
Francesca Caumo, MD, Padua, Italy (Presenter) Nothing to Disclose
Gisella Gennaro, PhD, Padua, Italy (Abstract Co-Author) Nothing to Disclose
Giovanna Romanucci, Verona, Italy (Abstract Co-Author) Nothing to Disclose
Marco Zappa, Firenze, Italy (Abstract Co-Author) Nothing to Disclose
PURPOSE
To compare the amount of potentially overdiagnosed breast cancers in two screening populations using tomosynthesis (DBT) and
digital mammography (FFDM) considering only tumor stage or also molecular features.
METHOD AND MATERIALS

Cancers detected within a prospective screening trials using DBT were compared with those obtained from an historical FFDM
screening cohort. The amount of potentially overdiagnosed breast cancers in the two cohorts was calculated according to two
different assumptions: (1) stages 0 and 1 cancers (DCIS, IDCs, ILCs); (2) considering low-grade DCIS and IDCs/ILCs that are
simultaneously grade 1, stage IA or IB, luminal A subtype. Rates (Fisher exact test) were used to compare the two hypotheses of
potential overdiagnosis. A p-value lower than 0.05 was considered statistically significant. The same analysis was performed on the
complementary subgroups for the two hypotheses: (1) stage > 1 tumors; (2) any other combination of features with the exception
of that defined with the second assumption.
RESULTS
Cancers detected in the two screening populations was 322 from the DBT trial and 153 from the FFDM cohort. Considering only
tumor stage, cancers potentially overdiagnosed were 268 with DBT and 116 with FFDM, equivalent to 83.2% and 75.8%,
respectively. Comparing rates, DBT found 7.9/1000 stage 0/1 cancers vs. 4.0/1000 found by FFDM (P<0.0001), with an incidence
rate ratio (IRR) equal to 1.99 (95%CI = [1.60-2.50]. Rates of cancers with stage >1 were 1.6/1000 with DBT vs. 1.3/1000 with
FFDM, not significantly different (P=0.2817); the IRR was 1.26 (95%CI = [0.81-1.97]). Including in the definition of potentially
overdiagnosed cancers also molecular features, the numbers become 61/322 (18.9%) with DBT and 13/153 (8.5%) with FFDM;
rates of overdiagnosed cancers with the second assumption were 1.8/1000 with DBT and 0.4/1000 with FFDM (P<0.0001)
corresponding to an IRR of 4.04 (95%CI = [2.20-8.02]). Rates of any other cancers (not overdiagnosed) were 7.7/1000 with DBT
vs. 4.8/1000 with FFDM (P<0.0001), with an IRR of 1.61 [95%CI = [1.30-1.99]].
CONCLUSION
DBT increased the amount of potentially overdiagnosed cancers with both definitions. However, the second assumption (inclusion of
stage, grade, and molecular features) showed that cancers detected by DBT were mostly not overdiagnosed.
CLINICAL RELEVANCE/APPLICATION
Overdiagnosis by DBT is overestimated using only cancer stage as metric for overdiagnosis.
 SSK02

Breast Imaging (Artificial Intelligence)

W ednesday, Nov. 28 10:30AM - 12:00PM Room: E451B

AI BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Despina Kontos, PhD, Philadelphia, PA (Moderator) Nothing to Disclose
Maryellen L. Giger, PhD, Chicago, IL (Moderator) Stockholder, Hologic, Inc; Shareholder, Quantitative Insights, Inc; Shareholder,
QView Medical, Inc; Co-founder, Quantitative Insights, Inc; Royalties, Hologic, Inc; Royalties, General Electric Company; Royalties,
MEDIAN Technologies; Royalties, Riverain Technologies, LLC; Royalties, Mitsubishi Corporation; Royalties, Canon Medical Systems
Corporation
Sub-Events
SSK02-01 Using Deep Convolutional Neural Networks to Predict Readers' Estimates of Mammographic Density
from Raw and Processed Mammographic Images
Participants
Georgia Ionescu, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Martin Fergie, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Michael Berks, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Elaine Harkness, PhD, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Johan Hulleman, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Adam Brentnall, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Jack Cuzick, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Gareth Evans, Manchester, United Kingdom (Abstract Co-Author) Nothing to Disclose
Susan M. Astley, PhD, Manchester, United Kingdom (Presenter) Nothing to Disclose
PURPOSE
Mean percentage density assessed visually by two independent readers using Visual Analogue Scales (VAS) has a strong
association with breast cancer risk, but is resource-intensive and impractical for stratified screening. We describe a fully-automated
method for predicting this mammographic percent density measure from raw (for processing) or processed (for presentation)
mammograms, and compare association of predicted VAS score with risk.
METHOD AND MATERIALS
Convolutional Neural Networks (CNNs) were trained using 67520 whole-image mammograms from 16968 women, each labelled with
the average VAS score of two independent readers. The networks learned a mapping between mammographic appearance and
mammographic density so that they can predict density for unseen images. To evaluate its use for risk assessment, we tested on
case-control datasets of contralateral mammograms of screen detected cancers (SDC) and prior screening mammograms of women
with cancers detected subsequently. Each cancer was matched to three controls on age, menopausal status, parity, HRT and BMI.
The test datasets contained 366 cancers (SDC) and 338 (priors). Odds ratios between the top and bottom quintile were derived,
and matched concordance indices were estimated. All images were acquired on GE Senographe systems, and none of the images
from the case-control test sets were used in the training process.
RESULTS
For density estimates derived from raw images, odds ratios of cancer in the highest vs lowest quintile were 3.07 (95%CI: 1.97 -
4.77) for SDC and 3.52 (2.22 - 5.58) for priors, with matched concordance indices of 0.59 (0.55 - 0.64) and 0.61 (0.58 - 0.65)
respectively. For processed images we obtained odds ratios of 3.22 (2.06 - 5.03) for SDC and 3.65 (2.27 - 5.88) for priors.
Matched concordance indices were 0.58 (0.53 - 0.62) for SDC and 0.61 (0.57 - 0.65) for priors.
CONCLUSION
Our fully automated method demonstrated encouraging results on both raw and processed mammographic images, indicating that
either image type could be used for screening stratification.
CLINICAL RELEVANCE/APPLICATION
Mammographic density is one of the most important risk factors for breast cancer. Our fully automated method could provide a
pragmatic solution for population-based stratified screening.
SSK02-02 Breast Density Classification with Deep Convolutional Neural (DCN) Networks Utilizing 200,000
Screening Mammograms
W ednesday, Nov. 28 10:40AM - 10:50AM Room: E451B

Participants
Krzysztof J. Geras, New York City, NY (Abstract Co-Author) Nothing to Disclose
Eric Kim, MD, New York, NY (Presenter) Nothing to Disclose
Nan Wu, New York City, NY (Abstract Co-Author) Nothing to Disclose
Yiqiu Shen, New York City, NY (Abstract Co-Author) Nothing to Disclose
Jingyi Su, New York City, NY (Abstract Co-Author) Nothing to Disclose
Sungheon Kim, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Stacey Wolfson, New York, NY (Abstract Co-Author) Nothing to Disclose
Linda Moy, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Kyunghyun Cho, New York City, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
kime18@nyumc.org
PURPOSE
To develop a DCN network to reliably assess mammographic breast density
METHOD AND MATERIALS
In this retrospective study, we trained a multi-column DCN network on 200,000 digital screening mammograms performed at our
institution from 2010-2016 to assess breast density. We extracted the textual reports associated with each exam to obtain the
breast density as determined by the original interpreting radiologist. The algorithm was trained on 80% of the data sets, validated
on a separate 10%, and tested on the remaining 10%. Once this convolutional neural network classifier was trained, we performed
a reader study comparing our model to 3 radiologists. All readers independently evaluated the breast density in 100 mammograms in
a randomized order. Breast density was assessed using the conventional BI-RADS categories: Class 0 - fatty, Class 1 - scattered
fibroglandular densities, Class 2 - heterogeneously dense, and Class 3 - extremely dense. Performance of the model and the readers
were assessed using the area under the ROC curve (AUC). Kappa score was used to assess for intra-observer and inter-observer
variability.
RESULTS
Both the radiologists and our DCN model achieved a fair agreement (k = 0.34 - 0.51) with the labels in the reader study. The
agreement between the predictions of our model and the labels in the data were higher (k = 0.65 - 0.72) compared to the inter-
observer agreement between the radiologists. There was higher agreement for the fatty and extremely dense breast tissue.
Comparing our CNN model to an average of the radiologists, the CNN achieved AUC of 0.934 (class 0: 0.971, class 1: 0.859, class
2: 0.905 and class 3: 1.000) while the radiologists achieved an AUC of 0.892 (class 0: 0.960, class 1: 0.812, class 2: 0.807 and
class 3: 0.990) (Figure 1).
CONCLUSION
The level of agreement between the trained classifier and the classes in the data was found to be similar to that between the
radiologists and the classes in the data, as well as among the radiologists.
CLINICAL RELEVANCE/APPLICATION
The classifier provides quantitative, reproducible prediction of breast density, while there is often poor intra-reader and inter-reader
correlation in the qualitative assessment of breast density.
SSK02-03 Improving Radiologists' Breast Cancer Detection with Mammography Using a Deep Learning-Based
Computer System for Decision Support
W ednesday, Nov. 28 10:50AM - 11:00AM Room: E451B

Participants
Alejandro Rodriguez-Ruiz, Nijmegen, Netherlands (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Krupinski, PhD, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
Jan-Jurre Mordang, MSc, Nijmegen, Netherlands (Abstract Co-Author) Nothing to Disclose
Kathy J. Schilling Colletta, MD, Boca Raton, FL (Abstract Co-Author) Nothing to Disclose
Sylvia H. Heywang-Koebrunner, MD, Munich, Germany (Abstract Co-Author) Nothing to Disclose
Ioannis Sechopoulos, PhD, Atlanta, GA (Abstract Co-Author) Research Grant, Siemens AG; Research Grant, Canon Medical Systems
Corporation; Speakers Bureau, Siemens AG; Scientific Advisory Board, Fischer Medical
Ritse M. Mann, MD, PhD, Nijmegen, Netherlands (Presenter) Researcher, Siemens AG ; Researcher, Seno Medical Instruments, Inc;
Researcher, Identification Solutions, Inc; Researcher, Micrima Limited; Researcher, Medtronic plc; Scientific Advisor, ScreenPoint
Medical BV; Scientific Advisor, Transonic Imaging, Inc; Stockholder, Transonic Imaging, Inc
For information about this presentation, contact:
ritse.mann@radboudumc.nl
PURPOSE
To compare the breast cancer detection performance of radiologists reading mammography exams unaided versus reading using an
interactive deep learning-based computer system for decision support (DS).
METHOD AND MATERIALS
A retrospective, fully-crossed (two sessions >4 weeks apart), multi-reader multi-case (MRMC) study was performed. 240 cases
(100 cancers, 40 false positive recalls, 100 normals) were scored by 14 MQSA-qualified radiologists, once with and once without
using DS. For each case, a forced BI-RADS® score and a level of suspicion (1-100) were provided. When reading with the DS
system (Transpara, Screenpoint Medical, Nijmegen, The Netherlands), radiologists could activate the DS for a specific breast region
by clicking on it and the system then displayed a cancer likelihood score (1-100). Additionally, traditional computer-aided detection
was available to prompt calcification and soft tissue lesion markers. Area under the receiver operating characteristic curve (AUC),
specificity and sensitivity, and reading time were compared using MRMC Analysis of Variance.
RESULTS
On average, with the DS system, the AUC increased significantly from 0.866 to 0.886 (P=0.0019) compared to unaided reading.
Sensitivity increased from 83% to 86% (P=0.046), while specificity only slightly improved from 77% to 79% (P=0.061). Considering
lesion type, AUC increased for soft tissue lesions (0.886 to 0.902, P=0.033), and calcifications (0.878 to 0.898, not significant,
P=0.1021). Reading time per case was similar in both situations (unaided = 146 s, with DS = 149 s, P=0.147). As a stand-alone, the
computer system had an equal detection performance (AUC=0.887) than the average of radiologists (P=0.333).
CONCLUSION
Radiologists significantly improved their cancer detection in mammography when using a deep learning-based computer system for
decision support without taking more time.
CLINICAL RELEVANCE/APPLICATION
The use of decision support might prevent overlook and interpretation errors that are relatively common in the reading of
mammography. The increase in performance when concurrently using DS does not lengthen radiologists reading time per case, as
opposed to traditional computer-aided detection systems. The use of single-reading in combination with the computer system might
achieve a performance similar to double human reading considering that the stand-alone performance of the system is similar to the
average of radiologists.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Elizabeth A. Krupinski, PhD - 2017 Honored Educator
SSK02-04 Data-Driven Imaging Biomarker for Breast Cancer Screening in Mammography-Reader Study
W ednesday, Nov. 28 11:00AM - 11:10AM Room: E451B

Participants
Eun-Kyung Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hyo-Eun Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Employee, Lunit Inc
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Boo-Kyung Han, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jung Yin Huh, MD, Seoul, Korea, Republic Of (Presenter) employee, Lunit Inc.
Kyunghwa Han, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ekkim@yuhs.ac
PURPOSE
Previously, we demonstrated data-driven imaging biomarker in mammography (DIB-MMG; an imaging biomarker that is derived from
large-scale mammography data by using deep learning technology) for detection of malignant lesions. Now, we assess the feasibility
of DIB-MMG as a diagnosis-support-tool for radiologists.
METHOD AND MATERIALS

Total 96,191 exams of 4-view digital mammograms were retrospectively collected from two institutions. All cancer exams were
proven by biopsy. Benign exams were proven by biopsy or at least 1 year of follow-up mammography, and normal exams were
proven by at least 1 year of follow-up mammography. 90,637 exams of training data (16,086 cancer, 31,237 benign, and 43,314
normal exams) and 5,554 exams of test data (1,692 cancer, 2,780 benign, 1,082 normal cases) were used for developing the DIB-
MMG. Sensitivity, specificity, and AUC of the final DIB-MMG on the test data were 82.6%, 93.3%, and 0.94, respectively. Total 120
exams of mammograms (38 cancer and 82 non-cancer exams) were independently collected for reader study, and five radiologists
participated. For each exam, readers first read the exam without the help of DIB-MMG and Task-1) annotate the most suspicious
lesion with DMIST 7-pt scores and Task-2) decide recall or not per breast. After reading of each exam, readers modify their
decision based on the heat-map of DIB-MMG which denotes the likelihood of malignancy.
RESULTS
Per-breast standalone performance of DIB-MMG for 120 exams was 0.942 of AUC in Task-1, and 89.7% of sensitivity, 89.6% of
specificity in Task-2. Average performance of five radiologists without DIB-MMG was 0.807 of AUC in Task-1, and 70.8% of
sensitivity, 86.2% of specificity in Task-2. With DIB-MMG, the average performance was improved to 0.879 of AUC (p=0.024) in
Task-1, and 79.5% of sensitivity, 86.5% of specificity in Task-2. Fig.1 shows exemplary DIB-MMG heat-maps.
CONCLUSION

This retrospective reader study showed the potential of DIB-MMG as a diagnosis support tool for radiologists in breast cancer
screening. Further clinical validation with prospective study is needed.
CLINICAL RELEVANCE/APPLICATION

DIB-MMG is purely based on data-driven features from a large-scale mammography data instead of manually designed features of
conventional computer-aided detection (CAD) algorithms. With further clinical validation, DIB-MMG can be practically used as a
diagnosis support tool for radiologists in breast cancer screening.
SSK02-05 Generative Neural Network Inserting or Removing Cancer into Mammograms Fools Radiologists and
Deep Learning Alike: Example of an Adversarial Attack
W ednesday, Nov. 28 11:10AM - 11:20AM Room: E451B

Participants
Anton S. Becker, MD, Zurich, Switzerland (Presenter) Nothing to Disclose
Lukas Jendele, Oberengstringen, Switzerland (Abstract Co-Author) Nothing to Disclose
Ondrej Skopek, Oberengstringen, Switzerland (Abstract Co-Author) Nothing to Disclose
Soleen Ghafoor, MD, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Nicole Berger, MD, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Magda Marcon, MD, Zurich, Switzerland (Abstract Co-Author) Nothing to Disclose
Ender Konukoglu, Sophia Antipolis, France (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

anton@becker.md
PURPOSE

To investigate whether a cycle-consistent generative adversarial network (CycleGAN) can insert or remove cancer-specific
features into mammographic images in a realistic fashion.
METHOD AND MATERIALS
From two publicly available datasets (BCDR and INbreast) 680 mammographic images from 334 patients were selected, 318 of which
exhibited potentially cancerous masses, and 362 were healthy controls. We trained a CycleGAN, using two pairs of generator and
discriminator networks to convert cancerous breast images to healthy and back, and vice versa for the controls, without the need
for paired images. The network, implemented in TensorFlow, was trained for 40 epochs on an augmented dataset enlarged ten-fold
by random rotation, scaling, and contrast perturbations. To investigate how realistic the images appear, we randomly selected 20
image pairs of original and generated images, and 10 single images of each category (60 images in total). The images were
presented to three radiologists (5 and 3 years of experience, and PGY-5 resident) who rated them on a 5-point Likert-like scale and
had to indicate whether the image was real or generated/modified. The readout was analysed with a receiver-operating-
characteristics (ROC) analysis, performance was expressed as area under the ROC curve (AUC).
RESULTS
For the most experienced radiologist, the modifications introduced by CycleGAN reduced diagnostic performance, with the AUC
dropping from 0.85 to 0.63 (p=0.06), respectively, while the two less experienced ones seemed unaffected at a lower baseline
performance (AUC 0.75 vs. 0.77 and 0.67 vs. 0.69). None of the radiologists could reliably detect which images were real and which
were modified by CycleGAN (AUC 0.50-0.66).
CONCLUSION

CycleGAN can inject or remove malignant features into mammographic images while retaining their realistic appearance. These
artificial modifications may lead to false diagnoses.
CLINICAL RELEVANCE/APPLICATION

Modern adversarial attacks may go undetected by humans as well as deep learning algorithms, and could be used in cyber warfare.
It is vital to secure healthcare devices and information systems against such attacks mediated by neural networks.
SSK02-06 Deep Learning for Detection of Breast Cancer and Negative Screening Exams Using an In-House
Million Mammogram Dataset
W ednesday, Nov. 28 11:20AM - 11:30AM Room: E451B

Participants
Hari Trivedi, MD, San Francisco, CA (Presenter) Nothing to Disclose
Peter Chang, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Dmytro Lituiev, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
April Liang, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Maryam Panahiazar, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jae Ho Sohn, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Yunn-Yi Chen, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Benjamin L. Franc, MD, Sacramento, CA (Abstract Co-Author) Nothing to Disclose
Bonnie N. Joe, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Dexter Hadley, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

hari.trivedi@gmail.com
PURPOSE

Breast cancer is the second leading cause of cancer death in women in the US. Screening mammography is effective for early
detection, however suffers from unnecessary recall imaging and biopsies. Deep learning shows promise in medical image recognition
tasks, but requires large-scale, robustly-annotated datasets. We expand upon our previously described end-to-end process of
constructing a million mammogram dataset using routine clinical data and present results of two preliminary deep learning models for
cancer detection and the identification of true negative images.
METHOD AND MATERIALS
923,685 DICOM images and 37,730 free-text pathology reports were used to generate an in-house database labeled with ground-
truth pathology results.The first deep learning model was created for cancer detection only in biopsy proven specimens - the most
difficult subset of data as each image contained a suspicious finding. The model was comprised of two components: patch-based
pre-training and end-to-end fine tuning. Training set size was 34,390 images (12,251 positive, 22,139 negative), and test set size
was 6,778 images equally split.The second model was designed to have a high NPV for screening and diagnostic studies. An
attention-based object detection network was used, with potential abnormalities identified by a region-proposal network and
resolved by a separate head classifier network. The model was trained with 359,574 images (4,738 positive, 354,837 negative).
RESULTS
The first model achieved an AUC of 0.81, sensitivity of 0.764, and specificity of 0.797. The second model when tested on 100
positive and 100 negative cases achieved an AUC of 0.90, sensitivity of .866, and specificity of .873 . If the test cases were
changed to a more clinically relevant distribution of 99% benign and 1% cancer, the AUC increased to 0.96.
CONCLUSION

We demonstrate the efficacy of deep learning for mammography in both cancer detection and the identification of negative studies.
Future work includes enrichment of the dataset with further clinical data such as history of breast cancer, prior surgeries, and
hormone replacement therapy. We also aim to improve model performance and efficiency through novel model architectures.
CLINICAL RELEVANCE/APPLICATION

We develop novel deep learning models for mammography using routine clinical data from a single institution with the potential to
decrease recall imaging and unnecessary biopsies.
SSK02-07 Improved Cancer Detection using Artificial Intelligence: A Retrospective Evaluation of Missed Cancers
on Mammography
W ednesday, Nov. 28 11:30AM - 11:40AM Room: E451B

Participants
Alyssa T. Watanabe, MD, Manhattan Beach, CA (Presenter) Consultant, CureMetrix, Inc
Vivian Lim, MD, La Jolla, CA (Abstract Co-Author) Consultant, CureMetrix, Inc
Jenna I. Liu, MD, La Jolla, CA (Abstract Co-Author) Consultant, Curemetrix, Inc
Eric Weise, La Jolla, CA (Abstract Co-Author) Software developer, CureMetrix, Inc;
Chi Yung Chim, La Jolla, CA (Abstract Co-Author) Researcher, CureMetrix, Inc
William G. Bradley JR, MD, PhD, La Jolla, CA (Abstract Co-Author) Officer, CureMetrix, Inc
Christopher E. Comstock, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
alyssa90266@gmail.com
PURPOSE

To determine whether artificial intelligence-based (AI) software can be used to improve radiologists' sensitivity in breast cancer
screening and detection.
METHOD AND MATERIALS

A set of 2-D Digital Mammograms originally interpreted with R2 ImageChecker CAD (Hologic, Sunnyvale, CA) and performed between
October 2011 to March 2017 was collected from a community facility. Of the 317 cancer patients with available prior mammograms,
139 had retrospective findings, and 90 of those were deemed actionable. A blinded retrospective study was performed with a panel
of seven radiologists comprised of false negative actionable mammograms obtained up to 5.8 years prior to diagnosis and 32 normal
studies. Each radiologist viewed the cases without and then with benefit of cmAssist TM (CureMetrix, La Jolla, CA) AI based
computer-aided detection (AI-CAD) flags and neuScore TM (quantitative AI-based probability for malignancy of flagged lesions, 1-
100 scale). Reader decision making changes in true and false positive recalls with and without AI were analyzed.
RESULTS

All radiologists showed a significant improvement in their cancer detection rate (CDR) with the use of AI-CAD and neuScore (p
=0.0069, C.I. = 95%). With the assistance of AI software, the sensitivity of less experienced general radiologists improved to a
level higher than a fellowship-trained academic mammographer. The readers detected between 25% and 71% (mean 51%) of the
early cancers without assistance. With AI software results,, overall reader CDR was 41% to 76% (mean 62%). Overall, there was
less than 1% increase in the readers' false positive recalls with use of the AI software.
CONCLUSION

There was a statistically significant improvement in radiologists' sensitivity for cancer detection in this enriched data set of primarily
false negative mammograms with the benefit of the AI-CAD with neuScore. The percentage increase in CDR for the radiologists in
the reader panel, ranged from 6% to 64% (mean 27%) with the use of AI-CAD, with negligible increase in false positive recalls.
CLINICAL RELEVANCE/APPLICATION
This study shows a measurable, significant benefit for radiologists in mammography interpretation with the use of artificial
intelligence (AI) based computer-aided detection software with quantitative scoring. The use of AI in clinical practice may
potentially expedite workflow, enhance earlier detection of cancer, and reduce false negative mammograms.
SSK02-08 Data-Driven Imaging Biomarker for Breast Cancer Screening in Digital Breast Tomosynthesis
W ednesday, Nov. 28 11:40AM - 11:50AM Room: E451B

Participants
Sungwon Kim, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Hyo-Eun Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Employee, Lunit Inc
Jin Chung, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jee Eun Lee, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Minsung Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Employee, Lunit Inc
Eun-Kyung Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

dinbe@yuhs.ac
PURPOSE

To assess feasibility of a data-driven imaging biomarker in digital breast tomosynthesis (DIB-DBT) using the deep learning
technology and evaluate its potential for detection of breast cancer.
METHOD AND MATERIALS

We retrospectively collected 49,577 exams of 4-view digital mammograms (MMG) and 1,196 exams of 4-view digital breast
tomosynthesis images (DBT) from a single institution. We also collected 41 (10 cancer, 16 benign, 15 normal) exams of 4-view DBT
retrospectively from another institution for external validation. 49,577 exams of MMG consists of 47,719 (5,599 cancer, 17,971
benign, and 24,149 normal) and independent 1,858 (619 cancer, 620 benign, 619 normal) exams of training and validation data,
respectively. 1,196 exams of DBT consists of 996 (822 cancer, 40 benign, 134 normal) and independent 200 (120 cancer, 30
benign, 50 normal) exams of training and validation data, respectively. Previously, we assessed the feasibility of DIB-MMG as a
screening tool for breast cancer detection in mammograms through external validation and pilot reader study. Thus, we exploit DIB-
MMG for developing DIB-DBT in this study. Training of DIB-DBT consists of two stages - semi-supervised pre-training with partially-
annotated large-scale MMG followed by fully-supervised fine-tuning with fully-annotated small-scale DBT. Residual network for
image recognition is used as a baseline model. Diagnostic accuracy of DIB-DBT was assessed using receiver operating characteristic
analysis.
RESULTS

Area under the curve (AUC) on the internal validation dataset of DIB-DBT with and without the pre-training stage of DIB-MMG was
0.9227 and 0.9081, respectively. AUC of the external validation dataset of DIB-DBT with and without the pre-training stage of DIB-
MMG was 0.9710 and 0.9232, respectively.
CONCLUSION
This study showed the feasibility of DIB-DBT as a screening tool for breast cancer detection in DBT. This research also showed the
potential of DIB-MMG as a base model for DIB-DBT. Further clinical validation of DIB-DBT is needed for using it as a reliable
screening tool for breast cancer screening.
CLINICAL RELEVANCE/APPLICATION
With further clinical validation, DIB-DBT could be practically used as a second-reader to help radiologists detecting and diagnosing
breast cancer in DBT efficiently.
SSK02-09 Improved Performance of Machine Learning-Based Analysis of Mammography by Using Digital Breast
Tomosynthesis Versus 2D Mammography
W ednesday, Nov. 28 11:50AM - 12:00PM Room: E451B

Participants
Bill Lotter, Boston, MA (Abstract Co-Author) Officer, DeepHealth Inc
Jerrold L. Boxerman, MD, PhD, Providence, RI (Abstract Co-Author) Nothing to Disclose
A. Gregory Sorensen, MD, Belmont, MA (Presenter) Employee, DeepHealth, Inc; Board member, IMRIS Inc; Board member, Siemens
AG; Board member, Fusion Healthcare Staffing; Board member, DFB Healthcare Acqusitions, Inc; Board member, inviCRO, LLC;
For information about this presentation, contact:
sorensen@deep.health
PURPOSE
Digital Breast Tomosynthesis (DBT) has been shown to be clinically superior to both full-field digital mammography and synthetic
two-dimensional mammography (2D) for breast cancer detection. However, few studies to date have compared machine learning
(ML) algorithmic performance in DBT versus 2D in large data sets. Technically, the much larger size of a DBT acquisition could
actually be a hindrance for training convolutional neural networks (CNNs), for example via overfitting. Such technical issues could in
turn imply impracticality of ML for DBT or a need for much larger training datasets. We sought to implement CNNs for both DBT and
synthetic 2D X-ray mammograms and compare their performance.
METHOD AND MATERIALS

We compiled two separate datasets consisting of de-identified images and linked reports, collected from multiple mammography
centers following an IRB-approved protocol. Data originated from equipment from the same manufacturer across all sites, and
included presentation DBT and synthetic 2D images. We developed a novel CNN architecture and trained this model on the first
dataset consisting of 22,000 DBT studies (323 cases of confirmed malignancy), where radiology reports and MQSA outcome data
were used as estimates of ground truth. To simulate a more realistic evaluation scenario, the CNN was then tested on the second
dataset collected from a different center. Using a test set of 1,750 screening DBT studies (94 confirmed cancers), receiver
operating characteristic (ROC) curves and the corresponding area-under-the-curve (AUC) were calculated on both the full DBT
study, and on just the synthetic 2D data alone.
RESULTS
AUC values for performance on the test dataset were: 2D: 0.894. DBT: 0.915. (p < 0.01 for difference between 2D and DBT on the
full test dataset). At typical operating points (sensitivity 0.75 to 0.90) this corresponds to an average 19.6% relative decrease in
model callback rates for the model (e.g., at sensitivity=0.8, from ~15% to ~11%).
CONCLUSION
ML can be applied successfully to DBT and results in improved performance over synthetic 2D mammography.
CLINICAL RELEVANCE/APPLICATION
Machine learning could play an important role in screening mammography, not only for traditional 2D mammography, but also when
used with DBT; thus, ML is not in conflict with DBT but complementary and could further improve breast cancer screening
performance.
 VW 42

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Self-guided Reading
Sessions: Siemens Healthineers Vendor Workshop
W ednesday, Nov. 28 10:30AM - 5:00PM Room: Booth 5530

Program Information
You are invited to our self-guided reading sessions with syngo® Breast Care workstations configured especially to allow you to
work at your own place at a time that suits you! A series of breast tomosynthesis and contrast enhanced mammography cases
presented as problem cases with a solution enables you to develop and test your reading skills.
 VW 43

Automated Breast Volume Scanner (ABVS) Self-guided Reading Sessions: Siemens Healthineers Vendor
Workshop
W ednesday, Nov. 28 10:30AM - 5:00PM Room: Booth 5530

Program Information
With syngo® Ultrasound Breast Analysis (sUSBA) software, self-guided reading sessions with real clinical cases will enable you to
become familiar with the coronal plane while providing practical approaches to interpretation of 3D automated breast ultrasound.
 VW 67

Integrating Contrast Enhanced 2D Imaging into Your Practice to Optimize Patient Care: Hologic Vendor
Workshop
W ednesday, Nov. 28 10:30AM - 11:30AM Room: Booth 5524

Participants
Matthew Covington, MD, St. Louis, MO (Presenter) Nothing to Disclose
Program Information
Explore the benefits of adding contrast enhanced mammography to your practice. Discussion will highlight: • Cost effectiveness •
Patient contraindications • Alternative to bMRI (I-View™ Software, SecurView® Workstations)
Registration
https://hologicrsna.com
 ML42

Machine Learning Theater: AI-driven Mammography: Applying the Right Filter: Presented by Densitas, Inc.
W ednesday, Nov. 28 11:30AM - 11:50AM Room: Machine Learning Show case North Hall

Participants
Ryan Duggan, Halifax, NS (Presenter) Nothing to Disclose
Mohamed Abdolell, MSc, Halifax, NS (Presenter) Founder and CEO, Densitas Inc
PROGRAM INFORMATION
Densitas builds machine learning and AI solutions for breast screening that provide actionable information for radiologists and
technologists to help improve existing workflows and quality, and lead to better care for patients.
 VW 45

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
W ednesday, Nov. 28 11:40AM - 12:50PM Room: Booth 5530

Participants
Ingolf Karst, MD, Chicago, IL (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 VW 68

A Clinical Perspective on Increasing Confidence with Synthesized 2D Imaging Technology: Hologic Vendor
Workshop
W ednesday, Nov. 28 12:00PM - 1:00PM Room: Booth 5524

Participants
Jacqueline S. Holt, MD, Wilmington, DE (Presenter) Nothing to Disclose
Program Information
Clinical perspective of Synthesized 2D Imaging Technology to increase reading confidence. This session includes a facilitated review
of relevant cases.
Registration
https://hologicrsna.com
 BRS-W EA

Breast Wednesday Poster Discussions

W ednesday, Nov. 28 12:15PM - 12:45PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

Participants
Stamatia V. Destounis, MD, Scottsville, NY (Moderator) Research Grant, Hologic, Inc; Research Grant, Delphinus Medical
Technologies, Inc
Sub-Events
BR254-SD- Associations Between Magnetic Resonance Imaging (MRI) Biomarkers and Tumor-Infiltrating
WEA1 Lymphocytes (TILs) in Breast Cancer: Results from a Preliminary Study
Station #1
Participants
Elena Venturini, MD, Milan, Italy (Presenter) Nothing to Disclose
Marta Maria Panzeri, MD, Milan, Italy (Abstract Co-Author) Nothing to Disclose
Claudio Losio, MD, Milan, Italy (Abstract Co-Author) Nothing to Disclose
Silvia Ravelli, MD, Milan, Italy (Abstract Co-Author) Nothing to Disclose
Francesca Gallivanone, Milan, Italy (Abstract Co-Author) Nothing to Disclose
Pietro Panizza, Milan, Italy (Abstract Co-Author) Speaker, Koninklijke Philips NV; Research Grant, Koninklijke Philips NV
For information about this presentation, contact:
venturini.elena@hsr.it
PURPOSE
TILs reflect the attempt of the host immune system to eradicate malignancies and own an interesting prognostic value in breast
cancer. Our purpose is to assess the role of multiparametric breast MRI in the prediction of the presence of TILs.
METHOD AND MATERIALS
We retrospectively reviewed the examinations of patients who underwent a multiparametric breast MRI from 01/2015 to 04/2017
and fulfill the following inclusion criteria: invasive ductal carcinoma histotype, core needle/VABB performed at our Institute with TILs
evaluation, mass-like lesion at breast MRI. MRI protocol included T2, diffusion-weighted imaging (DWI) and dynamic contrast-
enhanced (DCE) study (1.5T). The immunohistochemical and histologic data were evaluated on core biopsies by experienced
pathologists. On T2 presence of intratumoral necrosis, pseudocapsule and perilesional edema was assessed. On subtracted images
presence of early peripheral and delayed rim enhancement (DRE) was recorded. DWI, DCE and T2 images were analyzed using OLEA
software; tumor VOIs were manually depicted slice by slice avoiding necrosis. The software generated histograms for each VOIs
and first order texture parameters were obtained. Univariate and multivariate regression analyses were performed.
RESULTS
The study population was composed by 45 women with 47 breast carcinomas (23 G1-G2 24 G3; 19/47 40% HER2- luminal, 23/47
26% HER2+, 16/47 34% triple negative). The presence of TILs was significantly associated to high tumor grade and molecular
subtype quale ? (p<0,05). A significant association was found between TILs and DRE (p<0,05). A correlation was found between
presence of TILs and lower mean ADC (Apparent Diffusion Coefficient) (p<0,05), higher ADC skewness and kurtosis (p<0,05).
Focusing on G3 and TN tumors the association between the presence of TILs and DRE or lower mean ADC remained significant
(p<0,05). At multivariate analysis, the variables associated to presence of TILs are tumor grade, DRE and mean ADC (p<0,05).
CONCLUSION
The results of our exploratory study provide a new insight into the relationships between MRI biomarkers and tumor environment.
The presence of TILs seems to be related to lower mean ADC, reflecting an increased cellularity in the tumoral stroma.
CLINICAL RELEVANCE/APPLICATION
Multiparametric MRI may provide new insights into breast cancer micro-environment suggesting the presence of Tumor-Infiltrating
Lymphocytes.
BR255-SD- Patient and Tumor Characteristics to Predict the Benefit of Pre-Operative Breast MRI: Results from a
WEA2 Machine Learning Approach at a High Volume Academic Center
Station #2
Awards
Student Travel Stipend Award
Participants
Safia H. Cheeney, MD, Seattle, WA (Presenter) Nothing to Disclose
Habib Rahbar, MD, Seattle, WA (Abstract Co-Author) Research Grant, General Electric Company
Daniel S. Hippe, MS, Seattle, WA (Abstract Co-Author) Research Grant, Koninklijke Philips NV; Research Grant, General Electric
Company; Research Grant, Canon Medical Systems Corporation; Research Grant, Siemens AG
Yifan Wu, MS, Seattle, WA (Abstract Co-Author) Nothing to Disclose
Christoph I. Lee, MD, Mercer Island, WA (Abstract Co-Author) Research Grant, General Electric Company; Investigator, General
Electric Company
Savannah C. Partridge, PhD, Seattle, WA (Abstract Co-Author) Research Grant, General Electric Company
PURPOSE
Pre-operative breast MRI (pMRI) is a powerful tool for identifying additional mammographically occult disease in women newly
diagnosed with breast cancer. However, it also prompts many unnecessary biopsies, which are costly and can lead to treatment
delays. Surgeons vary in their use of pMRI, often basing their decisions on breast density, age, and tumor type without evidence to
support these approaches. We sought to identify patient and tumor characteristics that can predict which patients benefit from
pMRI using a machine learning approach.
METHOD AND MATERIALS
In this IRB-approved retrospective study, we identified all patients who underwent pMRIs (1/2005-2/2015) less than 6 months after
a core needle biopsy (CNB) diagnosis of breast cancer from our prospectively populated MRI database linked to pathology
outcomes. CNBs that occurred after pMRI were classified by worst outcome (invasive>DCIS>high-risk>benign). Patient and tumor
features, including age, density, tumor type, grade, hormone receptor, HER2, and Ki-67 were also extracted from the database.
The LASSO machine learning algorithm was used to generate multivariate models using these factors to predict additional, true
positive (TP) (DCIS or invasive breast cancer), and false positive (FP) (benign or high-risk pathology) CNBs after pMRI. The
resulting models were internally validated using the bootstrap with overall performance summarized using the c-statistic.
RESULTS
1396 women underwent pMRI during the study period, and 30% underwent a pMRI-prompted CNB (13% TP, 17% FP). While women
with dense breasts more often underwent pMRI-prompted CNB (32% vs 26%, p=0.02) with greater FP CNB rates (19% vs 14%,
p=0.02), there was no significant difference in TP CNB rate (13% vs 12%, p=0.6). During multivariate analysis, the LASSO selected
age, density, and HER2 status for predicting additional CNBs and FP CNBs; however, overall prediction performance was low (c-
statistics 0.55 and 0.56, p<0.05). The LASSO did not find any factors with sufficient predictive value to create a model for
predicting TP CNBs.
CONCLUSION

Our study demonstrates that clinical features, including age, density, and tumor features are weak predictors of who will benefit
most from pMRI.
CLINICAL RELEVANCE/APPLICATION

Pre-operative breast MRI is not more or less likely to benefit patients based on patient age, breast density, or tumor characteristics
and should not be denied to patients solely on these factors.
BR256-SD- Peritumoral Fat Content Correlates with Histologic Prognostic Factors in Breast Carcinoma: Iterative
WEA3 Decomposition of Water and Fat with Echo Asymmetry and Least-Squares Emission (IDEAL) Study
Station #3
Participants
Sachi Hisanaga, MD,PhD, Fukuoka, Japan (Presenter) Nothing to Disclose
Takatoshi Aoki, MD, PhD, Kitakyusyu, Japan (Abstract Co-Author) Nothing to Disclose
Shohei Shimajiri, MD, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Masanori Hisaoka, MD, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Toshiyuki Nakayama, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Yukunori Korogi, MD, PhD, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Akitaka Fujisaki, MD, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Chihiro Chihara, MD, Kitakyushu, Japan (Abstract Co-Author) Nothing to Disclose
Yoshiko Hayashida, MD, Fukuoka, Japan (Abstract Co-Author) Nothing to Disclose
PURPOSE

Breast cancer cells express receptors for adipokines secreted by adipocytes, which can affect tumor growth. In vitro and in vivo
data indicate that adipocytes are modified by cancer cells to acquire characteristics different from naive adipocytes (cancer-
associated adipocytes: CAAs). Histologically, CAAs located around breast cancer display smaller sizes and are less lipid. The
purpose of this study is to correlate peritumoral fat content using IDEAL with histologic prognostic factors in breast carcinoma.
METHOD AND MATERIALS
This study consisted of 100 patients who were diagnosed with invasive carcinoma of breast and underwent breast MRI including
IDEAL before surgery. The scan time of IDEAL fat fraction (FF) map imaging was 23 sec. Four regions of interests (ROIs), which are
a distance of 5mm from the tumor edge, and 4 ROIs in the mammary fat of the healthy side were set on the FF map. Then average
peri-tumoral FF values (FT), average FF values in the healthy side (FH), and peri-tumoral fat ratio (pFTR: defined as FT/FH) were
calculated. Histologically, the presence of lymph node metastasis and the MIB-1 index were evaluated by 2 pathologists.
RESULTS

FT and pTFR for breast carcinoma with lymph node metastasis (79.27±10.36 and 0.897±0.078) were significantly lower than those
without (86.23±4.53 and 0.945±0.032) (p<.001 and p=.005). Spearman rank correlation suggested that the FT correlated with the
MIB1 index (r=-340, p=.001).
CONCLUSION
The peritumoral fat content calculated with IDEAL is associated with the histologic prognostic factors, and may therefore be a
useful prognostic biomarker for breast carcinoma.
CLINICAL RELEVANCE/APPLICATION
In vivo IDEAL imaging is simple to perform without extrinsic contrast agent and the quantification of the peritumoral FF using IDEAL
may be useful for therapeutic strategy for breast carcinoma.
BR257-SD- Cancer Detection Rate for Stereotactic Biopsies Performed on Initially Categorized BI-RADS 3
WEA4 Calcifications
Station #4
Participants
Monica Froicu, MD, Danville, PA (Presenter) Nothing to Disclose
Katherine Chung, Scranton, PA (Abstract Co-Author) Nothing to Disclose
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
For information about this presentation, contact:

mfroicu@geisinger.edu
PURPOSE

To determine the malignancy rate and indications for biopsy of calcifications initially categorized as BI-RADS 3 and converted to BI-
RADS 4/5 during the ensuing surveillance period.
METHOD AND MATERIALS
Following IRB approval, our Radiology Information System (RIS) was searched for all BI-RADS 4/5 categorized mammograms
performed from January 1, 2013 to July 30, 2016 that underwent stereotactic biopsy and had been classified as BI-RADS 3 for
calcifications within the prior 2 years. Data collected included: patient age, prior biopsies, personal/family history of breast cancer,
pathogenic mutations, breast density, calcification morphology, extent, distribution and increase in size or change in morphology
during surveillance. BI-RADS scores changed from 3 to 4/5 with no change in the grouping were categorized as difference in
interpretation.
RESULTS

The search identified 162 patients with mean age 55.7 (range 32-81) years. 55% (90/162) were heterogeneously dense, 15%
(25/162) had personal history and 17.8% (29/162) had family history of breast cancer, none had pathogenic mutations and 14%
(23/162) had prior benign biopsies. The average BI-RADS conversion time was 9.5 months with 41.9 % (69/162) changed at 6
months. Change was due to increasing calcifications in 42.3% (69/162), morphology change 9.8% (16/162), distribution change
2.4% (4/162) or difference in interpretation 16.6% (27/162). Distribution was 78.4% (127/162) grouped, 11.11% (18/162) regional,
10.5% (17/162) linear, 1.2% (2/162) segmental, 0.6% (1/162) diffuse. PPV at biopsy was 11.7 % (19/162) including 13 DCIS and 6
invasive carcinomas. There were 67.2% (109/162) benign and 20.9% (34/162) high risk lesions 2 of which were upgraded to low
grade DCIS at surgery.
CONCLUSION

Calcifications initially classified as BI-RADS 3 that increase in number or change in morphology during surveillance have a significant
upgrade rate to malignancy.
CLINICAL RELEVANCE/APPLICATION
BI-RADS 3 lesions that change during surveillance should undergo biopsy.
BR258-SD- Radiologists versus Deep Learning Model Inter-Observer Variability in Mammographic Breast Density
WEA5 Assessment
Station #5
Participants
Brian N. Dontchos, MD, Boston, MA (Presenter) Nothing to Disclose
Regina Barzilay, PhD, Cambridge, MA (Abstract Co-Author) Nothing to Disclose
Adam Yala, Cambridge, MA (Abstract Co-Author) Nothing to Disclose
Tal Schuster, Cambridge, MA (Abstract Co-Author) Nothing to Disclose
Kyle Swanson, Boston, MA (Abstract Co-Author) Nothing to Disclose
Manisha Bahl, MD,MPH, Boston, MA (Abstract Co-Author) Nothing to Disclose
Christine E. Edmonds, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Constance D. Lehman, MD,PhD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Medical Advisory
Board, General Electric Company
For information about this presentation, contact:

bdontchos@mgh.harvard.edu
PURPOSE
Qualitative breast density assessment by radiologists is highly variable and currently available automated quantitative tools are
limited. Our purpose was to evaluate the inter-observer variability amongst radiologists vs our deep learning (DL) model in
mammographic breast density assessment.
METHOD AND MATERIALS

Four expert radiologists independently assessed BI-RADS density on a random subset of 100 screening mammograms, blinded to the
original radiologist's assessment. Our DL model (previously reported) was trained and tested using a deep convolutional neural
network on 50,071 consecutive digital screening mammograms performed June 2009 to June 2014 to predict the original radiologist's
assessment. We estimated both our DL model agreement and our radiologists blinded reader agreement with the original radiologist's
density assessment using percent agreement with Wilson confidence intervals (CI) and with linear-weighted kappa statistics,
compared across 5,000 bootstrap samples to assess significance.
RESULTS
Radiologist agreement with the original assessment was 80.5% (95% CI 76.8, 84.5) for binary assessment (dense vs not dense) and
was 66.3% (95% CI 61.5, 70.9) for 4-category density assessment. Our DL model agreement for the same cases was 85.0% (95%
CI of 76.7, 90.7) for binary assessment and 71.0% (95% CI of 61.5, 79.0) (p>0.05 for each comparison). Compared to the original
density assessment, our DL model showed substantial agreement (K=.61, 95% CI .49-.74) compared to moderate agreement by
radiologists (K=.57, 95% CI .50- .63).
CONCLUSION

Our DL model can accurately and consistently assess breast density, particularly into clinically relevant dense vs non-dense
categories.
CLINICAL RELEVANCE/APPLICATION

Implementation of a deep learning model that accurately and consistently assesses breast density could lead to more precise
identification of women who might benefit from supplemental imaging.
BR259-SD- Convolutional Neural Network Based Breast Cancer Risk Stratification Using a Mammographic Dataset
WEA6
Station #6
Participants
Simukayi Mutasa, MD, New York, NY (Presenter) Nothing to Disclose
Peter Chang, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jenika Karcich, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Eduardo Pascual Van Sant, BS, New York, NY (Abstract Co-Author) Nothing to Disclose
Mary Q. Sun, MD, Manhasset, NY (Abstract Co-Author) Nothing to Disclose
Michael Z. Liu, MS, New York, NY (Abstract Co-Author) Nothing to Disclose
Sachin Jambawalikar, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Richard S. Ha, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

s.mutasa@columbia.edu
PURPOSE

We propose a novel convolutional neural network (CNN) based pixel-wise breast cancer risk model using a mammographic dataset.
METHOD AND MATERIALS

An IRB approved retrospective case-control study of 1474 mammographic images was performed in average risk women. First, 210
patients with new incidence of breast cancer were identified. Mammograms from these patients at least two years prior to
developing breast cancer were identified and made up the case group [420 bilateral craniocaudal (CC) mammograms]. The control
group consisted of 527 patients without breast cancer from the same time period. Prior mammograms from these patients made up
the control group [1054 bilateral CC mammograms]. A CNN architecture was designed for pixel-wise breast cancer risk prediction.
Briefly, each mammogram was normalized as a map of z-scores and resized to an input image size of 256x256. Then a contracting
and expanding fully convolutional CNN architecture was composed entirely of 3x3 convolutions, a total of four strided convolutions
instead of pooling layers, and symmetric residual connections. L2 regularization and augmentation methods were implemented to
prevent overfitting. Cases were separated into training (80%) and test sets (20%). Fivefold cross validation was performed.
RESULTS
The average age of patients between the case and the control groups was not statistically different [case: 57.4 years (SD, 10.4)
and control: 58.2 years (SD, 10.9), p=0.33]. Mammographic breast density (BD) was significantly higher in the case group [2.39
(SD, 0.7)] than the control group [1.98 (SD, 0.75), p<0.0001]. On multivariate logistic regression analysis, both CNN pixel-wise
mammographic risk model and BD were significant independent predictors of breast cancer risk (p < 0.0001). The CNN risk model
showed greater predictive potential [OR=4.42 (95% CI, 3.4-5.7] compared to BD [OR =1.67 (95%CI, 1.4 -1.9). The CNN risk model
achieved an overall cross validation accuracy of 72% (95%CI, 69.8-74.4%) in predicting patients in the case group.
CONCLUSION

A novel pixel-wise CNN architecture can stratify breast cancer risk in mammography, independent of the BD. A larger dataset will
likely improve our model.
CLINICAL RELEVANCE/APPLICATION

Personalized breast cancer risk stratification may be aided by using a novel pixel-wise CNN model. This may have clinical
implications in screening guidelines.
BR260-SD- The Significance of Asymmetries in Screening Mammograms with Digital Breast Tomosynthesis
WEA7
Station #7
Awards
Student Travel Stipend Award

Participants
Yarisma Frometa, MD, New Haven, CT (Presenter) Nothing to Disclose
Maryam Etesami, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Liva Andrejeva-Wright, MD, New Haven, CT (Abstract Co-Author) Nothing to Disclose
Liane E. Philpotts, MD, New Haven, CT (Abstract Co-Author) Consultant, Hologic, Inc
For information about this presentation, contact:
yarisma.frometa@yale.edu
PURPOSE

Digital breast tomosynthesis (DBT) can result in significantly decreased recall rate (RR) of screening mammograms. However,
asymmetries remain one of the most frequently recalled abnormalities and a major source of false positives. The purpose of this
study is to determine the significance and characteristics of asymmetries detected in DBT screening and their diagnostic work up to
aid in further reducing false positive findings on DBT.
METHOD AND MATERIALS

In this retrospective Institutional Review Board approved study, we reviewed all DBT screening mammograms performed from
10/1/2014 through 9/30/2016 at our academic center. The number of recalled exams and type of recalled abnormalities were
identified. For each recalled asymmetry, imaging characteristics, diagnostic work up, biopsy, and imaging follow up were
documented. RR and positive predictive value (PPV1, PPV2, and PPV3) of asymmetries were calculated. The characteristics of true
and false positive asymmetries were compared.
RESULTS

Of 15620 DBT screening mammograms, 1100 exams were recalled for further evaluation (overall RR=7.04%). 482 asymmetries were
recalled in 440 exams (asymmetry RR=3.09%). The false positive rate of recalled asymmetries was 97.93% with only 10 true
positive cancers (asymmetry PPV1=2.07%). There was a significant difference in cancer yield of "developing" asymmetries (7%)
versus asymmetries not otherwise specified (1.2%) (p=0.002). Focal asymmetries had higher cancer yield (3.3%) compared to
asymmetries seen only on one view (1.3%), but the difference was not significant (p=0.19). The majority of true positive
asymmetries (9 out of 10) had a suspicious ultrasound correlate at the time of first diagnostic work up and underwent ultrasound
guided biopsy yielding invasive carcinoma. Only one case of ductal carcinoma in situ did not have an immediate sonographic
correlate.
CONCLUSION
Asymmetries recalled on DBT screening have a high false positive rate. Identifying asymmetries with a higher cancer yield such as
developing asymmetries may further reduce RR while preserving the cancer detection rate. In addition, our findings suggest that
follow up imaging of recalled asymmetries without suspicious ultrasound correlates may not be necessary.
CLINICAL RELEVANCE/APPLICATION

Given the relatively high RR and low PPV1 of asymmetries on DBT screening, efforts can be made towards reducing recall of benign
asymmetries and avoiding unnecessary follow up.
BR199-ED- The Forgotten Male Breast: A Comprehensive Review of Male Breast Disease
WEA8
Station #8
Awards
Certificate of Merit

Participants
Cheryce P. Fischer, MD, Santa Monica, CA (Presenter) Nothing to Disclose
Cecil Patel, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Melissa M. Joines, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
Brian L. Dubin, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

cpfischer@mednet.ucla.edu
TEACHING POINTS

Although males make up a small percentage of our clinical practice, male breast disease is important to recognize and be familiar
with. Gynecomastia is the most common disease affecting males. However, there is a wide spectrum of other conditions which can
also occur in the male breast. Two major teaching points the learner should accomplish by viewing this exhibit is: 1) To become
familiar with the appearance and work up of gynecomastia 2) To become familiar with the numerous other conditions that affect
the male breast
TABLE OF CONTENTS/OUTLINE

This is an educational exhibit demonstrating the spectrum of disease in the male breast. The exhibit will include the following cases
seen in the male breast at our institution and will be presented with a brief history and pertinent mammographic and sonographic
images. 1. Unilateral gynecomastia 2. Bilateral gynecomastia 3. Pseudogynecomastia 4. Epidermal inclusion cyst 5. Abscess 6.
Hematoma 7. Lipoma 8. Hibernoma 9. Intraductal papilloma and ductal carcinoma in situ 10. Papillary carcinoma 11. Metastatic
infiltrating ductal carcinoma 12. Infiltrating ductal carcinoma and ductal carcinoma in situ 13. Lymphoma 14. Transgender male to
female breast after hormone replacement therapy
BR200-ED- What the Breast Radiologist Needs to Know About Imaging Transgender Patients
WEA9
Station #9
Awards
Identified for RadioGraphics

Participants
Ujas N. Parikh, MD, New York, NY (Presenter) Nothing to Disclose
Elizabeth V. Mausner, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Chloe M. Chhor, MD, Brooklyn, NY (Abstract Co-Author) Nothing to Disclose
Yiming Gao, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Samantha L. Heller, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ujas.parikh@nyumc.org
TEACHING POINTS

There is increasing awareness of the need for optimal breast imaging approaches for transgender individuals, as well as for
evidence-based understanding of transgender breast cancer risk. In this case-based review, we will focus on screening and
diagnostic imaging scenarios involving transgender individuals and will also explore what is known about breast cancer risk and
breast cancer detection in the transgender population.
TABLE OF CONTENTS/OUTLINE

1. Intro a. Define "transgender" i. Terminology: transgender, transsexual, male-to-female (MtF), female-to-male (FtM)2.
Mammographic screeninga. Review of literature/paucity of datab. Screening and risk factors i. Exogenous Hormone use (MtF)ii.
Breast densityiii. Residual breast tissue (FtM)iv. Risk associated with natal sexv. High risk mutations and family historyc. Screening
and suggested guidelinesi. UCSF Center of Excellence for Transgender Healthii. Fenway Guide3. Diagnostic imaging a. Review of the
literatureb. Diagnostic scenarios with attention to pearls and pitfalls i. Hormone related symptoms and imagingii. Altered breast 1.
Augmentation/silicone injections (MtF)2. Post-mastectomy (FtM)4. Special considerations and challengesa. Healthcare access and
public educationb. Patient and physician awareness
BR201-ED- Unilateral Palpable Male Breast Lump: What Is Underneath!
WEA10
Station #10
Participants
Pramod K. Gupta, MD, Plano, TX (Presenter) Nothing to Disclose
Soume D. Foshee, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
Francisco Garcia-Morales, MD, Plano, TX (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

pramod.gupta@va.gov
TEACHING POINTS
Broad range of lesions from benign to malignant can present in male breast as palpable mass. The purpose of this scientific exhibit
is: 1. To describe the imaging findings of various lesions that present as unilateral masses. 2. To help the viewer learn an imaging
pattern-based approach to develop a reasonable differential diagnosis and in many cases make the specific diagnosis. 3. To
familiarize the viewer with those imaging findings which can be definitely considered benign, require no further work-up and to
familiarize from those findings which should be considered sufficiently suspicious or indeterminate to warrant further work-up.
TABLE OF CONTENTS/OUTLINE
The contents of this exhibit will be organized as follows with short discussions, illustrated examples and images: 1. Male breast
anatomy 2. Imaging work-up Algorithm 3. Categories: A. Skin and subcutaneous tissues origin: Sebaceous cyst/epidermal inclusion
cyst -Lipoma -Angiolipoma -Hematoma -Fat necrosis -Oil cyst B. Glandular and Stromal elements origin: -Gynecomastia -
Pseudogynecomastia -Pseudoangiomatous stromal hyperplasia -Myofibroblastoma -Hemangioma -Mason tumor -Schwanomma -
Granular cell tumor -Fibroadenoma -Abscess C. Malignancies: -Invasive ductal carcinoma -Recurrent breast carcinoma -Lymphoma -
Angiosarcoma -Metastasis
BR202-ED- Multimodality Imaging of Lobular Neoplasia: Review of Imaging Features and Updates on
WEA11 Controversies in Management
Station #11
Participants
Linda Ratanaprasatporn, MD, Boston , MA (Presenter) Nothing to Disclose
Lisa Ratanaprasatporn, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Sona A. Chikarmane, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Beth T. Harrison, Boston, MA (Abstract Co-Author) Nothing to Disclose
Eva C. Gombos, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
lratanaprasatporn@partners.org
TEACHING POINTS
Atypical lobular hyperplasia (ALH) and lobular carcinoma in situ (LCIS) are noninvasive proliferations of lobular cells. The term
'lobular neoplasia' (LN) is used to encompass both ALH and LCIS. LN is not an obligate precursor of invasive carcinoma. However, LN
is associated with increased risk of developing invasive carcinoma or ductal carcinoma in situ in either breast. Usually incidental
findings at core needle biopsy, LN has no reliable imaging features attributable to them. Knowing the most common imaging findings
of LN on mammography, ultrasound, and MRI is important to understand if there is imaging-histologic concordance or discordance as
this will affect management. The management of incidental ALH and LCIS at core biopsy is controversial and ranges from immediate
surgical excision to monitoring with clinical and imaging evaluation.
TABLE OF CONTENTS/OUTLINE

Review the definition of ALH and LCIS and their distinction on histopathology. Describe LN's natural history, its clinical significance,
and its associated increased risk of developing invasive cancer and DCIS. Demonstrate the most common imaging findings of LN on
mammography, ultrasound, and MRI using a case-based approach. Examine the different strategies and controversies for
management of LN including surgical excision, chemoprevention, and monitoring with imaging.
BR203-ED- Inserting Colors into the Diagnosis: When Doppler Makes the Difference in Breast Ultrasound
WEA12
Station #12
Participants
Cecilia S. Goldman, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana H. Catani, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Nestor Barros, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Pedro Henrique Hasimoto e Souza, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Flavia T. Horigome, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carlos Shimizu, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Tomie H. Ichihara, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Carla C. Caravatto, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Heni D. Skaf, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Caio D. Pinheiro, MBBS, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rafael L. Macedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

1- Review types of lesion vascularity and their blood flow characteristics 2-Describe the importance of color Doppler to distinguish
solid and cystic lesions, including complicated and complex cysts, as well as assessing vascular lesions.3- Correlate the amount and
distribution of blood-flow in breast lesions and their tendency to malignancy.4- Describe how color and pulsed-wave Doppler can
influence BIRADS classification as an associated feature, and consequently the patient's management and prognosis.
TABLE OF CONTENTS/OUTLINE

1. Doppler as an associated feature in the BI-RADS classification2. Imaging aspects of lesion vascularity and its assessment by
power/ color Doppler, and pulsed-wave Dopplerinternal and rim-like vascularitynumber of vesselsblood-flow velocityintra-tumoral
vessel resistance3. Applicability of color Doppler in the evaluation of breast lesionsCorrelation of blood flow and lesion
suspiciousnessCan vascularity evidence in a breast lesion change its BI-RADS category?4. Illustration of vascular breast lesions,
including:Cystic lesionsSolid lesionsVascular lesionsNon-mass lesions
 BRS-W EB

Breast Wednesday Poster Discussions

W ednesday, Nov. 28 12:45PM - 1:15PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

Participants
Stamatia V. Destounis, MD, Scottsville, NY (Moderator) Research Grant, Hologic, Inc; Research Grant, Delphinus Medical
Technologies, Inc
Sub-Events
BR261-SD- Utility of Targeted Ultrasound for the Evaluation of Palpable Breast Symptoms in Breastfeeding
WEB1 Women
Station #1
Participants
Amie Y. Lee, MD, San Francisco, CA (Presenter) Nothing to Disclose
Anna Knobel, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Kimberly M. Ray, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Heather I. Greenwood, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Bonnie N. Joe, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jessica H. Hayward, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
PURPOSE
To determine the accuracy of targeted breast ultrasound as the primary imaging modality for the evaluation of palpable symptoms
in breastfeeding women.
METHOD AND MATERIALS
A retrospective database review identified all breastfeeding women with a palpable symptom evaluated by targeted breast
ultrasound over a 15-year period (1/1/2000 -1/1/2015) at an academic facility. No patients were pregnant at time of imaging. Each
palpable site was designated as a case and all analyses were performed at the case level. BI-RADS assessments and lesion
characteristics were obtained from review of the radiology reports. Malignant outcomes were determined by pathology results.
Benign outcomes were determined by biopsy or >2 years clinical or imaging follow-up. Descriptive statistics and 2x2 contingency
table analyses were performed.
RESULTS
The final study cohort consisted of 139 palpable cases in 119 patients. Mean age was 36, and 24% were at least 40-years-old.
Sixty-three (45%) underwent targeted ultrasound alone and 76 (55%) underwent mammography in addition to ultrasound. On
ultrasound, 53 had no sonographic correlate (BI-RADS 1), 41 had benign findings (BI-RADS 2), and 3 were probably benign (BI-RADS
3). In the 42 positive ultrasound cases, 39 were suspicious (BI-RADS 4) and 2 were highly suggestive of malignancy (BI-RADS 5).
The most common biopsy results were galactocele, lactating adenoma, and lactational change. Frequency of malignancy was 1.4%
(n=2), and all malignancies were assessed as BI-RADS 5. Targeted ultrasound had high sensitivity (100%) and NPV (100%), but
limited specificity (71%) and low PPV2 (5%). In cases with mammography, 95% had heterogeneously or extremely dense breasts.
The addition of mammography yielded no additional cancers, and mammography missed one sonographically identified malignancy.
Mammography detected 7 false positive lesions unrelated to the palpable symptom.
CONCLUSION
Targeted ultrasound detected all malignancies in breastfeeding patients with palpable symptoms. The addition of mammography
increased false positives and yielded no additional cancers. Our results suggest that targeted ultrasound alone may be sufficient for
evaluation of symptomatic breastfeeding women.
CLINICAL RELEVANCE/APPLICATION
Current literature on breast imaging for symptomatic lactating women is sparse, and practice patterns vary widely. Our results
support the use of ultrasound as the primary imaging modality in this setting.
BR262-SD- Ductal Carcinoma in Situ Detected on Ultrasound Only Showed More Favorable Features Than Ductal
WEB2 Carcinoma in Situ Detected on Mammography in Asymptomatic Women with Dense Breasts
Station #2
Participants
Hee Jung Moon, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Eun-Kyung Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Min Jung Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jung Hyun Yoon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Vivian Y. Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
artemis4u@yuhs.ac
artemis4u@yuhs.ac
PURPOSE
To compared the clinical and pathological characteristics of ductal carcinoma in situ detected on mammography and ultrasound
(US) in asymptomatic patients with dense breasts.
METHOD AND MATERIALS
From February 2014 to September 2016, 236 asymptomatic patients with primary pure ductal carcinoma in situ and dense breasts
were included. The patients were classified into two groups. The mammography group (n=165) included patients with ductal
carcinoma in situ detected on mammography, and the US group (n=71) included patients with ductal carcinoma in situ detected on
US only. Clinico-pathological characteristics were compared between the two groups. Subgroup analyses were performed with a
cut-off age of 50 years and a cut-off tumor size of 20 mm.
RESULTS

Tumor size was significantly smaller in the US group (11.8±9.9 mm vs. 17.9±13.8 mm, p<0.001). Younger age, smaller tumor size,
low nuclear grade, no comedo necrosis, and progesterone receptor positivity were observed more in the US group (p<0.05). HER2
and Ki67 positivity were observed more in the mammography group (p<0.05). Similar results were found in 168 patients with ductal
carcinoma in situ < 20 mm. Regardless of the patient age, smaller tumor size, low nuclear grade, and no comedo necrosis were
observed significantly more in the US group.
CONCLUSION
Ductal carcinoma in situ in the US group showed significantly more low nuclear grade, no comedo necrosis, and hormone receptro
positivity, while HER2 and Ki67 positivity were observed significantly more in the mammography group. Ductal carcinoma in situ
detected on US only showed more favorable prognostic features than ductal carcinoma in situ detedted on mammography in
asymptimatic patients with dense breasts.
CLINICAL RELEVANCE/APPLICATION

Ductal carcinoma in situ deteted on US only in asymptomatic patients showed nore favorable prognostic factors than ductal
carcinoma in situ detected on mammography.
BR263-SD- Is the Contrast Enhanced Mammography an Alternative to Magnetic Resonance Imaging for the
WEB3 Presurgical Evaluation of Tumor Response in Breast Cancer Patients Treated with Neoadjuvant
Chemotherapy?
Station #3
Participants
Angela Iglesias Lopez, MD, A Coruna, Spain (Presenter) Nothing to Disclose
Joaquin J. Mosquera Oses, MD, La Coruna, Spain (Abstract Co-Author) Nothing to Disclose
Jose Ramon Varela Romero, MD, La Coruna, Spain (Abstract Co-Author) Nothing to Disclose
Alberto Bouzon Alejandro, A Coruna, Spain (Abstract Co-Author) Nothing to Disclose
Andres Vega Chaves, MD,MD, A Coruna, Spain (Abstract Co-Author) Nothing to Disclose
Diego Dominguez Conde, MD, A Coruna, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
a-iglesias@hotmail.com
PURPOSE
The neoadjuvant chemotherapy (NAC) is the standard therapeutic strategy for locally advanced breast cancer, allowing to increase
the breast conservation rate and it is increasingly being used for patients with early stage breast cancer. The magnetic resonance
imaging (MRI) is the most accurate diagnostic tool for assess the residual invasive disease after NAC. The Contrast Enhanced
Mammography (CE2D) is a recent imaging method that allows to show if tumor neovascularization exists in the breast parenchyma.
To evaluate the CE2D diagnostic accuracy to assess the residual disease extension of breast cancer patients receiving NAC.
METHOD AND MATERIALS

A prospective CE2D study was performed in 43 breast cancer patients treated with NAC from March to December 2017. All patients
underwent a CE2D and breast MRI after NAC. The residual tumor size determined by both techniques was correlated with the
pathological tumor size of the specimen.
RESULTS
CE2D showed a superior interclass correlation coefficient than MRI (0.9 vs. 0.7). The CE2D sensibility and specificity (83,9% and
83,3% respectively) were high and comparable to the MRI ones (74,2% and 91,6% respectively). Besides, the negative predictive
value of the CE2D was higher than the MRI one (66,7% vs. 57,9%) and the positive predictive value was similar (92,9% vs. 95,8%).
All these results could lead us to reduce the overall costs of diagnostic tests during the follow-up of breast cancer patients in the
neoadjuvant setting. CE2D could even replace pre- and post-treatment MRI studies, particularly for claustrophobic patients.
CONCLUSION

CE2D can replace MRI for the presurgical assessment of residual tumor size as well as on additional breast lessions visualization in
the ipsilateral breast or in the contralateral one, and therefore leading to costs savings and better accessibility.
CLINICAL RELEVANCE/APPLICATION
The CE2D could be an effective alternative to MRI for the residual tumor evaluation after neoadjuvant chemotherapy in breast
cancer patients. This technique could significantly save costs and improve patient flows. Furthermore, CE2D has almost no
contraindications.
BR264-SD- Microcalcifications of Ductal Carcinoma in Situ of the Breast: Correlation between Breast Imaging and
WEB4 Pathological and Biological Features
 Station #4
Participants
Eunji Lee, MD , Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Yun Woo Chang, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jinah Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hwa jin Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jiyoung Hwang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Seong Sook Hong, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
ywchang@schmc.ac.kr
PURPOSE
To evaluate the correlation between microcalcifications of ductal carcinoma in situ (DCIS) visible in breast imaging and pathological
and biological features.
METHOD AND MATERIALS
From December 2003 to December 2016, we retrospectively reviewed the mammography findings of 159 lesions of 155 patients who
were diagnosed with pure DCIS, and comparable sonographic images of 150 lesions of 146 patients, according to the Breast Imaging
Reporting and Data System (BI-RADS) lexicon. Sonographic findings of microcalcifications were divided into three groups compared
with mammographic microcalcifications: G1 (MMG -, US -), G2 (MMG +, US -) and G3 (MMG +, US +). Pathological findings (nuclear
grades, comedo necrosis) and biological features (ER positive group, HER2 positive group, triple negative group, Ki-67 index) were
compared with mammography and sonographic features. P values less than 0.05 were considered statistically significant.
RESULTS
Mammographic microcalcifications were observed in 90 out of 159 lesions (56.6%) of DCIS. Presence of microcalcifications on
mammography was significantly more common in high nuclear grade (p=0.001) and comedo necrosis (p=0.001). Fine pleomorphic,
fine linear or linear branching microcalcifications is the only morphologic feature that is significantly associated with nuclear grade
(p=0.004). Presence of microcalcifications on mammography were significantly associated with ER-negative group (p=0.010), HER2-
positive group (p=0.025), and increased Ki 67 index (p=0.001) respectively. Lesions with no visible microcalcifications in ultrasound
(G1+G2) were 93 out of 150 lesions (62%) and 57 out of 150 lesions (38.0%) showed microcalcifications in ultrasound (G3).
Calcification outside of a mass was the most common feature of sonographic microcalcification (36/150, 25.4%). Presence of
microcalcifications in the ultrasound were associated with high nuclear grade (p=0.001), comedo necrosis (p=0.001), ER-negative
group (p=0.028), HER2-positive group (p=0.028) and high Ki-67 index (p=0.001).
CONCLUSION
Microcalcifications of DCIS visible on mammography and ultrasound showed statistically significant association with poor
pathological and biological features.
CLINICAL RELEVANCE/APPLICATION
When DCIS, precursor of invasive breast cancer, developing to invasive cancer, ER and HER2 status and breast cancer subtype are
usually maintained and emphasize the evolution of breast cancer subtype specificity.
BR265-SD- Potential of Deep Learning and Conventional Radiomics in the Task of Distinguishing Between
WEB5 Malignant and Benign Breast Lesions in a Large Clinical MRI Dataset from China
Station #5
Participants
Hui Li, PHD, Chicago, IL (Presenter) Nothing to Disclose
Yu Ji, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Natalia O. Antropova, Chicago, IL (Abstract Co-Author) Nothing to Disclose
John Papaioannou, MSc, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc
Alexandra V. Edwards, Chicago, IL (Abstract Co-Author) Research Consultant, QView Medical, Inc; Research Consultant,
Quantitative Insights, Inc
Peifang Liu, MD, PhD, Tianjin, China (Abstract Co-Author) Nothing to Disclose
Maryellen L. Giger, PhD, Chicago, IL (Abstract Co-Author) Stockholder, Hologic, Inc; Shareholder, Quantitative Insights, Inc;
Shareholder, QView Medical, Inc; Co-founder, Quantitative Insights, Inc; Royalties, Hologic, Inc; Royalties, General Electric
Company; Royalties, MEDIAN Technologies; Royalties, Riverain Technologies, LLC; Royalties, Mitsubishi Corporation; Royalties,
Canon Medical Systems Corporation
For information about this presentation, contact:

huili@uchicago.edu
PURPOSE

To evaluate the potential of combining deep learning and lesion-based radiomic methods in the task of distinguishing between
malignant and benign breast lesions in a large MRI dataset from China.
METHOD AND MATERIALS
Our research involved a HIPAA-compliant, DCE-MRI database of 600 breast cases [average patient ages of the 300 benign and 300
malignant patients were 41.8 and 47.2 years with a standard deviation of 9.5 and 9.6 years, respectively]. MRIs had been obtained
using gadodiamide-enhanced T1-weighted spoiled gradient-recalled acquisition in the steady state sequence. A breast radiologist
located the lesions on the MRIs as input to our artificial intelligence workflow, which subsequently conducted two analyses. In the
first analysis, each lesion was automatically segmented from the surrounding parenchyma, and lesion phenotypes were extracted
serving as "lesion-segmented radiomics", including categories of size, shape, morphology, enhancement texture, kinetics, and
enhancement-variance kinetics. In the second analysis, transfer learning with a pre-trained deep convolutional neural network
(CNN), VGGNet, allowed for the extraction of features derived directly from the MRI data, yielding CNN-based radiomics. Each
method was investigated separately and in combination in the task of distinguishing between malignant and benign lesions, with
area under the ROC curve (AUC) serving as the figure of merit.
RESULTS

Both methods yielded promising classification performances with AUC values of 0.88 (se=0.01) and 0.84 (se=0.02) for the lesion-
segmented radiomics and the CNN-based radiomics methods, respectively. Combination of the two methods enhanced the
performance in malignancy assessment resulting in an AUC value of 0.90 (se=0.01), a statistically significant improvement over the
performance of the lesion-segmented radiomics method alone (p=0.0017).
CONCLUSION

Deep learning and lesion-segmented radiomics methods provide different information to the diagnostic classification task as
demonstrated by both yielding high but different performance levels, thus promoting the use as a combined predictor of malignancy.
CLINICAL RELEVANCE/APPLICATION
Our computerized combination of deep learning and segmentation-based radiomics CADx methods has potential to help radiologist
improve breast cancer diagnostic accuracy.
BR266-SD- Performance of Screening Breast MRI According to Different Risk Categories
WEB6
Station #6
Participants
Min Sun Bae, MD, New York, NY (Presenter) Nothing to Disclose
Janice S. Sung, MD, New York, NY (Abstract Co-Author) Research Grant, Hologic, Inc
Christopher E. Comstock, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

minsunb@gmail.com
PURPOSE

To evaluate screening breast MRI performance according to different risk categories in women at elevated risk of breast cancer
METHOD AND MATERIALS

This retrospective IRB-approved, HIPPA-compliant study included all screening breast MRI examinations performed between 2010
and 2012 in 5039 women (9208 examinations). Risk category were categorized as follows: (1) known BRCA1 or 2 mutation or
untested first degree relative of known BRCA carrier (BRCA), (2) history of chest radiation (CR), (3) high (>=20%) lifetime risk
(HLR), (4) personal history of breast cancer (PH), and (5) low (<20%) lifetime risk (LLR). Recall rate (BI-RADS assessment 0, 3, 4,
or 5), cancer detection rate (CDR), positive predictive value of biopsies performed (PPV3), sensitivity, and specificity were
compared for the risk categories.
RESULTS

Of the 9208 MRI examinations, there were 5310 (57.7%) in PH, 2442 (26.5%) in HLR, 973 (10.5%) in BRCA, 257 (2.8%) in LLR, and
226 (2.5%) in CR. 138 cancers were diagnosed, 114 at MRI, 19 at mammography, and 5 interval cancers. Recall rate ranged from
8.6% to 15.6% and the highest recall rate was found in the BRCA group (15.6%; 95% CI: 13.4%, 18.1%). The CDR was highest in
the BRCA group (19.5/1000), followed by CR (17.7/1000), HLR (14.3/1000), PH (10.2/1000), and LLR (7.8/1000). The PPV3 was
highest in the CR group (36.4%; 95% CI: 10.9%, 69.2%) and lowest in the LLR group (16.7%; 95% CI: 2.1%, 48.4%). Overall
sensitivity was 84.1% (95% CI: 76.9%, 89.7%) for all cancers and 84.8% (95% CI: 76.2%, 91.3%) for invasive cancers. Sensitivity
was lowest in the BRCA group (79.2%; 95% CI: 57.8%, 92.9%) and higher in the PH group (88.7%; 95% CI: 78.1%, 95.3%) and
the LLR group (100%). Specificity was highest in the PH group (92.5%; 95% CI: 91.7%, 93.2%) and lowest in the BRCA group
(86.1%; 95% CI: 83.7%, 88.2%).
CONCLUSION

MRI performance differed among women with various risk categories for breast cancer. Recall rate and CDR were highest and
sensitivity lowest in the BRCA group. Performance of screening breast MRI in women with PH was similar to that in other high-risk
women.
CLINICAL RELEVANCE/APPLICATION

Screening MRI may be possibly expanded to women at intermediate risk of breast cancer such as a personal history of breast
cancer or high-risk lesion.
BR267-SD- Multi-reader Multi-case Virtual Clinical Trial of Lesion Detection in Digital Mammography and Digital
WEB7 Breast Tomosynthesis
Station #7
Participants
Andrew D. Maidment, PhD, Philadelphia, PA (Presenter) Research support, Hologic, Inc; Research support, Barco nv; Research
support, Analogic Corporation; Spouse, Employee, Real-Time Tomography, LLC; Spouse, Stockholder, Real-Time Tomography, LLC;
Scientific Advisory Board, Real-Time Tomography, LLC;
Predrag R. Bakic, PhD, Philadelphia, PA (Abstract Co-Author) Research collaboration, Barco nv; Research collaboration, Hologic, Inc;
Bruno Barufaldi, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Ali N. Avanaki, PhD, Beaverton, OR (Abstract Co-Author) Employee, Barco nv
Kathryn S. Espig, MSc, Beaverton, OR (Abstract Co-Author) Employee, Barco nv
Albert Xthona, BS, Beaverton, OR (Abstract Co-Author) Employee, Barco nv
Susan Weinstein, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Tom Kimpe, Kortrijk, Belgium (Abstract Co-Author) Employee, Barco nv
For information about this presentation, contact:

predrag.bakic@uphs.upenn.edu
predrag.bakic@uphs.upenn.edu
PURPOSE

To compare detectability of lesions between digital mammography (DM) and digital breast tomosynthesis (DBT) in virtual clinical
trials (VCTs) using multi-reader multi-case (MRMC) analysis.
METHOD AND MATERIALS

VCTs were conducted using an open-source VCT pipeline (OpenVCT) with observer modeling software (MeVIC), to assess the
detection of masses and microcalcifications in DM and DBT. Twelve breast phantoms were used (simulating 700ml breasts with
compressed thickness of 6.33cm) with variations in simulated anatomy. Breast masses were simulated by oblate ellipsoids with
various diameter and thickness. Single calcifications were simulated as polycubes with various numbers of 100 µm cubes. A total of
1920 lesions were simulated. A clinical imaging geometry (Selenia Dimensions, Hologic) was simulated; exposure settings were
selected to match AEC performance. Mammographic phantom projections were processed and tomosynthesis images reconstructed
using commercially available software (Briona, Real-Time Tomography). Lesion detection was modeled by channelized Hotelling
observers with 15 Laguerre-Gauss channels and a spread of 22, using independent training and testing sets (each set consisting of
160 ROIs). The detectability of simulated lesions was assessed by the area under the ROC (AUC) using a one-shot MRMC approach,
assuming five independently trained observers. Two-sided Students' t-tests were used to test the statistical significance between
estimated AUCs.
RESULTS
For detection of simulated calcifications, we observed AUC of 0.802±0.023 and 0.799±0.026, for DM and DBT, respectively; these
AUCs do not differ significantly (p=0.856; 95% CI -0.040, 0.034). For the detection of masses, we observed AUC of 0.794±0.022
and 0.900±0.017, for DM and DBT, respectively, which is statistically significant (p<0.001; 95% CI 0.089, 0.124). Our observations
show close agreement with clinical results by Rafferty et al., who obtained AUC difference of 0.025 and 0.096, for calcifications and
masses respectively.
CONCLUSION

VCTs have demonstrated significant improvement in the detection of masses between DM and DBT, while AUCs for simulated
microcalcifications did not differ, in close agreement with published clinical results.
CLINICAL RELEVANCE/APPLICATION
We have demonstrated that virtual clinical trials can predict the clinical performance improvements seen with DBT over DM in the
detection of breast masses and microcalcifications.
BR205-ED- Interval Breast Cancers in the Era of Precision Medicine: A Multimodality Approach
WEB8
Station #8
Participants
Sarah G. Mizuguchi, MD, Louisville, KY (Abstract Co-Author) Nothing to Disclose
Lane M. Roland, MD, Louisville, KY (Abstract Co-Author) Nothing to Disclose
Stacey M. Crawford, MD, Louisville, KY (Abstract Co-Author) Nothing to Disclose
Elizabeth Riley, MD, Louisville, KY (Abstract Co-Author) Nothing to Disclose
Mary Ann Sanders, MD,PhD, Louisville, KY (Abstract Co-Author) Nothing to Disclose
Jason D. Messinger, MD, ST LOUIS, MO (Presenter) Nothing to Disclose
TEACHING POINTS

1. Review the definition of interval breast cancers and how they present. 2. Display imaging findings of interval cancers. 3. Discuss
how interval breast cancers typically have a worse prognosis. 4. Review receptor sensitivity, with histopathologic correlation, to
correlate with prognosis. 5. Review the current trend of precision medicine with medical oncology perspective. 6. Highlight the role
of the radiologist in precision medicine for breast cancer.
TABLE OF CONTENTS/OUTLINE

Interval breast cancer subtypes - true negative, false negative, occult, minimal sign - prognosis of interval cancers vs screen
detected - imaging findings Receptor sensitivity and tumor marker - review ER, PR, HER2 - Luminal A, Luminal B, HER2 enriched, and
Normal like - Ki-67 - histopathologic correlation Review of precision medicine - history - how receptor sensitivity influences
treatment - treatment strategies and outcomes from medical oncology persepctive Case examples to include imaging,
histopathology, and treatment with multimodality approach Summary
BR206-ED- A Wolf in Sheep's Clothing: Cancer Masquerading as a Benign-appearing Breast Mass
WEB9
Station #9
Awards
Certificate of Merit
Participants
Lindsey Storer, MD, Los Angeles, CA (Presenter) Nothing to Disclose
Cheryce P. Fischer, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
Stephanie A. Lee-Felker, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Melissa M. Joines, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Although most circumscribed breast masses are benign, up to 10-20% of breast malignancies can be relatively circumscribed.
Invasive ductal carcinoma, not otherwise specified, is the most common form of breast cancer and most often presents as an
irregular mass with spiculated margins; however, less common types can present as round or oval masses with predominately
circumscribed margins. After reviewing this presentation, participants will: Expand their differential diagnosis of circumscribed breast
masses to include less common types of breast cancer, including mucinous carcinoma, medullary carcinoma, papillary carcinoma,
malignant phyllodes tumors, and high grade triple negative intraductal carcinoma, as well as lymphoma and metastases; describe
the common and distinct imaging characteristics of these tumors; and integrate radiologic assessment with clinical scenario to
provide appropriate BIRADS category and management recommendation.
TABLE OF CONTENTS/OUTLINE

Mutimodality imaging review of the types of breast cancer that more often present as circumscribed masses on 2D and 3D
mammography, ultrasound, and MRI with clues that may help favor a specific diagnosis among these entities Epidemiology of these
breast cancer subtypes Clinical clues that may aid in diagnosis and review of prognosis Correlation with histologic features
BR207-ED- Coming Out: Making Your Breast Imaging Center LGBT Friendly
WEB10
Station #10
Awards
Certificate of Merit

Participants
Valerie J. Fein-Zachary, MD, Boston, MA (Presenter) Nothing to Disclose
Jordana Phillips, MD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Consultant, General Electric
Company
Hannah Perry, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Nancy Littlehale, Boston, MA (Abstract Co-Author) Nothing to Disclose
Tejas S. Mehta, MD, MPH, Boston, MA (Abstract Co-Author) Nothing to Disclose
Vandana M. Dialani, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Michael D. Fishman, MD, Boston, MA (Abstract Co-Author) Consultant, Zebra Medical Vision Ltd
Evguenia J. Karimova, MD, Memphis, TN (Abstract Co-Author) Nothing to Disclose
Priscilla J. Slanetz, MD, MPH, Belmont, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
vfeinzac@bidmc.harvard.edu
TEACHING POINTS
1. Diversity and inclusion in healthcare includes respect for and understanding of lesbian, gay, bisexual and transgender (LGBT)
people. 2. Improving access to care includes creating a welcoming environment. 3. Improving healthcare outcomes can be achieved
as providers increase their knowledge about issues specific to LGBT people and support patients by empowering them. 4. Including
lesbian, gay, bisexual and transgender women, and transgender men in breast cancer outcomes research is critical to fill in gaps in
medical knowledge.
TABLE OF CONTENTS/OUTLINE

1. Diversity and Inclusion: Understanding Healthcare Disparities in the LGBT population a. Historic discrimination by the medical
community b. Factors contributing to health disparities c. Current disparity issues 2. Improving access to care a. Create a
welcoming space, with verbal and visual clues of inclusion b. Provide staff education and training about LGBT issues c. Publicize
your efforts 3. Improve health outcomes for your LGBT patients a. Understand the unique needs of lesbians, gay women and
bisexual patients b. Know appropriate screening examinations and intervals for transgender patients c. Improve patient compliance
with recommendations d. Expand knowledge of the LGBT community through increased inclusion in research
BR208-ED- Anything but Binary: Exploring the Spectrum of Imaging Concerns in the Transgender Patient
WEB11
Station #11
Participants
Christopher McAdams, MD, Atlanta, GA (Presenter) Nothing to Disclose
Jean M. Kunjummen, DO, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
Margaret Fleming, MD, Atlanta, GA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1. Discuss approrpiate terminology and the review the gender spectrum concept 2. Identifiy pre-imaging considerations for
transgender patients with suggestions for application to your practice 3. Highlight unique pathologies and teaching points through
case examples
TABLE OF CONTENTS/OUTLINE

I. Introduction II. Patient-Centered Approach & Terminology (Gender Spectrum, etc.) III. Barrier to Care (Discrimination, Provider
Training, Insurance Coverage, Clinic Environment) IV. Current Screening Recommendations V. Case Vignettes (e.g. - Abscess,
Ectatic Ducts, Kaposi's Sarcoma, etc.) VI. Conclusions & Future Needs
BR209-ED- A Comprehensive Breast Cancer Risk Management Program Developed in a Radiology Department to
WEB12 Share with Referring Physicians
Station #12
Participants
Richard G. Barr, MD, PhD, Campbell, OH (Presenter) Consultant, Siemens AG; Consultant, Koninklijke Philips NV; Research Grant,
Siemens AG; Research Grant, SuperSonic Imagine; Speakers Bureau, Koninklijke Philips NV; Research Grant, Bracco Group; Speakers
Bureau, Siemens AG; Consultant, Canon Medical Systems Corporation; Research Grant, Esaote SpA; Research Grant, BK Ultrasound;
Research Grant, Hitachi, Ltd
TEACHING POINTS

The purpose of this exhibit is: 1. To review the different risk methods and describe their pros and cons 2. To highlight the
importance of the involvement of refering physicians for a better patient's empowerment
TABLE OF CONTENTS/OUTLINE
- Different breast cancer risk scores Breast density and breast cancer risk management The unsuitability of lifetime risk - Breast
cancer risk management program development - Evaluation of the referring physicians point of view
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Richard G. Barr, MD, PhD - 2017 Honored Educator
 MSRT44

ASRT@RSNA 2018: Hindsight is 20/20 - The Earliest Appearance of Breast Cancer

W ednesday, Nov. 28 1:00PM - 2:00PM Room: N230B

BR OI
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Rita Gidwaney, MD, San Rafael, CA (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) To review some of the key imaging features on mammography, tomosynthesis, ultrasound and MRI of detecting early breast
cancer. 2) To review common pitfalls in breast imaging and how to avoid them in the future. 3) To be better equipped at handling
such hindsight wisdom cases in your own practice.
ABSTRACT
This session will review how our advanced technology can help us detect the earliest signs of breast cancer, using mammography,
tomosynthesis, ultrasound and MRI, but also how positioning and technique can play a huge role in cancer detection. We will
examine cases and look at the pattern of how pathology develops in the breast, both malignant and benign. Common image
aquisition and interpretation pitfalls will be reviewed as well. The session will also discuss ways of evaluating such cases in our
practices to achieve the best patient outcomes, in spite of hindsight wisdom.
 VW 46

Wide-angle Digital Breast Tomosynthesis Reading Session: Siemens Healthineers Vendor Workshop
W ednesday, Nov. 28 1:05PM - 2:15PM Room: Booth 5530

Participants
Paula M. Grabler, MD, Oak Park, IL (Presenter) Nothing to Disclose
Brandie L. Fagin, MD, Glenview, IL (Presenter) Nothing to Disclose
Program Information
During this hands-on workshop you will learn to evaluate 2D mammography and 3D Breast Tomosynthesis. An expert tutor will lead
you through cases that will both fascinate and challenge you! All cases have been acquired with Siemens Mammomat Inspiration
and are displayed on our syngo® Breast Care workstations so you will become familiar with the quality of our HD Tomo images and
ease of use of our systems.
 VW 15

3D ABUS: Hands-on Workshop: GE Vendor Workshop

W ednesday, Nov. 28 1:30PM - 2:30PM Room: Booth 8156

Participants
Susan G. Roux, MD, Carmel Valley, CA (Presenter) Nothing to Disclose
Program Information
Peer educator led workshop includes a live Invenia ABUS 2.0 scan station acquisition and hands-on review of clinical cases using
the Invenia ABUS Viewer. Learn about the importance of the coronal view and how 3D ABUS screening helps increase cancer
detection in women with dense breast tissue. Registration is required; adding this session to the RSNA calendar tool alone does not
secure your seat in this session. Click the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 69

Tomosynthesis Guided Prone Breast Biopsy Solutions in a Community Based Practice: Hologic Vendor
Workshop
W ednesday, Nov. 28 1:30PM - 2:30PM Room: Booth 5524

Participants
Harriet B. Borofsky, MD, San Mateo, CA (Presenter)
Program Information
Clinical benefits of tomosynthesis guided biopsy which includes a hands-on demonstration of the Affirm® Prone Biopsy System and
the Brevera® Breast Biopsy System (Affirm® Prone Biopsy System, Brevera® Breast Biopsy System and Viera™ Portable Breast
Ultrasound)
Registration
https://hologicrsna.com
 VW 47

A Practical Approach to Breast Magnetic Resonance Imaging (MRI) Interpretation: An Interactive Session:
Siemens Healthineers Vendor Workshop
W ednesday, Nov. 28 2:30PM - 3:40PM Room: Booth 5530

Participants
Susan Weinstein, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Program Information
This interactive session will include both didactic and hands-on case review at workstations equipped with syngo® MR Brevis. A
practical approach to breast MRI interpretation will be discussed as well as utilizing the available sequences and techniques to
improve interpretive skills.
 SSM01

Breast Imaging (Interventional Techniques and Path Correlation)

W ednesday, Nov. 28 3:00PM - 4:00PM Room: E351

BR
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Michael A. Cohen, MD, Atlanta, GA (Moderator) Nothing to Disclose
Gary J. Whitman, MD, Houston, TX (Moderator) Nothing to Disclose
Sub-Events
SSM01-01 Cryoablation as a Primary Treatment of Low-Risk Breast Cancers: An Interim Update of the Ice 3
Trial
W ednesday, Nov. 28 3:00PM - 3:10PM Room: E351

Participants
Kenneth R. Tomkovich, MD, Freehold, NJ (Presenter) Consultant, Scion Medical Technologies, LLC;
For information about this presentation, contact:
ktomkovich@princetonradiology.com
PURPOSE
The Ice 3 Trial is the first of its kind large scale multi center trial in the world to assess image guided cryoablation as a primary
treatment for breast cancer without surgical lumpectomy. We report updated interim results and important imaging findings.
METHOD AND MATERIALS
This HIPPA compliant and IRB approved trial seeks enrollment of between 150 and 200 patients for cryoablation of low risk
carcinoma of the breast. The study is limited to female patients ages 60 and over with biopsy proven primary, unifocal cancer.
Cancers must measure 1.5cm or less with tumor prognostic panels that are ER+/PR+ or ER+/PR-and HER 2-. All patients underwent
ultrasound guided cryoablation using the IceSense 3 system (IceCure Medical). Following local anesthesia, patients underwent a
freeze, thaw, freeze cycle of cryoablation with the goal of a visible ice ball producing at least a 10mm margin of ice around the
tumor. Patients will be followed for recurrence with mammography at 6 and 12 months and then annually for 5 years. Additional
imaging with MRI or Ultrasound may be utilized as needed but is not a requirement. All patients have the option of post cryoablation
chemotherapy, hormone therapy and or radiation therapy as clinically indicated. Patients will not undergo surgical lumpectomy
following cryoablation.
RESULTS
A total of 157 patients have been treated with since enrollment began in October 2014 at 17 participating centers across the
United States. Patients ranged from 60-90 years of age. Tumor sizes ranged from 3 to 15mm. No serious adverse events were
reported. There has been 100% procedural success. All patients have had at least 6 months follow up. 78 patients have had at
least 24 months follow up. 24 patients have had at least 36 months follow up. There has been no recurrence in 156/157 patients
with at least 6 months follow up (99.4% success rate). Common imaging findings include fat necrosis, scarring and a mammographic
"halo" effect.
CONCLUSION
Cryoablation of the breast is safe and well tolerated with a 100% initial procedural success rate. The overall clinical success rate
for 157 patients with at least 6 months follow up is 99.4%. Long term results are also promising.
CLINICAL RELEVANCE/APPLICATION
Interim results suggest that cryoablation is a safe and effective primary treatment for women with small low risk breast cancers as
an alternative to surgical lumpectomy.
SSM01-02 Do Eligibility Criteria for Ductal Carcinoma in Situ (DCIS) Active Surveillance Trials Identify Patients at
Low Risk for Upgrade to Invasive Carcinoma?
W ednesday, Nov. 28 3:10PM - 3:20PM Room: E351

Participants
Manisha Bahl, MD,MPH, Boston, MA (Presenter) Nothing to Disclose
Tawakalitu Oseni, Boston, MA (Abstract Co-Author) Nothing to Disclose
Charmi Vijapura, MD, Iowa City, IA (Abstract Co-Author) Nothing to Disclose
Niveditha Pinnamaneni, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Constance D. Lehman, MD,PhD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Medical Advisory
Board, General Electric Company
For information about this presentation, contact:
mbahl1@mgh.harvard.edu
PURPOSE
Due to concerns regarding overtreatment of ductal carcinoma in situ (DCIS), randomized controlled trials are currently underway in
the United States and Europe to determine the safety and efficacy of active surveillance versus usual care for women with DCIS.
The purpose of this study is to determine upgrade rates of DCIS at core needle biopsy to invasive carcinoma at surgical excision
among women who meet eligibility criteria for three active surveillance trials.
METHOD AND MATERIALS
In this IRB-approved and HIPAA-compliant study, a retrospective review was performed of consecutive female patients diagnosed
with DCIS at vacuum-assisted core needle biopsy from 2007 to 2016. Medical records were reviewed for mode of presentation,
imaging findings, core biopsy pathology results, and surgical outcomes. DCIS cases were classified based on eligibility criteria for
the COMET, LORD, and LORIS Trials. Of note, eligibility for the LORIS Trial also requires real-time central pathology review with
features not routinely reported and thus not captured in this retrospective study.
RESULTS
Over a ten-year period, 1378 patients were diagnosed with unilateral DCIS at core biopsy and 12 were diagnosed with bilateral
DCIS, for a total of 1390 patients (mean age 57, range 27-89) with 1402 cases of DCIS. 79.8% (n=1119) were detected on
screening mammography. 17.3% of cases (n=243) were low nuclear grade, 47.2% (n=662) were intermediate nuclear grade, 26.2%
(n=368) were high nuclear grade, and 9.2% (n=129) were unreported. The overall upgrade rate to invasive disease was 19.1%
(268/1402). 485 were eligible for the COMET Trial, 163 for the LORD Trial, and 489 for the LORIS Trial. The rates of upgrade to
invasive carcinoma were 12.2% (59/485), 7.4% (12/163), and 12.5% (61/489) for the COMET, LORD, and LORIS Trials,
respectively. Of the 69 cases that upgraded to invasive carcinoma, 31.9% (n=22) upgraded to microinvasive disease (less than or
equal to 1 mm) and 91.3% (n=63) were node-negative.
CONCLUSION

The upgrade rates for women with DCIS who meet eligibility criteria for active surveillance trials range from 7.4 to 12.5%. Of the
upgraded cases, nearly one-third were microinvasive disease and more than 90% were node-negative.
CLINICAL RELEVANCE/APPLICATION

The risk of missing occult invasive carcinoma in women eligible for active surveillance trials ranges from 7.4 to 12.5%, and the
majority of these cancers have favorable biologic profiles.
SSM01-03 Role of Vacuum Assisted Excision (VAE) in Managing Ductal Atypia Such as Flat Epithelial Atypia
(FEA) and Atypical Intraductal Epithelial Proliferation (AIDP)
W ednesday, Nov. 28 3:20PM - 3:30PM Room: E351

Participants
Nisha Sharma, MBChB, Leeds, United Kingdom (Presenter) Nothing to Disclose
Isobel Haigh, Leeds, United Kingdom (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
nisha.sharma2@nhs.net
PURPOSE

It is recognised that surgical excision for B3 lesions is considered over treatment and vacuum excision (VAE) is playing a role in
managing these lesions.. In Europe and USA ductal atypia is still treated with surgical excision. In the UK ductal atypia is managed
with vacuum excision rather than surgical excision.
METHOD AND MATERIALS

Retrospective audit performed from a prospective database of all B3 lesions identified through the Breast Screening Programme. All
cases of pure ductal atypia, including AIDP and FEA were identified from April 2009 to March 2016. The mode of biopsy and upgrade
rate were recorded and follow up data obtained until March 2018.
RESULTS
268990 women were screened from April 2009 to March 2016, of which 12434 were recalled to assessment (4.6%). 5582 biopsies
were performed of which 688 were B3 lesions (12.3%). Ductal atypias (FEA and AIDP) (excluding papilloma and radial scars with
ductal atypias) accounted for 39.8% of the biopsies. 69% (190/274) were managed with vacuum assisted excsion (VAE) and annual
mammographic follow-up or routine screening surveillance. 3% (7/190) developed a cancer during surveillance period, of which 4
were in the same quadrant. 13% (35/274) were upgraded to malignancy following VAE and were treated with therapeutic surgery. 2
developed further cancer on surveillance in the same breast. 8% (21/274) had a vacuum excision and a surgical biopsy due to
radiological or pathological concern and 14/21 was benign and 7/21 upgraded to malignancy. One case developed cancer in the
contralateral breast on cancer follow up. 8% (22/274) had a surgical diagnostic biopsy instead of vacuum excision and 13/22 were
benign and 9/22 were upgraded to malignancy. 2% (6/274) did not go on to have either vacuum excision or surgery due to co-
morbidities. 2 developed cancer on surveillance. 12/274 (4%) developed malignancy during surveillance period of which 8/274 were
in the same breast.
CONCLUSION

Our study shows that managing ductal atypia with vacuum assisted excision (VAE) is a safe alternative to surgical excision as a
primary intervention but multidisciplinary review is important to determine if further surgery is required. Vacuum excision allowed
13% of our women to have a therapeutic surgery as preoperative diagnosis of malignancy was made and 69% avoided surgery
altogether.
CLINICAL RELEVANCE/APPLICATION
Vacuum assisted excision is a safe alternative to surgical biopsy in managing ductal atypias
SSM01-04 Papillary Breast Lesions Without Atypia Diagnosed by Core Biopsy: Should They Be Surgically
Excised?
W ednesday, Nov. 28 3:30PM - 3:40PM Room: E351

Participants
Sarah Dhundass, MD, Saint Cloud, France (Presenter) Nothing to Disclose
Pascal Cherel, Saint Cloud, France (Abstract Co-Author) Nothing to Disclose
Adriana Langer, MD, Saint Cloud, France (Abstract Co-Author) Nothing to Disclose
PURPOSE

To evaluate the upgrade rate of benign papilloma without atypia diagnosed after core needle biopsy (CNB) or Vacuum assisted
biopsy (VAB) in order to determine whether they should require surgical excision or not.
METHOD AND MATERIALS

Histological results of 15615 biopsy procedures were reviewed from January 2001 to December 2014 in our institution. Sampling
methods included large gauge VAB by ultrasonographic (US) or stereotaxic guidance and US guided 14G CNB. A total of 179
papillary lesions without atypia that were diagnosed in 159 women were included: 121 had surgical excision following initial
detection and 58 underwent imaging surveillance. Initial histological results were compared to the histological results after surgery
or to the follow up findings. Statistical analysis was performed to evaluate the association of clinical and radiological variables with
the upgrade rate to malignancy.
RESULTS

After exclusion of patients lost to follow up, 158 lesions have been analyzed. 90 were diagnosed by VAB and 68 by CNB. Surgical
excision revealed 7 malignant lesions (6 DCIS and 1 carcinoma). The upgrade rate to malignancy was 4.4%. Among the 42 followed
up patients, two necessitated secondary surgery, one of them corresponded to a malignant lesion (45 months after biopsy). The
median follow-up period was 30 months (6-93). The upgrade rate was statistically higher in the group diagnosed with CNB 8.9% vs
1.1% in the VAB group (p=0.042). Age, history of cancer, size of the lesion, ultrasound and mammographic findings including
BIRADS category were not associated with underestimation of malignancy (p>0.05).
CONCLUSION
Papilloma without atypia detected with VAB could not require surgery (upgrade rate 1.1%) if the excision is complete but should be
radiologically followed up at least 5 years to screen for potential biopsy site changes. Papilloma without atypia diagnosed with CNB
require surgical excision as the risk of coexisting carcinoma is significantly higher (8.9% in our study)
CLINICAL RELEVANCE/APPLICATION

Papilloma without atypia diagnosed with vacuum assisted biopsy can be radiologically followed up given the low rate of
underestimation.
SSM01-05 Follow-Up Outcomes of BI-RADS Category 3 Solid Nodules Identified on Screening and Diagnostic
Breast Ultrasound
W ednesday, Nov. 28 3:40PM - 3:50PM Room: E351

Awards
Student Travel Stipend Award

Participants
Joanna Marie D. Choa, MD, Taguig, Philippines (Presenter) Nothing to Disclose
Anna Lyn C. Egwolf, MD, Taguig, Philippines (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

choa.jmd@gmail.com
PURPOSE

Our study aims to determine significant difference in BI-RADS category from initial imaging to its subsequent follow-up and the
minimum length of time that the BI-RADS category will change for solid nodules. Our study also wants to assess the characteristics
and growth pattern of BI-RADS category 3 solid nodules.
METHOD AND MATERIALS

We did a retrospective cohort study reviewing patient reports across a 10-year follow-up period. All solid nodules with at least 2
follow-up ultrasound imaging within a minimum of 22 months were included. Patients were segregated into 2 groups by age: one
group below 40 years old and another group age 40 years old and above. Mammogram results were not viewed. Patients with
known carcinoma and those with other lesions having a higher BI-RADS assessment in either breast were excluded. Nodule size,
characteristics, and BI-RADS score for each follow-up were obtained. We employed ANOVA for repeated measures for statistical
analysis.
RESULTS

A total of 511 nodules (5.1% of identified solid nodules) were included and showed no significant difference in BI-RADS scores over
the course of 2 follow-ups, regardless of age. Significant change was identified at the sixth and seventh follow-ups with
downgrading of BI-RADS scores to 2. Significant changes in size between the second and third follow-up and comparing the first
with the third follow-up were appreciated in patients 40 years old and above. Patients in this study were also found out to have an
average interval between follow-up imaging of 14.39 months. The minimum length of time for a BI-RADS category 3 solid nodule to
become category 4 on follow-up was approximately 1 year (354 days). Only 1 out of the total 511 nodules turned out malignant, 44
months after initial imaging.
CONCLUSION

There is no significant change between BI-RADS scores of solid nodules regardless of age within a 2-year follow-up period. The
minimum time for change in BI-RADS from 3 to 4 is about 1 year. Low patient compliance to BI-RADS category 3 follow-up
guidelines should be taken into consideration in formulating institutional protocols in the management of these solid nodules.
CLINICAL RELEVANCE/APPLICATION
Patient compliance to follow-up recommendations of BI-RADS category 3 solid nodules is very low, hence can be re-classified as
category 2 and follow an annual schedule of follow-up imaging. Only 1 out of 511 nodules turned out malignant after 44 months
(~0.2%, low turn-out rate).
SSM01-06 Comparison of Upright Digital Breast Tomosynthesis-Guided Vacuum-Assisted Biopsy with
Conventional Prone Stereotactic Vacuum-Assisted Biopsy
W ednesday, Nov. 28 3:50PM - 4:00PM Room: E351

Participants
Mary Maunglay, MD, Pontiac, MI (Abstract Co-Author) Nothing to Disclose
Manisha Bahl, MD,MPH, Boston, MA (Presenter) Nothing to Disclose
Helen Anne D'Alessandro, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Constance D. Lehman, MD,PhD, Boston, MA (Abstract Co-Author) Research Grant, General Electric Company; Medical Advisory
Board, General Electric Company
For information about this presentation, contact:
mbahl1@mgh.harvard.edu
PURPOSE

Breast imaging practices are rapidly transitioning from digital 2D mammography (DM) to digital breast tomosynthesis (DBT), but
there is limited research on the clinical performance of DBT-guided vacuum-assisted breast biopsy (VABB). The purpose of this
study was to compare the performance of upright DBT VABB to that of conventional prone stereotactic (PS) VABB.
METHOD AND MATERIALS

In this IRB-approved and HIPAA-compliant study, a retrospective review was performed of consecutive patients who underwent PS
VABB from August 2014 to December 2015 (PS VABB group) and DBT VABB from February 2016 to June 2017 (DBT VABB group).
Tissue sampling methods and materials (9-gauge needles) were the same for PS VABB and DBT VABB. Student's t-tests and chi-
square tests were used to compare the following variables between the PS VABB and DBT VABB groups: sampling success,
procedure times, exposures, lesion types, histologic results, and complications.
RESULTS
Over a 17-month period before the introduction of DBT VABB, 444 PS VABBs in 410 patients (mean age 57, range 32-84) were
performed (PS VABB group). Over a 17-month period after complete integration of DBT VABB, 709 DBT VABBs in 682 patients (mean
age 58, range 23-90) were performed (DBT VABB group). Technical success was achieved for more lesions with DBT VABB than PS
VABB (98.4% [698/709] vs 93.2% [414/444], p<0.001). Mean procedure time was shorter with DBT VABB (12 vs 28 minutes,
p<0.001), and significantly fewer exposures were acquired with DBT VABB (3 vs 11, p<0.001). A higher percentage of lesions
biopsied with DBT VABB were non-calcified lesions (eg, architectural distortion, asymmetry, and mass) (29.5% [206/698] with DBT
VABB vs 3.1% [13/414] with PS VABB, p<0.001). There were no differences in the distribution of histologic results (benign, high-
risk, or malignant) (p=0.94). No major complications were observed in either group. Two patients in the DBT VABB group (2/682,
0.3%) experienced self-limited vasovagal symptoms.
CONCLUSION

DBT VABB has higher technical success than PS VABB and can be performed in less than half the time and with less than one third
of the radiation. In addition, more distortions and asymmetries are amenable to biopsy with DBT VABB.
CLINICAL RELEVANCE/APPLICATION

Clinical performance of DBT VABB is superior to PS VABB. DBT VABB can replace PS VABB for routine use in patients with suspicious
findings identified on DBT alone and on conventional DM.
 SSM02

Science Session with Keynote: Breast Imaging (Risk-Based Screening: Should We Do It?)
W ednesday, Nov. 28 3:00PM - 4:00PM Room: E350

AI BR SQ
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Elizabeth A. Morris, MD, New York, NY (Moderator) Nothing to Disclose
Daniel B. Kopans, MD, Waban, MA (Moderator) Royalties, Cook Group Incorporated; Research Consultant, Deep Health; Scientific
Advisory Board, Dart, Inc
Sub-Events
SSM02-01 Breast Keynote Speaker: Risk Based Screening
W ednesday, Nov. 28 3:00PM - 3:10PM Room: E350

Participants
Elizabeth A. Morris, MD, New York, NY (Presenter) Nothing to Disclose
SSM02-02 Risk-Based Screening Mammography for Women Age <40: Outcomes from the National
Mammography Database
W ednesday, Nov. 28 3:10PM - 3:20PM Room: E350

Participants
Cindy S. Lee, MD, Garden City, NY (Presenter) Nothing to Disclose
Heidi Ashih, PhD, Reston, VA (Abstract Co-Author) Nothing to Disclose
Debapriya Sengupta, MBBS,MPH, Reston, VA (Abstract Co-Author) Nothing to Disclose
Edward A. Sickles, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Margarita L. Zuley, MD, Pittsburgh, PA (Abstract Co-Author) Investigator, Hologic, Inc
Etta D. Pisano, MD, Charleston, SC (Abstract Co-Author) Researcher, Freenome Holdings Inc; Researcher, Real Imaging Ltd;
Researcher, Therapixel; Researcher, DeepHealth, Inc; Researcher, ToDos
For information about this presentation, contact:
Cindy.Lee3@nyumc.org
PURPOSE
There is insufficient large-scale evidence supporting screening mammography in women <40 years with risk factors. This study
compares risk-based screening of women ages 30-39 versus women age 40-49 with no known risk factors, using screening
mammography performance metrics from the National Mammography Database (NMD).
METHOD AND MATERIALS
This HIPAA compliant and IRB approved study analyzed data from 150 mammography facilities in 31 states in the NMD. The NMD
collects clinical practice data including self-reported patient demographics, clinical findings, screening mammography interpretation
and biopsy results. Patients were stratified by 5-year age intervals and specific risk factors for breast cancer: family history of
breast cancer (any first degree relative regardless of age), personal history of breast cancer and breast density of heterogeneously
or extremely dense (C or D). Prior mammograms were identified by patient date of birth and facility-assigned identification number.
Four performance metrics for screening mammography were calculated for each age and risk group: recall rate, cancer detection
rate, and positive predictive values for biopsy recommended (PPV2) and biopsy performed (PPV3).
RESULTS
5,772,730 screening mammograms were performed between January 2008 and December 2015 in 2,647,315 women. Overall, mean
cancer detection rate was 3.7 per 1000 (95% CI: 3.65-3.75), recall rate was 9.8% (9.8-9.8%), PPV2 was 20.1% (19.9-20.4%),
and PPV3 was 28.2% (27.0-28.5%). Overall, women age 30-34 and 35-39 had similar cancer detection rates, recall rates and PPVs,
with the presence of the three evaluated risk factors associated with significantly higher cancer detection rates. Moreover,
compared to a population currently recommended for screening mammography in the USA (age 40-44 with no known risk factors),
incidence screening (at least one prior screening examination) of women ages 30-39 with tthe three evauated risk factors has
similar cancer detection rates and recall rates.
CONCLUSION
Women ages 30-39 with 3 specific risk factors should benefit by starting screening at age 30 instead of the age 40 start
recommended for average-risk women.
CLINICAL RELEVANCE/APPLICATION
Women
SSM02-03 A Deep-Learning Breast Cancer Risk Prediction Network: Trained on the Population-based Swedish
CSAW Data
W ednesday, Nov. 28 3:20PM - 3:30PM Room: E350

Participants
Fredrik Strand, MD,PhD, Stockholm, Sweden (Presenter) Nothing to Disclose
Yue Liu, Stockholm, Sweden (Abstract Co-Author) Nothing to Disclose
Kevin Smith, Stockholm, Sweden (Abstract Co-Author) Nothing to Disclose
Hossein Azizpour, Stockholm, Sweden (Abstract Co-Author) Nothing to Disclose
Karin H. Dembrower, MD, Stockholm, Sweden (Abstract Co-Author) Nothing to Disclose
Peter Lindholm, MD, PhD, Stockholm, Sweden (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
fredrik.strand@ki.se
PURPOSE
Almost half of breast cancer diagnoses among women attending mammographic screening are interval cancers or large screen-
detected cancers. To enable more effective individualized screening, accurate risk prediction is paramount. In this study, we
examine how our trained deep learning network compares with mammographic density in risk prediction based on negative screening
mammograms.
METHOD AND MATERIALS
The Swedish cohort of screen-age women (CSAW) contains over 500,000 women linked to the cancer registry and to an image
database. Our deep learning network was trained on negative mammograms from incident cases from one uptake area 2008 to
2011. The test set consisted of cases from 2013 and 2014. In each set, we included a random sample of concurrent non-
overlapping controls. The input was each negative mammogram downscaled as well as full-resolution central crops, age at
mammography and selected DICOM parameters. The prediction output is called deep learning risk score (DLR). For comparison,
mammographic density was calculated using the validated LIBRA software. Logistic regression models were fitted to examine odds
ratios.
RESULTS

The training set consisted of 3167 negative mammograms from women with subsequent breast cancer and 125,683 mammograms
from healthy women. The test set consisted of negative mammograms from 752 screening rounds of 326 women with subsequent
breast cancer and 6728 rounds of 2065 healthy women. AUC was higher for DLR (0.63; 95%CI: 0.61 to 0.66) than for density
(0.57; 95%CI: 0.54 to 0.60) and for age-adjusted density (0.58; 95%CI: 0.56 to 0.61). The proportion of cases were 10.1% in the
top quintile and 2.5% in the bottom quintile of DLR. The top-to-bottom quintile odds ratio was 4.37 (95%CI: 3.01 to 6.45) and 1.69
(95%CI: 1.23 to 2.32) for DLR and age-adjusted density respectively.
CONCLUSION

We have demonstrated that it is possible to train a deep learning network on negative screening mammograms from subsequent
breast cancer cases, and produce risk predictions with reasonable accuracy and ability to identify women at elevated risk.
CLINICAL RELEVANCE/APPLICATION

After external validation, our network may be used in individualizing breast cancer screening.
SSM02-04 Potential Role of Convolutional Neural Network based Algorithms in Patient Selection for DCIS
Observation Trials
W ednesday, Nov. 28 3:30PM - 3:40PM Room: E350

Participants
Simukayi Mutasa, MD, New York, NY (Presenter) Nothing to Disclose
Peter Chang, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Jenika Karcich, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Eduardo Pascual Van Sant, BS, New York, NY (Abstract Co-Author) Nothing to Disclose
Mary Q. Sun, MD, Manhasset, NY (Abstract Co-Author) Nothing to Disclose
Michael Z. Liu, MS, New York, NY (Abstract Co-Author) Nothing to Disclose
Sachin Jambawalikar, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Richard S. Ha, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
s.mutasa@columbia.edu
PURPOSE
Minimizing over-diagnoses and treatment of Ductal Carcinoma in Situ (DCIS) has led to clinical trials of observing patients with DCIS
instead of surgery. Despite careful selection for 'low risk' DCIS patients, there is evidence of occult invasive cancers in a significant
number of these patients. We investigated the feasibility of utilizing convolutional neural networks (CNN) for predicting patients
with pure DCIS versus DCIS with invasion using mammographic images.
METHOD AND MATERIALS

An IRB-approved retrospective study was performed. 246 unique images from 123 patients were used for our CNN algorithm. 164
images in 82 patients diagnosed with DCIS by stereotactic-guided biopsy of calcifications without any upgrade at the time of
surgical excision (pure DCIS group). 82 images in 41 patients with mammographic calcifications yielding occult invasive carcinoma as
the final upgraded diagnosis on surgery (occult invasive group). Two standard mammographic magnification views (CC and ML/LM)
of the calcifications were used for analysis. Calcifications were segmented using an open source software platform 3D Slicer and
resized to fit a 128x128 pixel bounding box. A 15 hidden layer topology based on residual convolutions was used to implement the
neural network. A class balanced holdout set with 40 patients was used for testing. 5-fold cross validation was utilized with cases
randomly separated into a training set [80%] and validation set [20%].
RESULTS
The CNN algorithm for predicting patients with pure DCIS achieved an overall validation accuracy of 74.6% (95%CI, ± 5) with area
under the ROC curve of 0.71 (95% CI, ±0.04), specificity of 49.4% (95% CI, ± 6%) and sensitivity of 91.6% (95% CI, ± 5%).
CONCLUSION

It's feasible to apply a CNN to distinguish pure DCIS from DCIS with invasion using mammographic images. A larger dataset will likely
improve our prediction model and could potentially be useful in appropriate patient selection for observation trials.
CLINICAL RELEVANCE/APPLICATION

Convolutional neural networks have demonstrated strong performance in various image classification tasks and may potentially be
used in appropriate patient selection for DCIS observation trials.
SSM02-05 The Effect of Screening Modality and Race on BI-RADS Breast Density in a Large Urban Screening
Cohort
W ednesday, Nov. 28 3:40PM - 3:50PM Room: E350

Participants
Aimilia Gastounioti, Philadelphia, PA (Presenter) Nothing to Disclose
Anne Marie McCarthy, Boston, MA (Abstract Co-Author) Nothing to Disclose
Lauren Pantalone, BS, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Marie Synnestvedt, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
For information about this presentation, contact:
aimilia.gastounioti@uphs.upenn.edu
PURPOSE
Increased breast density is an independent breast cancer risk factor and also limits the sensitivity and specificity of mammography.
We investigated the effect of screening mammography modality and race on BI-RADS breast density assessments, accounting for
age and body-mass index (BMI).
METHOD AND MATERIALS

We retrospectively analyzed data from 24,740 individual women (45% White, 55% Black) who underwent screening from September
2010 through February 2017 at our institution. 15,147 women (55%) had repeated screening studies (N = 60,774 studies). Over
this time period, three screening modalities were used: digital mammography alone (DM; N = 8,936); digital breast tomosynthesis
(DBT) with DM (DM/DBT, N = 30,786); and synthetic 2D with DBT (s2D/DBT, N = 21,052). BI-RADS density classifications ranging
from lower (fatty or scattered) to higher (heterogeneous or extremely dense) density were extracted from screening reports.
Random-effects ordered logistic regression (panel variable: individual woman) was performed to estimate the odds of being assigned
to higher BI-RADS density by each modality, adjusted for race, age, BMI and radiologist. The interaction of modality and race on
density was tested in the model, and analyses were stratified by race.
RESULTS

Women screened with DBT had significantly lower odds of high density compared to those screened with DM alone (DM/DBT vs. DM:
OR = 0.62, p < .0001; s2D/DBT vs. DM: OR = 0.48, p < .0001). Lower odds of high density were also observed in s2D/DBT
compared to DM/DBT (OR = 0.76, p < .0001). There was a significant interaction of modality and race on breast density (p =
.0003). All differences by modality maintained statistical significance in analyses stratified by race, with lower ORs for black
(DM/DBT vs. DM: OR = 0.61; s2D/DBT vs DM: OR = 0.40; s2D/DBT vs. DM/DBT: OR = 0.67) than for white women (DM/DBT vs. DM:
OR = 0.65; s2D/DBT vs. DM: OR = 0.58; s2D/DBT vs. DM/DBT: OR = 0.89).
CONCLUSION

Screening mammography modality has a significant effect on BI-RADS density assessment with an overall trend of assigning lower
density with DBT and s2D/DBT screening versus DM alone. Furthermore, this effect seems to be more prominent in black than in
white women.
CLINICAL RELEVANCE/APPLICATION

Our findings have direct implications for personalized screening since breast density assignments, which often drive
recommendations for supplemental screening, may vary greatly by modality and race.
SSM02-06 Breast Keynote Speaker: Risk Based Screening
W ednesday, Nov. 28 3:50PM - 4:00PM Room: E350

Participants
Daniel B. Kopans, MD, Waban, MA (Presenter) Royalties, Cook Group Incorporated; Research Consultant, Deep Health; Scientific
Advisory Board, Dart, Inc
 VW 16

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

W ednesday, Nov. 28 3:00PM - 3:30PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 70

True Patient Comfort: Genius™ 3D Mammography™ exam and SmartCurve™ Breast Stabilization System:
Hologic Vendor Workshop
W ednesday, Nov. 28 3:00PM - 4:00PM Room: Booth 5524

Program Information
A clinical perspective on the implementation of Genius™ 3D Mammography™ exam and the SmartCurve™ Breast Stabilization
System without compromising image quality, exam time, dose or workflow. This session includes clinical feedback and tips to
increase patient comfort. (Genius™ 3D Mammography™, SmartCurve™ Breast Stabilization System)
Registration
https://hologicrsna.com
 VW 17

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

W ednesday, Nov. 28 3:30PM - 4:00PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 48

Automated Breast Volume Scanner (ABVS) Physician Training Workshop: An Interactive Learning Experience:
Siemens Healthineers Vendor Workshop
W ednesday, Nov. 28 3:50PM - 5:00PM Room: Booth 5530

Participants
Ingolf Karst, MD, Chicago, IL (Presenter) Nothing to Disclose
Program Information
Under the guidance of a breast imaging expert you will develop your skills in the interpretation of 3D breast ultrasound acquired with
the ACUSON S2000™ Automated Breast Volume Scanner (ABVS), HELX Evolution with Touch Control and displayed on workstations
equipped with syngo® Ultrasound Breast Analysis (sUSBA) software. Active participation in real clinical cases will enable you to
become familiar with the unique coronal plane while providing practical approaches to interpretation of 3D automated breast
ultrasound.
 ED001-TH

Breast Thursday Case of the Day

Thursday, Nov. 29 7:00AM - 11:59PM Room: Case of Day, Learning Center

AMA PRA Category 1 Credit ™: .50

Participants
Jessica H. Porembka, MD, Dallas, TX (Presenter) Nothing to Disclose
Amy M. Fowler, MD, PhD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Susan O. Holley, MD, PhD, Raleigh, NC (Abstract Co-Author) Nothing to Disclose
Alexander B. Sevrukov, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Chandni Bhimani, DO, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Catherine A. Young, MD, JD, Saint Louis, MO (Abstract Co-Author) Research support, Hologic, Inc
Cheryl R. Herman, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Michelle V. Lee, MD, Saint Louis, MO (Abstract Co-Author) Nothing to Disclose
Mai A. Elezaby, MD, Madison, WI (Abstract Co-Author) Research Grant, Exact Sciences Corporation
Lonie R. Salkowski, MD,PhD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Roberta M. Strigel, MD, Madison, WI (Abstract Co-Author) Research support, General Electric Company
Ryan W. Woods, MD, MPH, Madison, WI (Abstract Co-Author) Nothing to Disclose
Urvi A. Tailor, MD, Madison, WI (Abstract Co-Author) Nothing to Disclose
Lindsay Compton, MD, Dallas, TX (Abstract Co-Author) Researcher, QT Ultrasound, LLC
Ramapriya Ganti, MD, Dallas, TX (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
1) Identify, characterize, and analyze abnormal findings on multimodality breast imaging studies. 2) Develop differential diagnostic
considerations based on the clinical information and imaging findings. 3) Recommend appropriate management for the patients
based on imaging findings.
 MSES51

Essentials of Breast Imaging

Thursday, Nov. 29 8:30AM - 10:00AM Room: S406B

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Sub-Events
MSES51A Missed Lesions in Mammography: How to Improve Performance
Participants
Athina Vourtsi, MD, Athens, Greece (Presenter) Consultant, General Electric Company; Educator, ABUS
LEARNING OBJECTIVES
1) Identify the most common factors that may lead to missed breast cancers. 2) Apply the appropriate steps when interpreting
mammography. 3) Enhance skills in order to avoid the possibility of missing a suspicious lesion.
MSES51B Update on Ductal Carcinoma in Situ
Participants
Cecilia L. Mercado, MD, New York, NY (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Describe the imaging characteristics of ductal carcinoma in situ as detected on various imaging modalities. 2) Identify the risk
factors for the development of ductal carcinoma in situ. 3) Discuss the various treatment recommendations for ductal carcinoma in
situ as supported by the recent multiple randomized controlled trials.
MSES51C Management of Non-simple Breast Cysts
Participants
A. Thomas Stavros, MD, San Antonio, TX (Presenter) Advisor, Devicor Medical Products, Inc Advisor, General Electric Company
Advisor, SonoCine, Inc Owner, Ikonopedia, LLC Medical Director, Seno Medical Instruments, Inc
For information about this presentation, contact:
atstavros@gmail.com
LEARNING OBJECTIVES

1) To understand the histologic bases for the appearances of cysts that are not simple. 2) To know the differences between
'suspicious' complex cystic and solid masses and non-suspicious benign complicated cysts. 3) To be aware of ancillary ultrasound
modality and dynamic maneuver contributions to distinguishing suspicious from non-suspicious non-simple cysts. 4) To know when
to use interventional procedures and which to use. 5) To know how pre-test probability (screening vs. diagnosis) affects
interpretation and when to apply the rule of multiplicity.
ABSTRACT
As a general rule, most non-simple breast cysts are part of the benign fibrocystic spectrum and have little risk of being malignant.
Malignant breast cysts are uncommon and cystic malignancies usually look more like necrotic or hemorrhagic solid masses than
benign appearing cysts. Nevertheless, needs a systematic approach to evaluation of non-simple breast cysts in order not to miss
the uncommon cystic malignancy. We present an algorithmic approach to evaluating non-simple breast cysts that is derived from
the mammographic and solid mass algorithms, where we look for suspicious features first. If there are no suspicious findings, we
then look for definitively benign findings. If we cannot find benign findings, we look for probably benign features, and failing that,
classify the lesion as suspicious and obtain histology. We show the histologic basis for echogenic cyst fluid and a variety of benign
and malignant excrescences that cause a cyst to appear to be non-simple. We discuss the difference between complicated cysts
and complex cystic and solid masses in appearances and risk. We show how Doppler can help us assess complex cystic and solid
masses and complex clustered microcysts. We discuss the need for histologic rather than cytologic assessment of suspicious cystic
breast masses. We show the appearances of acutely and chronically inflamed cysts before and after aspiration, how Doppler can
help this assessment, and how these appearances differ from those of malignant cysts, and the need for gram stain and culture,
but not cytology. We present a variety of definitively benign appearances for non-simple breast cysts such as scintillating echoes,
fat fluid-levels, fluid-debris levels, milk of calcium, calcium oxalate crystals, and skin cysts, and present maneuvers to improve their
assessment. We also discuss the complicated cyst that has fluid so echogenic that it simulates a solid nodule, such as a
fibroadenoma, and a variety of methods of further evaluating such cysts, including shear wave and strain elastography and
aspiration. Finally, we discuss the rule of multiplicity and how multiple similar appearing non-simple cysts can be downgraded to BI-
RADS 2, especially during supplemental screening ultrasound for women with dense breasts on mammography.
Active Handout:A. Thomas Stavros
http://abstract.rsna.org/uploads/2018/18000959/06 BUS of cysts that MSES51C.pdf
MSES51D Breast Anatomy and Physiology
Participants
Ellen B. Mendelson, MD,MA, Chicago, IL (Presenter) Advisory Board, Delphinus Medical Technologies, Inc; Speaker, Siemens AG;
Advisory Board, Seno Medical Instruments, Inc; ; ;
LEARNING OBJECTIVES

1) Describe the anatomic composition of the adult breast. 2) Correlate physiology with specific pathologic occurrences in young,
pregnant, and lactating women. 3) Assess the value of whole breast US in management of bilateral benign-appearing masses.
 RC615

Advanced Breast Imaging Technologies

Thursday, Nov. 29 8:30AM - 10:00AM Room: E451A

BR MR NM
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Maxine S. Jochelson, MD, New York, NY (Moderator) Nothing to Disclose
For information about this presentation, contact:

jochelsm@mskcc.org
LEARNING OBJECTIVES
I will put my learning objectives in under my course itself and presume the other 2 presenters will do the same. Don't think there
needs to be a separate learning objective for the moderator?
Sub-Events
RC615A PET and PET/MRI
Participants
Amy M. Fowler, MD, PhD, Madison, WI (Presenter) Research support, General Electric Company
For information about this presentation, contact:
afowler@uwhealth.org
LEARNING OBJECTIVES
1) Describe current approaches for performing breast PET imaging. 2) Assess diagnostic performance of breast PET imaging for
extent of disease and therapy response evaluation. 3) Examine potential uses of PET/MRI for breast imaging.
RC615B Molecular Breast Imaging
Participants
Carrie B. Hruska, PhD, Rochester, MN (Presenter) Institutional license agreement, CMR Naviscan Corporation
For information about this presentation, contact:
hruska.carrie@mayo.edu
LEARNING OBJECTIVES
1) Describe MBI instrumentation and clinical protocol for low-dose imaging. 2) Assess performance of MBI in screening of women
with dense breasts. 3) Examine the potential role of MBI as an imaging biomarker of breast cancer risk.
RC615C Contrast Enhanced Mammography & Tomosynthesis
Participants
Maxine S. Jochelson, MD, New York, NY (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) The utility of tomosynthesis in the screening setting. 2) The technique and risks of Contrast Enhanced Mammography. 3)
Potential uses for Contrast Enhanced Mammography in the screening and diagnostic setting.
 MSRT52

ASRT@RSNA 2018: The Missed Breast Cancer - Causes and Cures

Thursday, Nov. 29 9:15AM - 10:15AM Room: N230B

BR OI
AMA PRA Category 1 Credit ™: 1.00
ARRT Category A+ Credit: 1.00

Participants
Michael N. Linver, MD, Albuquerque, NM (Presenter) Medical Advisory Board, Solis; Scientific Advisory Board, Real Imaging Ltd;
Scientific Advisory Board, Seno Medical Instruments, Inc
For information about this presentation, contact:
mammomike@aol.com
LEARNING OBJECTIVES

1) Identify the major biological features of breast cancers which cause some cancers to be missed on mammography. 2) Identify
the technical issues which cause some breast cancers to be missed on mammography. 3) Apply various strategies to prevent
breast cancers from being missed on mammography.
ABSTRACT
This session will cover the major reasons that breast cancers are sometimes missed on mammography. Emphasis will be placed on
the biological factors of breast cancers that the radiologist cannot control, and the technical factors related to the process of
obtaining the images that the radiologist can and should control. For all of these issues, strategies will be offered to prevent
missing cancers on mammography.
 SSQ01

Breast Imaging (Abbreviated MRI, Ultrafast Imaging and Artificial Intelligence)

Thursday, Nov. 29 10:30AM - 12:00PM Room: E450A

AI BR MR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Christiane K. Kuhl, MD, Aachen, Germany (Moderator) Nothing to Disclose
Ritse M. Mann, MD, PhD, Nijmegen, Netherlands (Moderator) Researcher, Siemens AG ; Researcher, Seno Medical Instruments, Inc;
Researcher, Identification Solutions, Inc; Researcher, Micrima Limited; Researcher, Medtronic plc; Scientific Advisor, ScreenPoint
Medical BV; Scientific Advisor, Transonic Imaging, Inc; Stockholder, Transonic Imaging, Inc
Sub-Events
SSQ01-01 Assessing the Accuracy of an Abbreviated Breast MRI Protocol Compared to a Full MRI Protocol in
Women with a Personal History of Breast Cancer
Thursday, Nov. 29 10:30AM - 10:40AM Room: E450A

Participants
Jennifer Gillman, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
Ari Borthakur, PhD,MBA, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Elizabeth S. McDonald, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Alice Chong, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Susan Weinstein, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
PURPOSE
Women with a personal history of breast cancer have an elevated lifetime risk for a second breast cancer. However, the current
American Cancer Society guidelines do not recommend MRI screening in this population. Multiple studies have demonstrated that
the sensitivity of Abbreviated breast MRI (AB-MRI) is similar to full diagnostic protocols (FDP-MRI). In this study, we retrospectively
evaluate the use of surveillance AB-MRI in women with a personal history of breast cancer.
METHOD AND MATERIALS
An IRB approved and HIPAA compliant reader study was performed on 398 consecutive women with a personal history of breast
cancer who underwent full protocol clinical breast MRIs from 9/13-12/15. There were 14 cancers detected (3.8%). An enriched
reader study was performed consisting of 68 cases including the 14 cancer cases. Non-cancer cases had at least 1 year of follow-
up. Interpretations from a limited image set simulating an AB-MR protocol (T2, pre, and post contrast) were compared with
interpretations of the FDP-MRI clinical study.
RESULTS
The AB-MR interpretations were compared with those from the full, clinical protocol. The sensitivity (SN), specificity(SP), positive
predictive value (PPV), and the negative predictive value (NPV) for the simulated AB-MR vs the FDP-MRI interpretations were: SN -
50% vs 71%, SP - 96% vs 77%, PPV- 74% vs 43%, NPV - 88% vs. 91%. The mean difference between reader 1 and reader 2 was
0.29 with 95% confidence interval: [-0.33, 0.90]. There were significantly fewer false positives with AB-MRI than FDP-MRI, but
more false negatives were observed with AB-MRI.
CONCLUSION
Our preliminary results show higher specificity at the expense of sensitivity in our simulated AB-MRI reads compared to FDP-MRI in
women with a history of breast cancer. Further evaluation is warranted.
CLINICAL RELEVANCE/APPLICATION
A simulated AB-MRI protocol resulted in fewer false positive exams than with a full, clinical MR protocol in women with a personal
history of breast cancer, however, more research is needed.
SSQ01-02 Abbreviated Breast MRI : 'Ultrafast' DISCO Acquisition for Lesion Characterization
Thursday, Nov. 29 10:40AM - 10:50AM Room: E450A

Awards
Student Travel Stipend Award
Participants
Audrey Milon, MD, Paris, France (Presenter) Nothing to Disclose
Isabelle Thomassin-Naggara, MD, Paris, France (Abstract Co-Author) Speakers Bureau, General Electric Company
Julie Poujol, PhD, Vandoeuvre-les-Nancy, France (Abstract Co-Author) Nothing to Disclose
Asma Bekhouche, Paris, France (Abstract Co-Author) Nothing to Disclose
Saskia Vande Perre, Paris, France (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
audrey.milon.am@gmail.com
PURPOSE
The purpose of our study was to evaluate the diagnostic performance of a dynamic acquisition over-sampling the first minute after
contrast administration in an abbreviated dynamic-contrast-enhanced (DCE) breast-MRI.
METHOD AND MATERIALS
153 women were retrospectively consecutively included between July 2016 and March 2017, regardless of indication. All these
women had a full breast- MRI protocol, including a DISCO ultrafast acquisition with 7 phases, and an enhanced lesion histologically
proven (age= 55 (28-88)). Two readers analyzed 179 lesions (73 benign, 5 B3, 101 malignant lesions) with BIRADS classification for
each protocol: an abbreviated protocol (T1-weighted, T2-wheigthed, DISCO, T1-fat suppressed VIBRANT 2mn after contrast
administration) and a standard full protocol with late post-contrast phases. Then readers studied DISCO's early enhancement curve
with the following semi-quantitative parameters: Wash-In Rate (WIR), Maximal Slope Increase (MSI), Enhancement Amplitude (EA),
and Time of Half Rising (THS). Heterogeneity was also assessed using Standard Deviation (STD) at the different DISCO phases.
RESULTS
176/179 (98%) lesions were detected by the abbreviated protocol regarding to the full protocol : 122 mass and 57 non-mass-like
enhancement or foci (medium size : 18mm). The 3 undetected lesions were benign. Malignant lesions showed a WIR, a MSI a EA
higher, a THS shorter and were more heterogeneous at all DISCO phases than benign lesions (p<0.01). In the group of masses with
benign morphology (n= 42), THS was shorter for the malignant lesions (39.1 sec) than for the benign lesions (44.6 sec) (p=0.01).
CONCLUSION

Including an additional ultrafast-scan in an abbreviated breast-DCE-MRI protocol enables the early enhancement study that is
useful for lesion characterization and is time efficient.
CLINICAL RELEVANCE/APPLICATION

DCE-abbreviated breast-MRI with ultrafast-scan is efficient for lesion detection and characterization; so might be considered as a
screening tool in intermediate-risk women.
SSQ01-03 Ultrafast Breast DCE-MRI in the Evaluation of Tumor Size: Potential Utility in Moderate to Marked
Background Parenchymal Enhancement
Thursday, Nov. 29 10:50AM - 11:00AM Room: E450A

Participants
Sooyeon Kim, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Nariya Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Rihyeon Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Sil Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Min Sun Bae, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Su Hyun Lee, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jung Min Chang, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Research Grant, General Electric Company
Woo Kyung Moon, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
river7774@gmail.com
PURPOSE
Ultrafast breast DCE-MRI allows imaging of early kinetics within the first 30 seconds after contrast injection, when the background
parenchymal enhancement (BPE) is minimal. This study was performed to explore the clinical utility of ultrafast MRI focusing on
tumor size evaluation according to the level of BPE.
METHOD AND MATERIALS

A total of 360 consecutive women (median age, 54 years; range, 26 - 82 years) with 361 tumors (49 DCIS and 312 invasive) who
underwent both the ultrafast and conventional breast MRI before surgery were included. Ultrafast MR images were obtained using
TWIST or 4D-TRAK sequence (temporal resolution, 4.5 sec; voxel size, 1.1×1.1×1.0 mm3, TR/TE 4.1/1.3 ms). Then, conventional
DCE-MR images were obtained using 3D FLASH sequence (temporal resolution, 90sec; voxel size, 0.8×0.8×1.0 mm3, TR/TE 4.7/1.7
ms). Tumor size was independently measured on each scan, respectively. Agreement between tumor sizes on MRI and those on
surgical histopathology was assessed using the intraclass correlation coefficient (ICC) analysis.
RESULTS
The ICC on ultrafast MRI was comparable to that on conventional MRI (ICC = 0.657 vs. 0.634, P =.598). For conventional MRI, the
ICC was lower in women with moderate to marked BPE (ICC = 0.568) than in women with minimal to mild BPE (ICC = 0.650) with
borderline significance (P =.080). However, no difference was found on ultrafast MRI (ICC = 0.625 for moderate to marked vs. 0.663
for minimal to mild BPE, P =.385). In women with moderate to marked BPE, the ICC was slightly higher on ultrafast MRI than that on
conventional MRI, although the difference was not statistically significant (ICC = 0.625 vs. 0.568, P = .236). No difference was
found for the ICC according to the age, menopausal status, family history, histologic type, ER positivity, HER2 positivity, and lesion
type on MRI (mass vs. non-mass enhancement) (All P >.05).
CONCLUSION
In women with moderate to marked BPE, tumor size measurement might be more accurate on ultrafast MRI than on conventional
MRI.
CLINICAL RELEVANCE/APPLICATION

In women with moderate to marked BPE, ultrafast MRI can be applied for more accurate evaluation of tumor extent.
SSQ01-04 Maximum Slope as a Kinetic Parameter Based on Ultrafast Dynamic Contrast-Enhanced MRI of the
Breast Using K-Space Weighted Imaging Contrast
Thursday, Nov. 29 11:00AM - 11:10AM Room: E450A

Participants
Akane Ohashi, Kyoto-hu, Japan (Presenter) Nothing to Disclose
Masako Y. Kataoka, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Shotaro Kanao, MD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Mami Iima, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Makiko Kawai, MD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Natsuko Onishi, MD, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Yuta Urushibata, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Katsutoshi Murata, Tokyo, Japan (Abstract Co-Author) Nothing to Disclose
Elisabeth Weiland, Erlangen, Germany (Abstract Co-Author) Employee, Siemens AG
Masakazu Toi, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Kaori Togashi, MD, PhD, Kyoto, Japan (Abstract Co-Author) Research Grant, Bayer AG; Research Grant, DAIICHI SANKYO Group;
Research Grant, Eisai Co, Ltd; Research Grant, FUJIFILM Holdings Corporation; Research Grant, Nihon Medi-Physics Co, Ltd;
Research Grant, Canon Medical Systems Corporation
For information about this presentation, contact:
amaoh@kuhp.kyoto-u.ac.jp
PURPOSE
To investigate the diagnostic performance and inter-reader agreement of the maximum slope (MS) in breast malignant from benign
lesions obtained by ultrafast dynamic contrast-enhanced magnetic resonance imaging (DCE MRI). Comparison with washout index
(WI) was performed with the focus on discrepant cases.
METHOD AND MATERIALS
In total, 141 enhancing lesions (89 malignant, 52 benign) were included. Ultrafast DCE MRI sequences were acquired using a k-
space-weighted imaging contrast (KWIC) sequence, obtained 0 to 1 min after gadolinium injection (3.75 s/frame; 16 frames) and
followed by standard DCE MRI. The MS was calculated its percentage relative enhancement per second (%/s). The inter-reader
agreement of MS values by two radiologists were evaluated using intra-class correlation coefficients (ICC). As a semi-quantitative
parameter for conventional DCE MRI, washout index (WI: signal intensity [SI] delay - SI early) / SI pre × 100 (%) was calculated.
The diagnostic performance (malignant/ benign differentiation) of the MS and WI was compared using ROC analysis.
RESULTS
Intra-class correlation coefficients (ICC) of the reading was 0.98 (95% confidence interval 0.97-0.99) for all, 0.96 (0.95-0.98) for
malignant lesions and 0.99 (0.97-0.99) for benign lesions. The average MS was 25.4%/s (standard deviation: SD, 11.2 %/s) for
malignant lesions and 11.8%/s (SD, 10.7 %/s) for benign lesions. The AUC of the MS (ICC: 0.98) was almost same as that of the
WI (0.83 vs. 0.82, respectively; P = 0.80). Using the optimal cut-off points determined by the Youden index (>9.76% /s for the MS
and <-23 % for the WI), MS tended to have higher sensitivity (92.1%) and specificity (65.4%) compared with WI (91.1% and
61.5%, respectively). False positive cases based on MS were FA (n=5) and intraductal papilloma (n=1), while false positive cases
based on WI were fibrocystic change (n=6), intraductal papilloma (n=2) and flat epitherial atypia.
CONCLUSION

The overall diagnostic performance of MS in breast lesion was similar to the conventional kinetic parameter, with AUC of over 0.8.
Excellent ICC was obtained. MS helped to reduce false positive in fibrocystic change, while FA tended to be false positive on MS.
CLINICAL RELEVANCE/APPLICATION

Our results suggest that maximum slope can be an alternative kinetic parameter to conventional kinetic curve, potentially shorten
scan time, with excellent inter-reader agreement.
SSQ01-05 Combination of an Ultrafast TWIST VIBE Dixon Sequence Protocol and Diffusion-Weighted Imaging
to a Highly Accurate Clinically Applicable Classification Tool for Suspicious Masses in Breast MRI
Thursday, Nov. 29 11:10AM - 11:20AM Room: E450A

Participants
Stephan Ellmann, MD, Erlangen, Germany (Presenter) Nothing to Disclose
Sandra Peter, Erlangen, Germany (Abstract Co-Author) Nothing to Disclose
Evelyn Wenkel, MD, Erlangen, Germany (Abstract Co-Author) Nothing to Disclose
Elisabeth Weiland, Erlangen, Germany (Abstract Co-Author) Employee, Siemens AG
Rolf Janka, MD, PhD, Erlangen, Germany (Abstract Co-Author) Nothing to Disclose
Michael Uder, MD, Erlangen, Germany (Abstract Co-Author) Speakers Bureau, Bracco Group Speakers Bureau, Siemens AG Speakers
Bureau, Bayer AG Research Grant, Siemens AG
PURPOSE

To develop a statistical model for classification of suspicious masses in breast MRI when using TWIST VIBE Dixon (TVD) dynamic
sequences in combination with diffusion-weighted imaging (DWI) and compare it to a model based on a combination of conventional
dynamic contrast enhancement (DCE) and DWI. As ultrafast TVD sequences offer the potential to shorten breast MRI protocols,
diagnostic accuracy might be hampered due to reduced kinetic information. A special focus of this study was thus to maintain high
diagnostic accuracy in lesion classification.
METHOD AND MATERIALS
65 patients underwent clinically indicated breast MRI between 02/2014 and 04/2015, with 83 reported lesions (60 malignant, 23
benign). Inclusion criteria were suspicion of breast cancer or pre-therapeutic staging. Patients with non-mass-enhancements only
were excluded. The protocol consisted of our institute's standard protocol complemented by an ultrafast TVD sequence. The
apparent diffusion coefficient (ADC) and the peak enhancement of the TVD sequences were used to calculate a generalized linear
model (GLM) for prediction of malignancy. A second model was calculated using ADC and the curve type derived from the
conventional DCE sequence for the sake of comparison. Generalizability was ensured by applying leave-one-out cross validations.
For easy application of the GLMs in clinical workflows, nomograms were created.
RESULTS

The GLM based on peak enhancement of the ultrafast TVD sequences and ADC performed comparably accurate to the model based
on conventional DCE and ADC (Sensitivity 93.3% vs. 93.3%, specificity 91.3% vs. 87.0%, positive predictive value 96.6% vs.
94.9%, negative predictive value 84.0% vs. 83.3%; no significant differences).
CONCLUSION

This study presents a method to integrate ultrafast TVD sequences into a breast MRI protocol and reduce examination time while
maintaining diagnostic accuracy. A GLM based on the combination of TVD-derived peak enhancement and ADC provides high
diagnostic accuracy. The GLM can easily be applied in clinical routine using the supplied nomograms.
CLINICAL RELEVANCE/APPLICATION

One limiting factor hampering the comprehensive application of breast MRI is time. This study presents a breast MRI protocol with
less than 5 minutes duration along with a classification scheme reaching high diagnostic accuracy. Use of this protocol could
improve patient throughput and strenghten the role of breast MRI in screening.
SSQ01-06 Ultrafast Dynamic Contrast-Enhanced MRI for Detection of Invasive Components in Cases of Breast
Ductal Carcinoma in Situ by Biopsy
Thursday, Nov. 29 11:20AM - 11:30AM Room: E450A

Participants
Naoko Mori, MD, PhD, Sendai, Japan (Presenter) Nothing to Disclose
Hiroyuki Abe, MD, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Shunji Mugikura, MD, PhD, Sendai, Japan (Abstract Co-Author) Nothing to Disclose
Kei Takase, MD, PhD, Sendai, Japan (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

naokomori7127@gmail.com
PURPOSE

To evaluate whether ultrafast dynamic contrast-enhanced (DCE) MRI could identify invasive components in cases with ductal
carcinoma in situ (DCIS) diagnosed by percutaneous biopsy.
METHOD AND MATERIALS

Fifty-three consecutive women with 53 lesions diagnosed with DCIS by biopsy underwent IRB-approved ultrafast DCE-MRI including
a pre- and 18 post-contrast ultrafast 3D bilateral scans using a 3T system. Ultrafast 3D bilateral scans were acquired with temporal
resolution of 3 seconds per image. We evaluated the heterogeneity of enhancement in a target lesion using model-based analysis.
Regions of interest (ROIs) were placed where the strongest and weakest signal increases were found in ultrafast DCE-MRI to obtain
kinetic curves of maximum and minimum enhancement, respectively. The kinetic curve obtained from ultrafast DCE-MRI was
analyzed using an empirical mathematical model: ΔS(t)=A*(1-e-at). Where A is the upper limit of the signal intensity, a (min-1) is
the rate of signal increase. The initial slope of the kinetic curve is given by 'A*a'. Amax, Amin, amax, amin, A*amax, and A*amin
were obtained from ROIs for maximum and minimum enhancement, respectively. We obtained the following derivations for diagnostic
parameters showing heterogeneity of enhancement: A difference=Amax - Amin; a difference= amax -amin; A*a difference= A*amax
- A*amin.
RESULTS
Surgical specimens revealed 32 lesions with pure DCIS and the remaining 21 lesions with DCIS with invasive components (DCIS-IC).
The A difference for DCIS-IC (132±235) was significantly higher than that of pure DCIS (49±34) (p = 0.013). No significant
difference was found for a difference and A*a difference (p = 0.24 and 0.46, respectively). Receiver operating curve analysis
revealed that the area under the curve of A difference was 0.70. The most effective threshold for A difference was 68, and the
sensitivity, specificity, positive predictive value and negative predictive value were 62% (13/21), 72% (23/32), 59% (13/22), and
74% (23/31), respectively.
CONCLUSION

The A difference could suggest the presence of invasive components in cases with DCIS diagnosed by biopsy.
CLINICAL RELEVANCE/APPLICATION
The A difference showing the heterogeneity of enhancement of lesions in ultrafast DCE-MRI might suggest the presence of invasive
components in cases of DCIS by biopsy.
SSQ01-07 Ultrafast Dynamic Contrast Enhanced Breast MRI in Differentiating between Subcentimeter
Carcinomas and Benign Lesions: Quantitative versus Qualitative Assessments
Thursday, Nov. 29 11:30AM - 11:40AM Room: E450A

Participants
Natsuko Onishi, MD, PhD, New York, NY (Presenter) Nothing to Disclose
Meredith Sadinski, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Peter Gibbs, BSC,PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Katherine M. Gallagher, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Mary C. Hughes, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Theodore M. Hunt, New York, NY (Abstract Co-Author) Nothing to Disclose
Danny F. Martinez, BSC,MSc, New York, NY (Abstract Co-Author) Nothing to Disclose
Amita Shukla-Dave, PhD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth A. Morris, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Elizabeth J. Sutton, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

natsucom1981@gmail.com
PURPOSE

Ultrafast dynamic contrast enhanced (UF-DCE) breast MRI, characterized by high temporal and spatial resolution, enables image
acquisition at multiple time points starting simultaneously with the beginning of contrast injection. In a preliminary study comparing
several quantitative parameters calculated from UF-DCE MRI, we determined bolus arrival time (BAT) and maximum slope (MS) were
most useful in the differentiation between subcentimeter carcinomas and benign lesions. This study aims to compare the
performance of these parameters with qualitative assessments of UF-DCE MRI.
METHOD AND MATERIALS

We identified female patients between February-October 2017 with a: 1) UF-DCE MRI as part of hybrid protocol with conventional
DCE MRI performed with a 3.0T MRI with a 16-ch coil and 2) biopsy proven BI-RADS 4-6 lesion. UF-DCE MRI were acquired
continuously 15 times during the approximately 60 sec (temporal resolution, 3.0-4.3 sec) starting simultaneously with the beginning
of contrast injection. BAT and MS were computationally calculated based on 3D volumetric segmentation. Qualitative assessments
were visually performed by a reader, identifying the time from scan start to the beginning of lesion enhancement (vBAT) and
evaluating the degree of enhancement relative to background parenchymal enhancement (vE) by a 4-point grading scale from
'prominent' to 'indistinguishable'. Wilcoxon signed-rank test or Pearson's chi-squared test were used for the statistical analyses. P
value <0.05 was considered satistically significant. The diagnostic performance was evaluated using areas under the receiver
operating characteristic curve (AUC).
RESULTS

In total, 77 subcentimeter lesions (carcinomas, 33 [43%]; benign lesions, 44 [57%]) were analyzed. BAT, MS and vBAT presented
significant difference between carcinomas and benign lesions (p=0.0004, p<.0001, p=0.0063), while vE did not (p=0.0607). AUCs of
BAT (0.737) and MS (0.790) were higher than those of vBAT (0.683) and vE (0.605).
CONCLUSION

Quantitative assessments of UF-DCE MRI presented higher performance than qualitative assessments in differentiating between
subcentimeter carcinomas and benign lesions.
CLINICAL RELEVANCE/APPLICATION

There is no standardized way to evaluate ultrafast DCE breast MRI. Although diagnostic utility of some quantitative parameters is
known, little is known about the performance of qualitative assessment, especially for subcentimeter lesions.
SSQ01-08 Comparison of Machine Learning Based Measurement and Visual Assessment of Fibroglandular Tissue
and Background Parenchymal Enhancement in Breast MR Imaging: A Preliminary Study
Thursday, Nov. 29 11:40AM - 11:50AM Room: E450A

Participants
Heeyoung Chung, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Sung-Hun Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Yoonho Nam, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga-Eun Park, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

hoonhoony@naver.com
PURPOSE

To design and validate a machine learning model for the measurement of fibroglandular tissue (FGT) and background parenchymal
enhancement (BPE) in breast MR imaging, and compare with estimation of radiologist according to BI-RADS categories.
METHOD AND MATERIALS
195 women (mean age, 54.9 years; range 30 - 86 years) who were diagnosed with invasive breast cancer and underwent
preoperative breast MR, between January and December 2017 were enrolled in this study. Two radiologists independently assessed
the categories of FGT and BPE of contralateral breast, using with axial precontrast, early dynamic contrast enhancement T1-
weighted image, and subtraction image between them. In case of discordance, two radiologists reached consensus. Machine
learning model was designed to measure the volume of whole breast, FGT and BPE, using nonnegative matrix factorization (NMF). In
this study, 50 and 145 samples were assigned to train and valid, respectively. Areas under the receiver operating characteristic
curve was used to assess model performance of predicting dense breast (FGT category c, d) and prominent BPE (BPE category c,
d). Correlation between the visual assessment of radiologist and machine learning based measurement was assessed using
Spearman correlation analysis.
RESULTS
With the machine learning model, AUC of prediction of dense breast were 0.971 (0.880-0.998) in training set and 0.902 (0.784-
0.968) in validation set. AUC of prediction of prominent BPE were 0.959 (0.912-0.985) in training set and 0.819 (0.746-0.848) in
validation set (P < .001). Correlation between machine learning based measurement and visual assessment by radiologist was r =
0.871 of FGT, and r = 0.523 of BPE, respectively (P < .001).
CONCLUSION

Machine learning model showed reliable predictive power for FGT and BPE assessment and close correlation with FGT assessment by
radiologist.
CLINICAL RELEVANCE/APPLICATION

FGT and BPE are known as risk factors for breast cancer and are associated with poor prognosis. Machine learning can provide
quantitative and objective information of FGT and BPE volume in breast MR imaging and can be helpful to predict patient's
prognosis.
SSQ01-09 Deep Learning of Breast MRI Tumor Volume Improves Tumor Proliferation Marker Ki-67 Estimation
Thursday, Nov. 29 11:50AM - 12:00PM Room: E450A

Participants
Dooman Arefan, PhD, Pittsburgh, PA (Presenter) Nothing to Disclose
Aly A. Mohamed, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Hong Peng, MD, Xiangtan, China (Abstract Co-Author) Nothing to Disclose
Wendie A. Berg, MD, PhD, Pittsburgh, PA (Abstract Co-Author) Nothing to Disclose
Jules H. Sumkin, DO, Pittsburgh, PA (Abstract Co-Author) Research Grant, Hologic, Inc; Research Grant, General Electric Company
Shandong Wu, PhD, MSc, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

wus3@upmc.edu
PURPOSE

Ki-67 is a commonly used immunohistochemistry marker for cellular proliferation in invasive tumors. A few recent studies showed
some association between Ki-67 and DCE-MR imaging features. We performed an investigation to compare effects of a 3D deep
learning approach versus conventional radiomic features in deriving breast DCE-MRI information to predict Ki-67 rate.
METHOD AND MATERIALS
In an IRB-approved retrospective study of 141 patients, we identified 141 breast DCE-MRI scans (2011-2016) at our institution. All
patients have the Ki-67 proliferation rates measured that are further categorized into High vs Low category according to a clinically
defined threshold of 14. Breast tumor volume were automatically segmented in 3D space from the first post-contrast breast MR
sequence images. From the segmented 3D tumor volume, we extracted 30 common radiomic features, including morphological and
contrast enhancement kinetic characteristics of the tumor volume; those features were fed to a logistic least absolute shrinkage
and selection operator (LASSO) regression model to predict High vs Low Ki-67 categories. Also, a 3D convolutional neural network
(CNN) deep learning model was used to perform the same prediction but directly using the original image of the segmented 3D tumor
volume (i.e., here no any pre-defined imaging features extracted nor used). We performed 10-fold cross-validation for both logistic
regression and deep learning model evaluation and used average AUC as the metric of model classification accuracy.
RESULTS
There are 102 and 39 patients in the High and Low Ki-67 category, respectively. The average of the Ki-67 was 28.05%±21.63. The
AUC of the logistic regression model was 0.74 (95% CI: 0.73-0.75) for 4 LASSO-selected top ranked radiomic features (1
morphological and 3 contrast-enhancement related), while the 3D deep learning model achieved an AUC of 0.80 (95% CI: 0.75-
0.85).
CONCLUSION
In this study, the 3D CNN deep learning-based approach that automatically identifies and organizes hierarchical imaging features for
predicting Ki-67 outperformed the LASSO regression model coupled with pre-defined radiomic features.
CLINICAL RELEVANCE/APPLICATION

Deep learning of breast DCE-MRI tumor volume using CNN models may improve interpretation on the association between
radiological images and the immunohistochemistry tumor proliferation marker Ki-67.
 VW 18

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Thursday, Nov. 29 10:30AM - 11:00AM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 49

Wide-angle Digital Breast Tomosynthesis and Contrast Enhanced Mammography Self-guided Reading
Sessions: Siemens Healthineers Vendor Workshop
Thursday, Nov. 29 10:30AM - 5:00PM Room: Booth 5530

Program Information
You are invited to our self-guided reading sessions with syngo® Breast Care workstations configured especially to allow you to
work at your own place at a time that suits you! A series of breast tomosynthesis and contrast enhanced mammography cases
presented as problem cases with a solution enables you to develop and test your reading skills.
 VW 50

Automated Breast Volume Scanner (ABVS) Self-guided Reading Sessions: Siemens Healthineers Vendor
Workshop
Thursday, Nov. 29 10:30AM - 5:00PM Room: Booth 5530

Program Information
With syngo® Ultrasound Breast Analysis (sUSBA) software, self-guided reading sessions with real clinical cases will enable you to
become familiar with the coronal plane while providing practical approaches to interpretation of 3D automated breast ultrasound.
 VW 19

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Thursday, Nov. 29 11:00AM - 11:30AM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 VW 20

Invenia ABUS 2.0 - Live Scanning Demo: GE Vendor Workshop

Thursday, Nov. 29 11:30AM - 12:00PM Room: Booth 8156

Participants
Doug Whisler, Sunnyvale, CA (Presenter)
Program Information
This thirty minute session will cover the latest technological advancements in ABUS design and performance. Attendees will learn
how improvements in workflow and image quality have the potential to increase cancer detection in women with dense breast
tissue. Registration is required; adding this session to the RSNA calendar tool alone does not secure your seat in this session. Click
the link below to register.
Registration
http://ge.cvent.com/events/ge-breast-health-advantage-workshop-rsna-2018-/agenda-
e57e0b47e9aa4f5ba89b1a0da1e829b9.aspx
 ML53

Machine Learning Theater: State-of-the-art Deep Learning for Breast Cancer Screening: Presented by Kheiron
Medical Technologies
Thursday, Nov. 29 12:00PM - 12:20PM Room: Machine Learning Show case North Hall

Participants
Tobias Rijken, London, United Kingdom (Presenter) Stockholder, Kheiron Medical Technologies Ltd
Program Information
Kheiron is at the cutting edge of deep learning technology for breast cancer screening. This session will cover our multi-site clinical
trial, and how we have achieved state-of-the-art results using deep learning technology. In addition, we will discuss: - data
processing at scale - deep learning infrastructures - the future of breast screening empowered by deep learning.
 BRS-THA

Breast Thursday Poster Discussions

Thursday, Nov. 29 12:15PM - 12:45PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

Participants
Victoria L. Mango, MD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
BR010-EB- Imaging Findings of Breast Implant Complications: What Has Changed with Latest-Generation
THA Implants?
Hardcopy Backboard
Participants
Vinicius C. Felipe SR, MD, Sao Paulo, Brazil (Presenter) Nothing to Disclose
Marilia M. Azevedo, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Almir Bitencourt, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Juliana A. Souza, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Luciana Graziano, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Camila Guatelli, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Maria Luiza D. Albuquerque, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Elvira F. Marques, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Mirian R. Poli, MD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
Rubens Chojniak, MD, PhD, Sao Paulo, Brazil (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS
- Earlier-generation implants have high rates of complications, such as contracture and rupture. Latest-generation implants reduced
the number of complications by modifying its composition (cohesive silicone gel) and shell structure (stronger with textured
surface). Thereby, imaging aspects of breast implants complications has also changed. - Some breast implant complications became
less frequent with latest-generation implants, including displacement, rotation, capsular contracture and rupture. Besides, classical
imaging signs of breast implant rupture, such as 'teardrop' or 'keyhole', 'salad oil' and 'linguini' signs, may not be seen on modern
implants. Because of the semisolid consistence of these implants, silicone gel does not mix with water / saline content, and thus,
rupture looks more likely as a 'fracture' on imaging methods. - Some complications that were rarely seen before became more
frequent after the use of latest-generation textured implants, such as late seroma, silicone-induced granuloma of the breast
implant capsule and breast implant associated anaplastic large cell lymphoma.
TABLE OF CONTENTS/OUTLINE

Introduction breast implant types complications of latest generation implants implant rupture late seroma silicone-induced
granuloma breast implant associated anaplastic large cell lymphoma.
BR268-SD- Prediction of Axillary Response by Monitoring with Ultrasound and MRI During and After
THA2 Neoadjuvant Chemotherapy in Breast Cancer Patients
Station #2
Participants
Na Lae Eun, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Ji Hyun Youk, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Ju Son, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hye Mi Gweon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jeong-Ah Kim, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
PURPOSE
To investigate whether monitoring with ultrasound (US) and MR imaging can predict axillary response in breast cancer patients after
neoadjuvant chemotherapy (NAC).
METHOD AND MATERIALS
A total of 131 breast cancer patients with clinically positive axillary lymph node (LN) who underwent NAC and subsequent surgery
from January 2012 to August 2017 were enrolled. They had US and 3T MR examinations for breast and axilla before, during, and
after NAC. After reviewing US and MR images obtained at three different times, the number, size, shape, presence of hilum, and
cortical thickness of axillary LNs as well as index tumor size were noted. According to LN status after surgery, imaging features
were analyzed by using independent t-test, the Fisher exact test, multiple logistic regression analysis, and ROC analysis.
RESULTS
Of 131 patients, 60 (45.8%) had positive LNs after surgery. All US and MR features of LN before NAC showed no difference in LN
status. There was significant difference in transverse diameter, cortical thickness and its % change, and tumor size and its %
change during NAC and cortical thickness after NAC at US (P<0.003), and transverse diameter, cortical thickness and its % change,
hilum, and tumor size and its % change during and after NAC at MR (P<0.03). On multivariate analysis, cortical thickness at US
during NAC (odd ratio [OR], 1.8; 95% CI, 1.2-2.6; P=0.005) and tumor size at US (OR, 1.06; 95% CI, 1.0-1.1]; P=0.038) and
MR(OR, 1.04; 95% CI, 1.007-1.079; P=0.019) after NAC were independently associated with positive LN. The area under the ROC
curve for predicting LN status was 0.741, 0.639, and 0.692 for cortical thickness at US during NAC and index tumor size on US and
MR after NAC, respectively.
CONCLUSION
The US cortical thickness of axillary LNs during NAC and the index tumor size on US and MR after NAC can be useful to predict
axillary response in breast cancer patients.
CLINICAL RELEVANCE/APPLICATION
Monitoring morphologic features of axillary LNs and index tumor by US and MR imaging can help making treatment decision in breast
cancer patients receiving NAC.
BR275-SD- "Virtual" Full-Dose (VFD) Technology: Radiation Dose Reduction in Digital Breast Tomosynthesis
THA3 (DBT) by Means of Neural Network Convolution (NNC) Deep Learning
Station #3
Participants
Junchi Liu, MS, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Amin Zarshenas, MSc, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Syed Ammar Qadir, Chicago, IL (Abstract Co-Author) Nothing to Disclose
Limin Yang, MD, PhD, Iowa City, IA (Abstract Co-Author) Nothing to Disclose
Laurie L. Fajardo, MD, MBA, Park City, UT (Abstract Co-Author) Consultant, Hologic, Inc; Consultant, Siemens AG; Consultant,
FUJIFILM Holdings Corporation;
Kenji Suzuki, PhD, Chicago, IL (Presenter) Royalties, General Electric Company; Royalties, Hologic, Inc; Royalties, MEDIAN
Technologies; Royalties, Riverain Technologies, LLC; Royalties, Canon Medical Systems Corporation; Royalties, Mitsubishi
Corporation; Royalties, AlgoMedica, Inc
For information about this presentation, contact:
jliu118@hawk.iit.edu
TEACHING POINTS
1) To understand the basics of a deep-learning-based "virtual" full-dose technology. 2) To demonstrate and compare the image
quality of our VFD images generated from quarter- and half-dose acquisitions to that of real clinical full-dose images in DBT. 3) To
understand the clinical utility of VFD technology for reducing radiation dose in DBT.
TABLE OF CONTENTS/OUTLINE

Content OrganizationA. Radiation dose issues with DBT breast cancer screeningB. Basics of VFD technology for reducing radiation
doseC. Quantitative evaluation: Image quality vs. radiation dose reductionD. Blinded observer rating study with 35 breast
radiologists: virtual vs. real imagesE. Benefits and limitations of our radiation dose reduction technology for DBT This exhibit
presents1) Details on our deep-learning-based VFD technology that converted 25% dose images of cadaver phantoms to VFD
images, retaining the image quality equivalent to 119% dose images and achieving a 79% dose reduction. 2) The image quality of
VFD images of clinical cases was equivalent to that of real full-dose images. Our technology significantly reduced noise in half-dose
images, while preserving tissue and lesions. 3) Blinded observer rating study of 51 clinical cases: 60% of 35 radiologists preferred
our VFD images over real full-dose images or could not distinguish between the two.
PDF UPLOAD

http://abstract.rsna.org/uploads/2018/18004421/18004421_8gfq.pdf
BR270-SD- Cognitive Bias in Screening Mammography
THA4
Station #4
Participants
Ashutosh Shelat, MD, Santa Barbara, CA (Presenter) Nothing to Disclose
Myrna Wallace-Servera, MD, Santa Barbara, CA (Abstract Co-Author) Nothing to Disclose
Michael A. Trambert, MD, Santa Barbara, CA (Abstract Co-Author) Medical Advisor, IBM Corporation
Jeremy M. Wolfe, PhD, Cambridge, MA (Abstract Co-Author) Research collaboration, Koninklijke Philips NV;
Murray A. Reicher, MD, Rancho Santa Fe, CA (Abstract Co-Author) Chief Medical Officer, Merge Healthcare Incorporated Board
Member, Merge Healthcare Incorporated Co-CEO, Health Companion, Inc Former Chairman, DR Systems, Inc
PURPOSE
Gambler's fallacy and reverse gambler's fallacy are logical missteps based on the mistaken premise that previous outcomes of a
random event will affect future outcomes. Our purpose is to determine how significantly such a bias affects the reading of
screening mammograms, and specifically, to determine if recalling of one examination affects the likelihood of recalling subsequent
examinations.
METHOD AND MATERIALS

Over 70000+ screening digital mammographic results from two separate enterprises including 22 different radiologists over a span of
less than 5 years were obtained. Exams after the implementation of tomosynthesis were excluded. The data was anonymized, but
each radiologist was assigned a number. Exams were batched into different time frames (5 minutes, 10 minutes, 30 minutes, and 1
hour). Runs of exams were analyzed to determine if recalling a case would significantly impact the recall of subsequent cases.
RESULTS
Initial results demonstrated that the probability of recalling a case after a recent recall was significantly lower than the probability
of recalling any single case alone. However, after excluding the first study there was no significant difference between the relative
recall rates. Curiously, batching exams and aggregating results demonstrated that the recall rate for the first position of each batch
was statistically higher than the recall rate for all other examinations in a batch at an aggregate level for all time frames.
Additionally, using a 5 minute batch time, the results were statistically significant for all radiologists individually. Statistical
significance was achieved for the majority of all radiologists at all other batching time frames.
CONCLUSION

We found that radiologists do not suffer from the general premise of gambler's fallacy. Assuming no artifactual bias, however we
found that they appear to be primed to recall the first case in a series. This may represent an unknown cognitive bias in screening
mammography.
CLINICAL RELEVANCE/APPLICATION

Sensitivity of the first examination of any batch of screening digital mammographic examinations is expected to be higher on
average than other examinations and may reflect the only non-biased result.
BR271-SD- Imaging Features Associated with Pathological Complete Response in HER2 Positive Breast Cancer
THA5 after Neoadjuvant Chemotherapy with Dual Blockade
Station #5
Participants
Ga Young Yoon, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Eun Young Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hak Hee Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Joo Hee Cha, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hee Jung Shin, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Woo Jung Choi, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
3770ghwo@hanmail.net
PURPOSE
In human epidermal growth factor receptor 2 (HER2)-positive breast cancer, the incorporation of dual HER2-blockade to
neoadjuvant chemotherapy (NAC) has been shown to induce a higher rate of pathologic complete response (pCR). The purpose of
this study was to investigate associations between imaging features and pCR in HER2-positive breast cancer after NAC plus dual
blockade.
METHOD AND MATERIALS

This retrospective study was approved by the institutional review board. We evaluated 73 consecutive patients (mean age, 50 ±
9.99 years) with HER2-positive breast cancer who underwent NAC plus dual blockade with trastuzumab and pertuzumab between
April 2016 and March 2018. All patients had mammography, ultrasound, and MR imaging prior to NAC. pCR following NAC was defined
as the absence of residual invasive cancer in the breast and ipsilateral lymph nodes (ypT0/is, ypN0). Clinicopathological and initial
imaging features before NAC were assessed and compared according to the pathological response after surgery.
RESULTS
Of 73 patients, 41 (56.2%) showed pCR and the remaining 32 (43.8%) showed non-pCR. Segmental distribution of calcification on
mammography (odds ratio [OR], 13.57; P = 0.027), parallel orientation on ultrasound (OR, 4.03; P = 0.007), and the presence of
intratumoral high signal intensity on T2-weighted MR images (OR, 2.81; P = 0.037) were significantly associated with pCR.
Progesterone receptor-negative tumors (OR, 6.33; P = 0.004) were significantly associated with pCR. The presence of
mammographic calcification with or without a mass, associated nonmass enhancement on MR images, fine pleomorphic or fine linear
branching calcification morphology on mammography, and higher tumor infiltrating lymphocytes level were more common in the pCR
group, although these did not reach statistical significance.
CONCLUSION

Several imaging features showed association with pCR in HER2-positive breast cancer after NAC with dual blockade treatment.
CLINICAL RELEVANCE/APPLICATION

Our results may help differentiate patients who can benefit from adding dual blockade in HER2-positive breast cancer and determine
treatment.
BR272-SD- Comparison between Abbreviated Protocol (AB-MR) and Full Diagnostic Protocol (FD-MR) in the
THA6 Characterization of Lesions Detected by Breast MRI: A Multi-reader Study
Station #6
Participants
Eun Sil Kim, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Nariya Cho, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sooyeon Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ann Yi, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Rihyeon Kim, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Bo Ra Kwon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Min Sun Bae, MD, New York, NY (Abstract Co-Author) Nothing to Disclose
Su Hyun Lee, MD, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jung Min Chang, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Research Grant, General Electric Company
Woo Kyung Moon, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

esk19@naver.com
PURPOSE
Abbreviated breast MRI (AB-MR) has been shown to offer equivalent diagnostic accuracy compared with the respective full-
Abbreviated breast MRI (AB-MR) has been shown to offer equivalent diagnostic accuracy compared with the respective full-
diagnostic MRI (FD-MR) in a screening setting. We compared the diagnostic performance of AB-MR and FD-MR in the
characterization of lesions detected by MRI.
METHOD AND MATERIALS

A cohort of 111 biopsy-proven cases (34 malignant, 77 benign; median lesion size 1.2cm, range 0.3-6.6cm) identified at
contralateral breast MRI screening during the preoperative evaluation of newly diagnosed breast cancer patients between January
2011 and August 2016 was used for the reader study. Three blinded radiologists independently classified the likelihood of
malignancy and BI-RADS final assessment category for imaging data sets, i.e., AB-MR and FD-MR with 1-week interval. AB-MR
consisted of fat saturated pre-contrast, first post-contrast (90 sec) T1WI, and maximum-intensity projection (MIP) reconstruction
images. FD-MR consisted of T2WI, fat saturated pre-contrast, and five post-contrast T1WI series. Sensitivity, specificity, and
areas under the receiver operating characteristic (AUC) curve in distinguishing benign from malignant lesions were compared
between both protocols.
RESULTS

Sensitivity of AB-MR was slightly lower than that of FD-MR in all readers (82.4% [28/34] vs. 85.3% [29/34] in reader 1, P>.999;
82.4% [28/34] vs. 100% [34/34] in reader 2, P= not applicable; 58.8% [20/34] vs. 82.4% [28/34] in reader 3, P=.077). Specificity
of AB-MR was higher than that of FD-MR (41.6% [32/77] vs. 36.4% [28/77], P=.503; 39.0% [30/77] vs. 19.5% [15/77], P=.001;
74.0% [57/77] vs. 37.7% [29/77], P<.001), although statistical significance was only found for specificity differences in two
readers. AUC of AB-MR was comparable to that of FD-MR in all readers (0.706 vs. 0.705, P=.981; 0.700 vs. 0.685, P=.765; 0.738
vs. 0.698, P=.542).
CONCLUSION

Compared with FD-MR, AB-MR missed one, six, or eight of 34 cancers for each reader in the characterization of lesions detected by
MRI, although overall performances were similar in both protocols.
CLINICAL RELEVANCE/APPLICATION

Due to the limited characterization performance, AB-MR cannot replace FD-MR in a diagnostic setting.
BR211-ED- Easily Seen, Difficult Diagnosis: Spectrum of Lesions Involving the Nipple
THA7
Station #7
Participants
Rosa M. Lorente-Ramos, MD, PhD, Madrid, Spain (Presenter) Nothing to Disclose
Javier Azpeitia Arman, MD, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Carlos Oliva Fonte Sr, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Soledad Alonso Garcia, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
Almudena Blazquez Saez, MD, Salamanca, Spain (Abstract Co-Author) Nothing to Disclose
Eva Balbin, Madrid, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

rosa.lorenteramos@salud.madrid.org
TEACHING POINTS
To review the different causes of nipple lesions, both benign and malignant. To illustrate imaging findings of lesions involving the
nipple in different techniques, including mammogram, US and MR, providing clinical images and pathologic correlation. To discuss the
appropriate management of these lesions, emphasizing pitfalls, diagnostic difficulties and differential diagnosis.
TABLE OF CONTENTS/OUTLINE

Apart from breast cancers extending to the nipple, different lesions may appear in this area. From 2009 to 2016, 90 nipple biopsies
were performed at our Institution. We review nipple anatomy, as well as clinical and radiological imaging and pathology of the
different entities. We also analize special considerations of this challenging region for diagnosis, interventional procedures and
treatment. We present: Congenital: accesory nipple; Dermal lesions: Epidermal inclusion cyst, fibroepithelial polips, dermatitis,
melanoma and benign melanocytic nevus; Paget´s disease; Inflammatory lesions: galactophoritis, abscess; Nipple Tumors: adenoma
of the nipple; Breast lesions: papiloma, ductal carcinoma.
BR212-ED- Systematic Approach to Lesions Involving Nipple-Areolar Complex Using MRI
THA8
Station #8
Participants
Akane Ohashi, Kyoto-hu, Japan (Presenter) Nothing to Disclose
Masako Y. Kataoka, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Shotaro Kanao, MD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Maya Honda, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Mami Iima, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Ayami Ohno Kishimoto, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Kanae K. Miyake, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Tatsuki Kataoka, MD, PhD, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Takaki Sakurai, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Masakazu Toi, Kyoto, Japan (Abstract Co-Author) Nothing to Disclose
Kaori Togashi, MD, PhD, Kyoto, Japan (Abstract Co-Author) Research Grant, Bayer AG; Research Grant, DAIICHI SANKYO Group;
Research Grant, Eisai Co, Ltd; Research Grant, FUJIFILM Holdings Corporation; Research Grant, Nihon Medi-Physics Co, Ltd;
Research Grant, Canon Medical Systems Corporation
For information about this presentation, contact:

amaoh@kuhp.kyoto-u.ac.jp
TEACHING POINTS
The nipple-areolar complex (NAC) is located at the unique crossroad between skin and mammary duct. Therefore, wide variety of
lesions originated from skin or mammary duct can grow there, making their diagnosis challenging. Some parenchymal lesions invade
to the NAC. Identifying locations of NAC lesions in relation to anatomy and knowing typical progression pattern is crucial in
diagnosis and surgical planning. Lesions involving NAC may be continuous to the parenchymal component, which is often
overlooked. MRI is an ideal tool to examine NAC involvement and associated underlying parenchymal disease by obtaining cross-
sectional imaging with excellent tissue contrast.In this exhibit, we 1. review the anatomical structure of NAC and underlying
parenchyma, 2. illustrate the location of NAC lesions, 3. classify NAC lesions based on their location, and 4. propose systematic
approach to lesions involving NAC based on cross-sectional information on MRI.
TABLE OF CONTENTS/OUTLINE
1. Anatomy of NAC in relation to the ductal system of the breast parenchyma. 2. Classifications of lesions involving NAC based on
their location. 3. Case presentations, including nipple adenoma, Paget's disease, DCIS with intraductal spread, Pagetoid spread,
invasive cancer invading skin and nipple. 4. Systematic approach to lesions involving NAC.
BR213-ED- Preoperative Freehand Breast MRI Needle/Hookwire Localization of Lesions in Hard to Reach
THA9 Locations Where Grid Techniques Fail: Indications, Techniques, and Pitfalls
Station #9
Participants
Crystal Chang, MD, Stanford, CA (Presenter) Nothing to Disclose
Debra M. Ikeda, MD, Stanford, CA (Abstract Co-Author) Scientific Advisory Board, Grail, Inc; Reviewer, Siemens AG
Bruce L. Daniel, MD, Stanford, CA (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

To review indications for freehand MRI needle/hookwire localization in hard to reach locations where standard MRI grid techniques
may fail (examples: far posterior lesions, implants) To describe/demonstrate the step-by-step freehand MRI needle/hookwire
localization method To discuss potential pitfalls of each localization step
TABLE OF CONTENTS/OUTLINE
Indications for freehand MRI localization where grid techniques may fail Freehand MRI localization technique: A step-by-step
approach with examples Freehand localization technique pitfalls
BR214-ED- Intraoperative Use of 3D Printed Breast Models in the Setting of Breast Cancer
THA10
Station #10
Participants
Lumarie Santiago, MD, Houston, TX (Presenter) Nothing to Disclose
Cristina M. Checka, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Mark W. Clemens, Houston, TX (Abstract Co-Author) Nothing to Disclose
Beatriz E. Adrada, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Elsa M. Arribas, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

Lumarie.Santiago@mdanderson.org
TEACHING POINTS

1) 3D printed breast models allow accurate depiction of breast cancer extent by providing detailed relationships between the tumor,
chest wall, skin, nipple and surrounding blood vessels. 2) Intraoperative manipulation of 3D printed models complements the
information provided by concurrent breast imaging studies. 3) 3D printed breast models are useful in personalized patient education
regarding their disease and surgical management. 4) 3D printed breast models support multidisciplinary management of breast
cancer by enhancing surgical planning and execution.
TABLE OF CONTENTS/OUTLINE
We will present a pictorial essay of our experience in 3D printed breast models in the setting of multidisciplinary breast cancer
practice. 1) Review of imaging features of breast cancer cases in which a 3D printed breast model was requested 2) Preoperative
image guided localization correlating with the findings depicted in the 3D printed breast model and other breast imaging studies 3)
Intraoperative utilization and manipulation of 3D printed breast model during oncoplastic breast surgery.
BR215-ED- Non-mass Findings on Breast Ultrasound (US): Detection, Differential, Diagnosis
THA11
Station #11
Awards
Identified for RadioGraphics
Participants
Jihee Choe, MD, Boston, MA (Presenter) Nothing to Disclose
Sona A. Chikarmane, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Catherine S. Giess, MD, Wellesley, MA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

schikarmane@bwh.harvard.edu
TEACHING POINTS

The term 'nonmass finding' on ultrasound (US) is not part of current BI-RADS terminology, but is increasingly being used in the
radiology literature. Despite the current lack of a standardized approach to classification and evaluation of nonmass US findings,
recognition of this US finding can improve sonographic detection and correlation of mammographic and MRI lesions to help guide
biopsy. The purpose of this exhibit is to: 1) Review the various proposed classifications systems for nonmass US findings 2) Provide
imaging features of nonmass enhancement to help radiologists identify this sonographic finding 3) Demonstrate the wide range of
benign and malignant entities that may manifest as a nonmass finding 4) Correlate breast MRI and mammographic findings that may
present as sonographic nonmass findings
TABLE OF CONTENTS/OUTLINE
1) Review the definitions and various proposed classification systems for non-mass US findings described in the radiology literature
2) Illustrate the sonographic features of non-mass findings, any associated findings, and imaging techniques to detect non-mass
findings 3) Correlate sonographic nonmass findings with mammographic and breast MRI lesions to aid in biopsy and diagnosis 3)
Review benign and malignant etiologies for non-mass US findings through a rich pictorial review of illustrative cases
Honored Educators

Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Catherine S. Giess, MD - 2015 Honored EducatorCatherine S. Giess, MD - 2017
Honored Educator
BR210-ED- Genetic Testing for Hereditary Breast and Ovarian Cancer: A Primer for Radiologists
THA12
Station #12
Awards
Certificate of Merit
Identified for RadioGraphics

Participants
Puja Bharucha, MD, Baltimore, MD (Presenter) Nothing to Disclose
Fabienne Francois, Baltimore, MD (Abstract Co-Author) Nothing to Disclose
Nikki Tirada, MD, Baltimore, MD (Abstract Co-Author) Spouse, Research Grant, Siemens AG
Gauri R. Khorjekar, MD, Laurel, MD (Abstract Co-Author) Nothing to Disclose
Jessica Scott, Baltimore, MD (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

Explain the criteria for genetic risk evaluation that qualifies a patient to undergo genetic testing. Discuss the process of genetic
counseling, the different types of genetic testing used, and the implications of positive genetic results. Review the radiology
findings and imaging appearances of breast cancer typically associated with each mutation.
TABLE OF CONTENTS/OUTLINE

Breast imagers are often the first to initiate the conversation about genetic counseling with patients who have premenopausal
breast cancer or a strong family history of breast and ovarian cancer. Commercial genetic testing panels have gained popularity and
become more affordable in recent years. Therefore, it is imperative for radiologists to be able to provide counseling and to identify
which patients should be referred to genetic testing. Understanding of various breast cancer risk assessment tools such as Gail and
Tyrer-Cuzick models; and recognizing unique clinical presentations, specific imaging appearances, and genetic pedigree patterns
related to each mutation permits prompt identification of patients and their family members who carry mutations. The genetic test
results enable appropriate patient-specific screening that allows improvement of overall survival via early detection and timely
treatment.
 BRS-THB

Breast Thursday Poster Discussions

Thursday, Nov. 29 12:45PM - 1:15PM Room: BR Community, Learning Center

BR
AMA PRA Category 1 Credit ™: .50

FDA Discussions may include off-label uses.

Participants
Victoria L. Mango, MD, New York, NY (Moderator) Nothing to Disclose
Sub-Events
BR273-SD- Fully Automated Breast Lesion Segmentation on DCE-MRI Using a Convolutional Neural Network for
THB1 Radiomic Analysis
Station #1
Participants
Meghan A. Moriarty, MD, Port Jefferson, NY (Presenter) Nothing to Disclose
Karl D. Spuhler, MSc, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Jie Ding, MS, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Chunling Liu, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
Chuan Huang, PhD, Stony Brook, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
meghanannmoriarty@gmail.com
PURPOSE
To validate the accuracy of the automated segmentation of breast tumors using a fully convolutional neural network (CNN)
developed by our group. The goal is to create an automated DCE-MRI breast lesion segmentation system to be able to be utilized in
further radiomics studies in breast cancer, such as predicting sentinel lymph node metastases.
METHOD AND MATERIALS
This retrospective study was approved by the local IRB. A total of 316 DCE-MRI scans of breast cancer patients acquired on a 1.5T
MRI scanner were collected. We trained a CNN on a GPU server to automatically segment DCE-MRI breast lesions. Two radiologists
participated in this study. A total of 197 DCE-MRI scans were used for training the lesion segmentation neural network, these were
all drawn by radiologist 1. The network was then tested on a separate set of 119 DCE-MRI scans, using ROIs drawn by both
radiologists. Dice indices among the two radiologists' hand drawn ROIs and those generated by the automated network were
calculated.
RESULTS
On patient level, the mean Dice indices between ROIs were 0.64 (radiologist 1 vs 2), 0.60 (CNN vs radiologist 1), and 0.62 (CNN vs
radiologist 2). In order to reduce the effect of very small lesions on Dice index, the ROIs of all patients were grouped together; the
corresponding Dice indices were 0.64, 0.71 and 0.67. A Dice index of 0.7 is generally considered to have excellent agreement
between two segmentations in this condition . Comparing these two Dice indices indicates that the network segments lesions with
similar error rates to inter-human reader differences.
CONCLUSION
The proposed neural network-based automated breast lesion segmentation shows significant agreement compared to the labor-
intensive manual segmentation. The next step is to apply this in a task-based radiomic analysis to provide potential biomarkers and
guide clinical decisions for breast cancer patients.
CLINICAL RELEVANCE/APPLICATION
To determine whether such a fully automated neural network based segmentation could be employed for developing and
implementing radiomics pipelines.
BR274-SD- Comparison of Results from Three Different Density Assessment Methods on Mammographic Density
THB2 (MD) in Screening Patients Receiving Vitamin D (Vit D): Results of CALGB 70806 (Alliance)
Station #2
Participants
H. Carisa Le-Petross, MD, FRCPC, Houston, TX (Presenter) Nothing to Disclose
Drew K. Seisler, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Despina Kontos, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Heshan Liu, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Jayne R. Charlamb, MD, Syracuse, NY (Abstract Co-Author) Nothing to Disclose
Sin-Ho Jung, PhD, Durham, NC (Abstract Co-Author) Nothing to Disclose
James R. Marshall, PhD, Buffalo, NY (Abstract Co-Author) Nothing to Disclose
Marie Wood, MD, Burlington, NY (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
hlepetross@mdanderson.org
PURPOSE
To compare the different methods used for assessing MD in this screening population receiving Vit D for its breast cancer
prevention properties, in CALGB 70806, a randomized phase II trial.
METHOD AND MATERIALS
Premenopausal women randomized to receive either 2000IU of Vit D or placebo for 12 months had mammogram at baseline and at
12 months. MD was determined by Clinical Breast Imaging Reporting and Data System (BI-RADS), the semiautomatic software
Cumulus 6.0 (University of Toronto, Toronto, Canada), and fully automated method by the Laboratory for Individualized Breast
Radiodensity Assessment (LIBRA, by Computational Breast Imaging Group at University of Pennsylvania). Blinded central review of all
submitted mammograms was performed. Eligible women were premenopausal, age <55, with at least 25% dense breast tissue.
Kappa statics were used to measure agreement between local and central MD readings using BI-RADS. MD measurements were
compared using Wilcoxon rank-sum test.
RESULTS
300 women from 41 US centers were accrued from 2011 to 2013. 150 women received Vit D and 150 placebo. Mean age was 42.6
years with 14% Hispanic, 12% African American, and 74% Caucasian. 72% of participants completed treatment; the rest withdrew.
As previously reported, 1 year Vitamin D therapy did not significantly change MD (p=0.7048). Sub-analysis demonstrated moderate
agreement between local and central MD readings using BI-RADs classification at baseline and at 12 months, with Kappa
coefficients of 0.48 and 0.41 respectively. Increased MD from Cumulus and LIBRA were noted in heterogeneously dense and dense
BI-RADs cases (p <0.0001). When the readings for CC view was compared to MLO view, CC views showed slightly higher readings
at baseline and at 12 months (p=0.05 and 0.02 respectively). Cumulus readings were consistently higher than LIBRA readings (table
1, P <0.0001).
CONCLUSION
The subjective method of BI-RADS, semi-automated method CUMULUS, and automated method LIBRA were in agreement for the
majority of cases, with least variability noted from LIBRA than the other two methods.
CLINICAL RELEVANCE/APPLICATION
The automated method is more reliable and reproducible than the semi-automated or BI-RADs method for assessing breast density.
Support: UG1CA189823, U24CA196171. ClinicalTrials.gov Identifier: NCT01224678
BR269-SD- Comparison Between Radiation Dose of 2D Digital versus Digital Tomosynthesis Guided Stereotactic
THB3 Breast Biopsies: Tomosynthesis Wins!
Station #3
Participants
Tali Amir, MD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Bruno Barufaldi, PhD, Philadelphia, PA (Abstract Co-Author) Nothing to Disclose
Samantha P. Zuckerman, MD, Philadelphia, PA (Presenter) Nothing to Disclose
Andrew D. Maidment, PhD, Philadelphia, PA (Abstract Co-Author) Research support, Hologic, Inc; Research support, Barco nv;
Research support, Analogic Corporation; Spouse, Employee, Real-Time Tomography, LLC; Spouse, Stockholder, Real-Time
Tomography, LLC; Scientific Advisory Board, Real-Time Tomography, LLC;
Emily F. Conant, MD, Philadelphia, PA (Abstract Co-Author) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc;
Consultant, iCAD, Inc; Speaker, iiCME
PURPOSE

To compare radiation dose of digital breast tomosynthesis (DBT) versus digital mammographic (DM) guided breast biopsies
METHOD AND MATERIALS

We replaced a prone DM stereotactic biopsy (bx) unit with a prone dual mode, DM-DBT bx unit in 8/2017. All bxs performed 8/2016
to 1/2017 and 8/2017 to 1/2018 were retrospectively reviewed. The bxd finding, guidance modality (DM vs DBT), and Digital
Imaging and Communications in Medicine (DICOM) header data were recorded. Image metadata was extracted from the DICOM
header and stored into a client-side Structured Query Language (SQL) database. The average glandular dose (AGD) per image and
study were computed and stratified by modality and finding type.
RESULTS
25 DM guided bxs (24 calcifications,1 asymmetry) were performed on the DM unit between 8/2016 and 1/2017. The AGD per image
was 2.63 mGy (SD 1.16). The AGD per procedure was 28.77 mGy (SD 14.34) and average image number was 10.92. 97 DM/DBT
guided bxs (80 calcifications, 13 architectural distortions, 2 asymmetries, 2 masses) were performed on the DM-DBT unit from
8/2017 -1/2018. The AGD per image was 2.40 mGy (SD 1.09); DM AGD was 2.32 mGy (SD 1.11) while DBT AGD was 2.46 mGy (SD
0.95). The AGD per procedure was 18.18 (SD 13.66); DBT guided bxs had an AGD of 12.43 mGy (SD 9.08) while DM guided bxs had
an AGD of 21.20 mGy (SD 14.18). The average image number for DM bxs was 9.14 compared with 5.05 for DBT bxs. There was a
26% dose reduction for DM bxs on the dual mode unit compared to the DM only unit, due to a 12.7% reduced AGD per image
(p<0.0001). When assessing DBT compared with DM guided bxs on the dual unit, the DBT AGD was 36% lower than DM (p=0.0304),
despite a 13% increase in dose per individual DBT acquisition compared to a single DM view.
CONCLUSION

Significant dose reduction is achieved for mammographically guided bxs with newer equipment functioning with higher efficiency. In
addition, for bxs performed on a dual mode unit, use of DBT guidance can significantly reduce dose compared to using DM guidance.
This lower dose is due to a lower number of acquisitions in DBT guided bxs and because a single DBT acquisitions has a lower AGD
than DM stereo pairs.
CLINICAL RELEVANCE/APPLICATION

Newer, more efficient biopsy equipment and the use of DBT guidance can reduce the total radiation dose of mammographically
guided biopsy procedures.
BR276-SD- Comparative the Average Glandular Dose between Digital Breast Tomosynthesis (DBT) and Full-Field
THB4 Digital Mammography (FFDM): Correlation with Breast Thickness and Density
Station #4
Participants
Chanjuan Wen, Guangzhou, China (Presenter) Nothing to Disclose
Weimin Xu, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Hui Zeng, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Zilong He, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Weiguo Chen, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

182389433@qq.com
PURPOSE

To compare the average glandular dose (AGD) between single-view digital breast tomosynthesis (DBT) and single-view full-field
digital mammography (FFDM), and to evaluate the correlation of AGD with breast thickness and density.
METHOD AND MATERIALS
A total of 318 female patients who underwent both DBT and FFDM (DBT and FFDM were performed in the same compression
thickness in each breast) were included. 636 DBT images of unilateral breast mediolateral oblique (MLO) view and 636 FFDM images
of unilateral breast mediolateral oblique (MLO) view were analyzed. Mammographic breast density was determined according to BI-
RADS breast density grading, and breast thickness and AGD per exposure in MLO views retrieved from DICOM headers were
recorded. Breast thickness were divided into the following four groups: <= 30cm, 31 ~ 45cm, 46 ~ 60cm and > 60cm. The
statistical analyses used variance analysis and Pearson's correlation for parametric tests.
RESULTS

(1) The AGD of DBT had a weak negative correlation with breast density (correlation coefficient =-0.305, P<0.001), decreased as
the breast density increased. The AGD of FFDM did not change significantly with breast density increased (correlation coefficient
=-0.027, P=0.501). (2) Breast thickness was significantly associated with AGDs, and both AGDs of FFDM and DBT increased with
increased breast thickness (correlation coefficient =0.771 and 0.935, respectively, all P<0.001). (3) When breast density was >75%
and breast thickness was >60cm, the AGD of DBT was lower than that of FFDM, and the difference was statistical significant (P =
0.031).
CONCLUSION

The AGD of DBT increased with breast thickness increased and decreased with breast density. For thick and dense breast, the
radiation dose of DBT was lower than that of FFDM.
CLINICAL RELEVANCE/APPLICATION

In this study, we evaluated the AGD of MLO FFDM and DBT according to breast density and thickness.
BR277-SD- Fate of a BI-RADS 3 Lesion: An Analysis of the Characteristics, Follow Up, Diagnostic Workup, and
THB5 Cancer Rate of Probably Benign Lesions Seen at Breast MRI
Station #5
Participants
Margaret J. Wong, MD,MENG, Palo Alto, CA (Presenter) Nothing to Disclose
Rupa Patel, MD, Stanford, CA (Abstract Co-Author) Nothing to Disclose
Wendy B. Demartini, MD, Stanford, CA (Abstract Co-Author) Nothing to Disclose
Julia Todderud, Stanford, CA (Abstract Co-Author) Nothing to Disclose
Debra M. Ikeda, MD, Stanford, CA (Abstract Co-Author) Scientific Advisory Board, Grail, Inc; Reviewer, Siemens AG
PURPOSE

Our purpose was to evaluate the patient and imaging characteristics, follow-up frequency and timing, and rate of malignancy for
breast MRI examinations assessed as BI-RADS Category 3.
METHOD AND MATERIALS

From 4235 consecutive screening and diagnostic breast MRI examinations performed from May 2011 through December 2014, we
retrospectively identified examinations assessed as BI-RADS Category 3. For the study examinations, we collected patient
characteristics and breast MRI BI-RADS descriptors, follow-up frequency and timing, and biopsy outcomes. Benign versus malignant
outcome was determined by biopsy and/or imaging follow-up for two years. We calculated the frequencies of patient and imaging
characteristics, and the rates of follow-up compliance and of malignancy.
RESULTS

From 4235 consecutive breast MRI examinations in the study interval, 3.9% (167/4235) were assessed as BI-RADS Category 3.
88% (147/167) examinations were in patients designated at high risk for breast cancer. The most frequency MRI features
associated with BI-RADS Category 3 were minimal background parenchymal enhancement (50%, 83/167), lesion type mass (38%,
64/167), and lesion type focus/foci (39%, 65/167). Masses were most commonly oval (50%, 32/64) and circumscribed (39%,
25/64). Initial kinetics were most commonly medium (39%, 65/167), and late kinetics were most commonly persistent (60%,
100/167). The follow-up compliance rate was 75% (125/167) with an average time to follow up of 14.1 months. Out of the 125 MRI
examinations with adequate follow up, 21 lesions were biopsied (17%, 21/125) showing two cancers (1.6%, one IDC and one DCIS),
six high-risk lesions (4.8%; three papillomas, one ALH, one ADH, one low-grade spindle cell tumor) and 13 benign findings.
CONCLUSION

Breast MRI BI-RADS Category 3 was used in 3.9% of examinations, and the recommended follow-up occurred in 75%. The rate of
malignancy was 1.6% which is comparable to that for mammography, despite the higher risk MRI patient population.
CLINICAL RELEVANCE/APPLICATION

The BI-RADS MRI category 3 assessment category is not well established, largely due to variations in hardware, field strengths, and
pulse sequences between sites. Our analysis provides insight into what MRI characteristics warrant a BI-RADS 3 assessment.
Understanding this BI-RADS category is especially important to reduce unnecessary biopsies of probably benign lesions that can be
safely followed with short term follow up imaging.
BR216-ED- Superficial Breast Lesions That We Need to Know
THB6
Station #6
Participants
Marisela L. Curros, MD, Adrogue, Argentina (Presenter) Nothing to Disclose
Valeria Vidales, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Daniela E. Simbler, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Norma I. Pona, MD, CABA, Argentina (Abstract Co-Author) Nothing to Disclose
Felix Vigovich, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Adriana Garcia, MD, Banfield, Argentina (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

mcurros@hbritanico.com.ar
TEACHING POINTS

-To recognize the radiological features of the lesion that helps to identify its localization between the dermis, the subcutaneous fat
and the parenchyma. -To describe differential diagnosis and its classification in BI RADS system. -To emphasize in diagnostic
difficulties.
TABLE OF CONTENTS/OUTLINE
-Description of the anatomy of the superficial breast. -Description of different types of lesions that could be found in the breast
skin and the superficial breast parenchyma. -US and mammographic: principal findings and its correlations. -Sample cases.
BR217-ED- Mammographic Evaluation of Calcifications: Atlas for Residents
THB7
Station #7
Awards
Identified for RadioGraphics

Participants
Yesenia Bermudez Cano, MD, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Veronica Gonzalez, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Karina Pesce, Capital Federal, Argentina (Abstract Co-Author) Nothing to Disclose
Victoria Ardiles, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Maria Jose Chico, Buenos Aires, Argentina (Abstract Co-Author) Nothing to Disclose
Pamela I. Causa Andrieu, MD, La Plata, Argentina (Presenter) Nothing to Disclose
For information about this presentation, contact:

yesebeca@hotmail.com
TEACHING POINTS

1. Review the mammographic descriptors for breast calcifications, taking like reference the fifth edition of the BIRADS. 2. Analyze of
morphology and distribution of breast microcalcifications on mammography, thus aiding in their classification and management of
breast lesions. 3. Demonstrate the imaging features of calcifications and correlate with the histopathology obtained by Stereotactic
Large Core Needle Biopsy
TABLE OF CONTENTS/OUTLINE

Table of Contents: 1. Introduction 2. Images features: - Calcification descriptors - Distribution - Location - Associated findings -
Change, if previous films are compared - Use of microcalcification descriptors in BI-RADS 5th edition to stratify risk of malignancy 3.
Microcalcification in breast lesions: radio-pathologic correlation 4. Pseudocalcifications: Artifact 5. Management - Management of
microcalcifications that develop at the lumpectomy site after breast-conserving therapy 6. Conclusion.
BR218-ED- Contrast Enhanced Mammography: Where Does It Fail?
THB8
Station #8
Participants
Ignacio Gonzalez de la Huebra Rodriguez, MD, Pamplona, Spain (Presenter) Nothing to Disclose
Alejandra Garcia Baizan, MD, Pamplona, Spain (Abstract Co-Author) Nothing to Disclose
Ana Ezponda, MD, Pamplona, Spain (Abstract Co-Author) Nothing to Disclose
Marta Calvo-Imirizaldu, MD, Pamplona, Spain (Abstract Co-Author) Nothing to Disclose
Arlette Elizalde, Pamplona, Spain (Abstract Co-Author) Nothing to Disclose
Luis Pina, MD, PhD, San Sebastian, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
agarcia.13@unav.es
TEACHING POINTS

The purpose of this exhibit is: 1. To review the strengths, weaknesses and indications of Contrast Enhanced Mammography (CEM)2.
The purpose of this exhibit is: 1. To review the strengths, weaknesses and indications of Contrast Enhanced Mammography (CEM)2.
To show the features of breast tumors that were missed by CEM.
TABLE OF CONTENTS/OUTLINE

Description of the technique: -What is double energy?Intravenous administration of iodinated contrast medium (mammography
becomes a morpho-functional technique).Indications of CEM -Problem solving technique. -Screening of intermediate risk patients. -
Preoperative staging of breast cancer. -Patients not suitable for MRI.Contraindications of CEM: -Allergy to iodinated contrast
medium. -Renal insufficiency.Limitations of CEM -Evaluation of microcalcificationes. -Women with breast ImplantsFalse Negative
cases: -Lesions out of the field of view (peripherically located). -Pathology: More misses in DCIS than in invasive cancers. -Breast
density: More misses in dense patterns . -Imaging features: More misses in architectural distortions than masses. -Size: More
misses in small cancers (<10mm).False Positive cases: -Fibroadenoma -Papilloma -Fat necrosis -Sclerosing adenosis
BR219-ED- MRI Biopsy Radiology/Pathology Concordance: A Rapid Review of Common Pathologic Entities
THB9
Station #9
Participants
Marina Mohallem Fonseca, Ottawa, ON (Abstract Co-Author) Nothing to Disclose
Raman Verma, MD, Ottawa, ON (Presenter) Nothing to Disclose
Leslie Lamb, MD, Boston, MA (Abstract Co-Author) Nothing to Disclose
Jean M. Seely, MD, Ottawa, ON (Abstract Co-Author) Nothing to Disclose
TEACHING POINTS

This exhibit will: 1)Identify common benign and malignant pathologies diagnosed by MRI biopsy using an image rich review. 2)Outline
practical tips when assessing radiologic and pathologic concordance and the need for re-biopsy. 3)Provide an algorithm for
management of indeterminate lesions on MRI.
TABLE OF CONTENTS/OUTLINE

Breast MRI is routinely used in the screening and diagnostic settings, in conjunction with mammographic and sonographic
assessment. An indeterminate lesion on MRI, occult on conventional imaging, may require an MRI-guided biopsy for definitive
histopathologic analysis. Lesions biopsied under MRI guidance are frequently benign; however, there is variable and often
overlapping appearance of both benign and malignant etiologies, often rendering the concordance assessment difficult. Lesions
including the following will be illustrated, with all cases having follow-up imaging or surgical excision: 1)Benign (fibroadenoma,
fibrocystic change, pseudoangiomatous stromal hyperplasia, fat necrosis, mastitis) 2)High risk (lobular carcinoma in-situ, atypical
ductal hyperplasia, radial scar, papilloma) 3)Malignant (invasive ductal and lobular, ductal carcinoma in-situ) A summary based on
the literature will outline strategies to determine appropriate imaging modalities with which to further evaluate an MRI-detected
indeterminate lesion.
BR011-EB- Contrast Enhanced Spectral Mammography (CESM). The 'What, When and How' Guide to Using it in a
THB Symptomatic Tertiary Referral Center
Hardcopy Backboard
Participants
Rosanna Frost, MBBS,FRCR, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Ruxandra Pietrosanu, MD,FRCR, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Sultana Hasso, MBBS, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Konstantia Diana Stavrou, MBBS, BSC, London, United Kingdom (Presenter) Nothing to Disclose
For information about this presentation, contact:
rfrost@doctors.net.uk
TEACHING POINTS

The purpose of the exhibit to to outline: 1 Why to use CESM? 2 When to utilize CESM in clinical practice. 3 How to interpret CESM
TABLE OF CONTENTS/OUTLINE
Why do we use it? CESM is an evolving breast imaging technique, which combines standard full - field digital mammography (FFDM)
with an intravenous iodinated contrast medium to detect areas of increased angiogenesis. The technique is well tolerated by
patients, quick and cost effective whilst being highly sensitive comparable to CEMRI. When do we use it? The local staging of
patients under the age of 40 with a biopsy proven breast cancer who have not undergone a diagnositic mammogram due to their
age. The local staging of patients who have an indication for CEMRI but have a contraindication to MRI . Monitoring response to
neoadjuvant chemotherapy. We are currently conducting an ongoing study to demonstrate that CESM can be used as an
alternative to MRI for monitoring reponse during neoadjuvant chemotherapy. Annual monitoring of breast cancer patients who are
at high risk or who have dense breasts (birads 4-5). Current ongoing study. How to interpret it? Assessment of normal backgound
enhancement. Artefacts. The characteristics of benign vs malignanat mass lesions. Non mass enhancement. Non enhancing
malignanat calcification. Chest wall and retroareolar lesions. Our experience over the last three years.
BR007-EB- What Radiologists Should Know to Avoid Mistakes in Screening Breast US
THB
Hardcopy Backboard
Participants
Jin Hwa Lee, MD, Busan, Korea, Republic Of (Presenter) Nothing to Disclose
Eun Cho, MD, Busan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Young Mi Park, MD, PhD, Busan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

jhrad@dau.ac.kr
TEACHING POINTS

The purpose of this exhibit is To review the reasons for False-Negative and False-Positive in screening breast US To learn how to
reduce missed cancer and unnecessary recall or biopsy in screening breast US
TABLE OF CONTENTS/OUTLINE

Reasons for False-Negative Technical errors; high-resolution US equipment, adjustment of US settings, optimal scanning technique
Perception errors; isoechoic lesions, deeply located lesions in large breasts, peripherally located lesions, subareolar lesions, US
tissue composition; background echotexture Interpretation errors; misinterpretation of margin, multiple distracting lesions
Correlation errors; mammographic correlation, MRI correlation, clinical correlation (Symptomatic vs Asymptomatic, Patients' own risk
factors, past history (underlying extramammary disease) Reasons for False-Positive Inherent factor of US Technical errors Large
numbers of category 3 lesions
 MSCB51

Case-based Review of Breast (Interactive Session)

Thursday, Nov. 29 1:30PM - 3:00PM Room: N228

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Jiyon Lee, MD, New York, NY (Director) Nothing to Disclose
LEARNING OBJECTIVES

1) Identify appropriate application of multi-modality breast imaging for routine screening, supplemental screening, and diagnostic
indications. 2) Select appropriate methods for image-guided percutaneous biopsy and perform post-biopsy radiologic-pathologic
correlation for next management recommendation. 3) Review appropriateness criteria and performance benchmarks, and guidelines
for ongoing breast imaging audits as they apply. 4) Appreciate the range of reassuringly common and sometimes not-so common
among the international faculty's portrayal of their piece of the globe.
ABSTRACT
Title: Managing expectations in breast imaging around the world. "Best" versus sufficient? Abstract: Our case-based review course
will use the interactive audience response system (ARS) to walk and skip through the fundamentals of breast imaging. We will
present how we use mammography, ultrasound, and MRI in daily screening and diagnostic scenarios, along with reminders of the
overarching principles of BI-RADS lexicon for effective communication, and ACR appropriateness criteria and performance metrics as
applicable or adapted around the world. Our international faculty (sessions 1 and 2) will also add depth, and the fun added
dimensions of how breast imaging works around the world. Varying breast cancer statistics, possible innate ethnic variations,
differing cultural expectations and socioeconomic context can and do impact how we carry out our discretionary work. Such
interesting details will inform the narrative of the speakers' case scenarios, while the core diagnostic radiology skills aim to be
constant, and teachable. The focus is using lots of cases to demonstrate breast imaging now and evolving. Please join us for smart
fun!
Active Handout:Jiyon Lee
http://abstract.rsna.org/uploads/2018/18001608/MSCB51 52.pdf
Sub-Events
MSCB51A Tomosynthesis: Evolving Appreciation of the Better Mammogram
Participants
Jiyon Lee, MD, New York, NY (Presenter) Nothing to Disclose
MSCB51B From Andes to Patagonia: Breast Imaging in Argentina
Participants
Daniel E. Lehrer, MD, CABA, Argentina (Presenter) Speaker, Hologic, Inc; Institutional research agreement, Siemens AG
For information about this presentation, contact:
lehrerdan@cerim.com.ar
LEARNING OBJECTIVES
1) Identify appropriate application of multi-modality breast imaging for routine screening, supplemental screening, and diagnostic
indications. 2) Improve basic knowledge and skills relevant to clinical practice. 3) Recommend the appropriate technique and avoid
mistakes, incorporating others' clinical experiences.
ABSTRACT
We show cases from different parts of the country, with different realities and possibilities. This cases include a wide range of
sophistication, from the optimization of the basic knowledge to the ones that require the latest technologies. You can realize that
Tolstoy´s: Paint your village and you will paint the whole world is true for breast imaging.
Active Handout:Daniel E. Lehrer
http://abstract.rsna.org/uploads/2018/18001610/RSNA_Argentina MSCB51B.pdf
MSCB51C The Many 'Faces' of DCIS
Participants
Ana P. Lourenco, MD, Providence, RI (Presenter) Nothing to Disclose
For information about this presentation, contact:
alourenco@lifespan.org
LEARNING OBJECTIVES
1) Detect the varied appearances of DCIS on mammography, ultrasound, and MRI. 2) Compare how imaging findings may predict
pathology.
ABSTRACT
This case-based session will showcase the various appearances of DCIS on mammography, ultrasound and MRI, highlighting how
certain imaging findings may predict pathology. The interactive questions will cover management as well as follow-up
recommendations, and illustrate key findings that should be included in imaging reports.
Active Handout:Ana P. Lourenco
http://abstract.rsna.org/uploads/2018/18001611/Handout.Lourenco.DCIS MSCB51C.pdf
MSCB51D Spain Explains the Mundane and the Less So
Participants
Lucia Grana Lopez, MD, Lugo, Spain (Presenter) Nothing to Disclose
For information about this presentation, contact:
lucia.grana.lopez@sergas.es
LEARNING OBJECTIVES
1) Identify appropriate application of multi-modality breast imaging mainly for supplemental screening, diagnosis and define its
interventional indications. 2) Learn about the advantages of ultrasound-guided percutaneous removal of benign breast lesions and
when and how to perform this procedure, as we've experienced in our practice. 3) Preview emerging molecular breast dedicated
imaging tool, define its possible indications and potential use in clinical routine.
ABSTRACT
My case-based review course will use the interactive audience response system (ARS) to walk and skip through the fundamentals
of breast imaging. I will present how we use mammography, ultrasound, and MRI in daily supplemental screening and diagnostic
scenarios, along with reminders of the overarching principles of BI-RADS lexicon for effective communication, and ACR
appropriateness criteria and performance metrics as applicable.I will try to show how breast imaging works in Spain.
Active Handout:Lucia Grana Lopez

http://abstract.rsna.org/uploads/2018/18001612/Spain explains the mundane MSCB51D.pdf

 MSCB52

Case-based Review of Breast (Interactive Session)

Thursday, Nov. 29 3:30PM - 5:00PM Room: N228

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Jiyon Lee, MD, New York, NY (Director) Nothing to Disclose
LEARNING OBJECTIVES
1) Identify appropriate application of multi-modality breast imaging for routine screening, supplemental screening, and diagnostic
indications. 2) Select appropriate methods for image-guided percutaneous biopsy and perform post-biopsy radiologic-pathologic
correlation for next management recommendation. 3) Review appropriateness criteria and performance benchmarks, and guidelines
for ongoing breast imaging audits as they apply. 4) Appreciate the range of reassuringly common and sometimes not-so common
among the international faculty's portrayal of their piece of the globe.
ABSTRACT
Title: Managing expectations in breast imaging around the world. "Best" versus sufficient? Abstract: Our case-based review course
will use the interactive audience response system (ARS) to walk and skip through the fundamentals of breast imaging. We will
present how we use mammography, ultrasound, and MRI in daily screening and diagnostic scenarios, along with reminders of the
overarching principles of BI-RADS lexicon for effective communication, and ACR appropriateness criteria and performance metrics as
applicable or as adapted around the world. Our international faculty (sessions 1 and 2) will also add depth, and the fun added
dimensions of how breast imaging works around the world. Varying breast cancer statistics, possible innate ethnic variations,
differing cultural expectations and socioeconomic context can and do impact how we carry out our discretionary work. Such
interesting details will inform the narrative of the speakers' case scenarios, while the core diagnostic radiology skills aim to be
constant, and teachable. The focus is using lots of cases to demonstrate breast imaging now and evolving. Please join us for smart
fun!
Active Handout:Jiyon Lee
http://abstract.rsna.org/uploads/2018/18001613/MSCB51 52.pdf
Sub-Events
MSCB52A Breast Care for 'Challenging' Populations
Participants
Cherie M. Kuzmiak, DO, Chapel Hill, NC (Presenter) Research Grant, Delphinus Medical Technologies, Inc
LEARNING OBJECTIVES
1) Describe and discuss how to appropriately image and manage patients with special needs.
Active Handout:Cherie M. Kuzmiak
http://abstract.rsna.org/uploads/2018/18001614/US Prison System Health Care MSCB52A.pdf
MSCB52B Greek Philosophy and Cases to Ponder Personalized Screening
Participants
Athina Vourtsi, MD, Athens, Greece (Presenter) Consultant, General Electric Company; Educator, ABUS
LEARNING OBJECTIVES
1) To assess the benefits of DBT, US, and MRI in various screening and diagnostic studies. 2) Identify the applications of multi-
modality breast imaging of supplemental screening in women of average, intermediate, and high risk for developing breast cancer. 3)
Appreciate some of the Greek life style trends and health care system details with respect to breast cancer detection and clinical
management.
Active Handout:Athina Vourtsi
http://abstract.rsna.org/uploads/2018/18001615/RSNA-6. Vourtsis. Greece MSCB52B.pdf
MSCB52C False Positives and False Negatives: How to Minimize the Bunch
Participants
Elizabeth S. McDonald, MD, Philadelphia, PA (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Define false negative exam and identify common reasons for cancer misses in breast imaging and how to avoid them. 2) Define
false positive exam and discuss radiologic signs indicating that breast biopsy is not needed.
MSCB52D J-Start and Breast Density in the Land of the Rising Sun
Participants
Youichi Machida, MD, PhD, Chuo-City, Japan (Presenter) Nothing to Disclose
For information about this presentation, contact:
machida.yoichi@kameda.jp
LEARNING OBJECTIVES
1) Learn how 'dense breasts' are recognized by Japanese women and physicians, and how they will be involved in breast cancer
screening in Japan. 2) Learn the low examination rate of breast cancer screening, as well as other problems we are facing in breast
cancer care. 3) Learn the results of J-START, and the concept of 'combined assessment guideline', published by Japan Association
of Breast Cancer Screening.
Active Handout:Y ouichi Machida
http://abstract.rsna.org/uploads/2018/18001617/MSCB52D.pdf
 RC715

Tomosynthesis
Thursday, Nov. 29 4:30PM - 6:00PM Room: E353C

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Margarita L. Zuley, MD, Pittsburgh, PA (Moderator) Investigator, Hologic, Inc
Sub-Events
RC715A Use in Screening
Participants
Emily F. Conant, MD, Philadelphia, PA (Presenter) Grant, Hologic, Inc; Consultant, Hologic, Inc; Grant, iCAD, Inc; Consultant, iCAD,
Inc; Speaker, iiCME
For information about this presentation, contact:
emily.conant@uphs.upenn.edu
LEARNING OBJECTIVES
1) Review outcomes from breast cancer screening with digital breast tomosynthesis (DBT). 2) Discuss the implementation of
synthetic imaging in DBT screening. 3) Demonstrate case-based examples of pearls and pitfalls in DBT screening.
RC715B Current Trials
Participants
Valentina Iotti, MD, Reggio Emilia, Italy (Presenter) Speaker fee and travel grants from GE Healthcare.
For information about this presentation, contact:
valentina.iotti@ausl.re.it
LEARNING OBJECTIVES
1) List the current trials with digital breast tomosynthesis. 2) Compare the different study designs, interventions and setting. 3)
Examine the outcomes and potential impact on the future screening and clinical practice with tomosynthesis.
RC715C Use in Diagnostics
Participants
Margarita L. Zuley, MD, Pittsburgh, PA (Presenter) Investigator, Hologic, Inc
For information about this presentation, contact:
zuleyml@upmc.edu
 RC750

MR Imaging-guided Breast Biopsy (Hands-on)

Thursday, Nov. 29 4:30PM - 6:00PM Room: E260

BR MR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Amy L. Kerger, DO, Columbus, OH (Presenter) Nothing to Disclose
Rifat A. Wahab, DO, Cincinnati, OH (Presenter) Nothing to Disclose
Vandana M. Dialani, MD, Boston, MA (Presenter) Nothing to Disclose
Deepa Sheth, MD, Chicago, IL (Presenter) Research Grant, Guerbet SA
Lara D. Richmond, MD, Toronto, ON (Presenter) Nothing to Disclose
Gary J. Whitman, MD, Houston, TX (Presenter) Nothing to Disclose
Kirti M. Kulkarni, MD, Chicago, IL (Presenter) Nothing to Disclose
Jill J. Schieda, MD, Cleveland, OH (Presenter) Nothing to Disclose
Brandy Griffith, DO, Columbus, OH (Presenter) Nothing to Disclose
Amado B. del Rosario, DO, Tucson, AZ (Presenter) Nothing to Disclose
Karla A. Sepulveda, MD, Houston, TX (Presenter) Nothing to Disclose
Wendi A. Owen, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Laurie R. Margolies, MD, New York, NY (Presenter) Research Consultant, FUJIFILM Holdings Corporation
Mitra Noroozian, MD, Ann Arbor, MI (Presenter) Nothing to Disclose
Jeffrey R. Hawley, MD, Columbus, OH (Presenter) Nothing to Disclose
Nikki S. Ariaratnam, MD, Voorhees, NJ (Presenter) Nothing to Disclose
Su-Ju Lee, MD, Cincinnati, OH (Presenter) Spouse, Stockholder, General Electric Company; Spouse, Stockholder, Siemens AG
Mai A. Elezaby, MD, Madison, WI (Presenter) Research Grant, Exact Sciences Corporation
Anika N. Watson, MD, New York, NY (Presenter) Nothing to Disclose
Alena Levit, MD, Rochester, NY (Presenter) Nothing to Disclose
Esther N. Udoji, MD, Birmingham, AL (Presenter) Nothing to Disclose
For information about this presentation, contact:
Brandy.Griffith@osumc.edu
eudoji@uabmc.edu
alevit@ewbc.com
Jeffrey.hawley@osumc.edu
dsheth@radiology.bsd.uchicago.edu
NinaWatson@emory.edu
LEARNING OBJECTIVES
1) Explain why MR-guided breast biopsy is needed for patient care. 2) Identify relative and absolute contraindications to MR-guided
breast biopsy. 3) Describe criteria for MR-guided breast biopsy patient selection. 4) Debate risks and benefits of pre-biopsy
targeted ultrasound for suspicious MRI findings. 5) Understand the basic MR-guided biopsy procedure, protocol and requirements for
appropriate coil, needle and approach selection. 6) Manage patients before, during and after MR-guided breast biopsy. 7) Define
the benefits and limitations of MR-guided vacuum assisted breast biopsy. 8) How to problem shoot complicated cases due to lesion
location, patient anatomy, etc.
ABSTRACT
This course is intended to provide basic didactic instruction and hands-on experience for MR-guided breast biopsy. Because of the
established role of breast MRI in the evaluation of breast cancer through screening and staging, there is a proven need for MR-
guided biopsy of the abnormalities that can only be identified at MRI. This course will be devoted to the understanding and
identification of: 1) appropriate patient selection 2) optimal positioning for biopsy 3) target selection and confirmation 4) various
biopsy technologies and techniques 5) potential problems and pitfalls and 6) practice audits. Participants will spend 30 minutes in
didactic instruction followed by 60 minutes practicing MR-guided biopsy using provided phantoms. Various combinations of full size
state-of-the-art breast MRI coils, biopsy localization equipment and needles from multiple different vendors will be available for
hands-on practice. Some stations will have monitors loaded with targeting software. Expert breast imagers from around the world
will be at each of 10 stations to provide live coaching, tips, techniques and advice.
 RC752

US-guided Interventional Breast Procedures (Hands-on)

Thursday, Nov. 29 4:30PM - 6:00PM Room: E264

BR US
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Karen S. Johnson, MD, Durham, NC (Presenter) Nothing to Disclose
Jocelyn A. Rapelyea, MD, Washington, DC (Presenter) Speakers Bureau, General Electric Company; Consultant, Transmed7;
Anita K. Mehta, MD, MSc, Washington DC, DC (Presenter) Nothing to Disclose
Kathleen R. Gundry, MD, Atlanta, GA (Presenter) Nothing to Disclose
Michael N. Linver, MD, Albuquerque, NM (Presenter) Medical Advisory Board, Solis; Scientific Advisory Board, Real Imaging Ltd;
Scientific Advisory Board, Seno Medical Instruments, Inc
Tilden L. Childs III, MD, Fort Worth, TX (Presenter) Nothing to Disclose
Evguenia J. Karimova, MD, Memphis, TN (Presenter) Nothing to Disclose
Caroline M. Ling, MD, Darby, PA (Presenter) Nothing to Disclose
Sora C. Yoon, MD, Durham, NC (Presenter) Nothing to Disclose
Connie E. Kim, MD, Durham, NC (Presenter) Spouse, Consultant, ClarVista Medical, Inc Spouse, Royalties, Leica Biosystems
Nussloch GmbH Spouse, Intellectual property, Leica Biosystems Nussloch GmbH
Mary S. Soo, MD, Durham, NC (Presenter) Nothing to Disclose
Christina G. Marks, MD, Saint Louis, MO (Presenter) Nothing to Disclose
Margaret M. Szabunio, MD, Lexington, KY (Presenter) Nothing to Disclose
Jean M. Kunjummen, DO, Atlanta, GA (Presenter) Nothing to Disclose
For information about this presentation, contact:
mammomike@aol.com
cmarks@umc.edu
mary.soo@duke.edu
jrapelyea@mfa.gwu.edu
margaret.szabunio@uky.edu
LEARNING OBJECTIVES
1) Describe the equipment needed for ultrasound guided interventional breast procedures. 2) Review the basic principles of
ultrasound guidance and performance of minimally invasive breast procedures. 3) Practice hands-on technique for ultrasound guided
breast interventional procedures.
ABSTRACT
This course is intended to familiarize the participant with equipment and techniques in the application of US guided breast biopsy
and needle localization. Participants will have both basic didactic instruction and hands-on opportunity to practice biopsy
techniques on tissue models with sonographic guidance. The course will focus on the understanding and identification of: 1) optimal
positioning for biopsy 2) imaging of adequate sampling confirmation 3) various biopsy technologies and techniques 4) potential
problems and pitfalls
 RC815

Radiologist's Value in the Multidisciplinary Team

Friday, Nov. 30 8:30AM - 10:00AM Room: E350

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

Participants
Cherie M. Kuzmiak, DO, Chapel Hill, NC (Moderator) Research Grant, Delphinus Medical Technologies, Inc
LEARNING OBJECTIVES
1) Define high-risk lesions. 2) Review radiologic-pathologic features and characteristics of high-risk lesions. 3) Discuss approaches
to high-risk lesions with attention to management considerations and controversies. 4) Describe and discuss the imaging features of
the major subtypes of invasive breast cancer with integration of pathology. 5) Understand the molecular classification of breast
cancer and its impact on patient prognosis. 6) Describe patients' perspectives of breast imaging experiences with a focus on pain,
anxiety, and emotional distress relating to abnormal results and biopsy procedures. 7) Define interventions for reducing patients'
negative experiences related to abnormal results and biopsy procedures.
Sub-Events
RC815A Management of High Risk Lesions
Participants
Samantha L. Heller, MD, PhD, New York, NY (Presenter) Nothing to Disclose
For information about this presentation, contact:
Samantha.Heller@nyumc.org
LEARNING OBJECTIVES
1) Define high-risk lesions. 2) Review radiologic-pathologic features and characteristics of high-risk lesions. 3) Discuss approaches
to high-risk lesions with attention to management considerations and controversies.
RC815B Imaging Appearance of Cancer Subtypes
Participants
Cherie M. Kuzmiak, DO, Chapel Hill, NC (Presenter) Research Grant, Delphinus Medical Technologies, Inc
LEARNING OBJECTIVES

1) Describe and discuss the imaging features of the major subtypes of invasive breast cancer with integration of pathology. 2)
Understand the molecular classification of breast cancer and its impact on patient prognosis.
RC815C Improving the Patient's Experience
Participants
Mary S. Soo, MD, Durham, NC (Presenter) Nothing to Disclose
For information about this presentation, contact:
mary.soo@duke.edu
LEARNING OBJECTIVES
1) Describe patients' perspectives of breast imaging experiences with a focus on pain, anxiety, and emotional distress relating to
abnormal results and biopsy procedures. 2) Define interventions for reducing patients' negative experiences related to abnormal
results and biopsy procedures.
 RC852

Breast Elastography (Hands-on)

Friday, Nov. 30 8:30AM - 10:00AM Room: E264

BR US
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 0

Participants
Richard G. Barr, MD, PhD, Campbell, OH (Presenter) Consultant, Siemens AG; Consultant, Koninklijke Philips NV; Research Grant,
Siemens AG; Research Grant, SuperSonic Imagine; Speakers Bureau, Koninklijke Philips NV; Research Grant, Bracco Group; Speakers
Bureau, Siemens AG; Consultant, Canon Medical Systems Corporation; Research Grant, Esaote SpA; Research Grant, BK Ultrasound;
Research Grant, Hitachi, Ltd
Stamatia V. Destounis, MD, Scottsville, NY (Presenter) Research Grant, Hologic, Inc; Research Grant, Delphinus Medical
Technologies, Inc
Rajas N. Chaubal, MBBS,MD, Thane, India (Presenter) Nothing to Disclose
Nitin G. Chaubal, MD, MBBS, Mumbai, India (Presenter) Nothing to Disclose
Chander Lulla, MBBS, Mumbai, India (Presenter) Nothing to Disclose
Vito Cantisani, MD, Rome, Italy (Presenter) Speaker, Canon Medical Systems Corporation; Speaker, Bracco Group; Speaker,
Samsung Electronics Co, Ltd;
Maija Radzina, MD, PhD, Riga, Latvia (Presenter) Nothing to Disclose
Phan T. Huynh, MD, Houston, TX (Presenter) Nothing to Disclose
Paula B. Gordon, MD, Vancouver, BC (Presenter) Stockholder, OncoGenex Pharmaceuticals, Inc ; Stockholder, Volpara Health
Technologies Limited; Scientific Advisory Board, Real Imaging Ltd; Scientific Advisory Board, DenseBreast-info, Inc;
Tanya W. Moseley, MD, Houston, TX (Presenter) Nothing to Disclose
Catherine W. Piccoli, MD, Voorhees, NJ (Presenter) Stockholder, Qualgenix LLC;
Gary J. Whitman, MD, Houston, TX (Presenter) Nothing to Disclose
Anna I. Holbrook, MD, Atlanta, GA (Presenter) Nothing to Disclose
Rachna Dutta, MD, Cleveland, OH (Presenter) Nothing to Disclose
Valerio Forte, MD, Rome, Italy (Presenter) Nothing to Disclose
Daniele Fresilli, Roma, Italy (Presenter) Nothing to Disclose
Giuseppe Schillizzi, Roma, Italy (Presenter) Nothing to Disclose
Gregorio Alagna, Rome, Italy (Presenter) Nothing to Disclose
Valeria de Soccio, JD, Rome, Italy (Presenter) Nothing to Disclose
For information about this presentation, contact:
riaclinic@gmail.com
nitin.chaubal@gmail.com
rajas.chaubal@gmail.com
sdestounis@ewbc.com
rdutta@metrohealth.org
LEARNING OBJECTIVES
1) To explain the difference between strain and shear wave elastography. 2) To review how to characterize breast lesions as
benign or malignant on elastography. 3) To demonstrate how to perform both strain and shear wave elastography for breast
imaging.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Richard G. Barr, MD, PhD - 2017 Honored Educator
 SST01

Breast Imaging (Multimodality Breast Imaging)

Friday, Nov. 30 10:30AM - 12:00PM Room: E353B

BR
AMA PRA Category 1 Credits ™: 1.50
ARRT Category A+ Credit: 1.75

FDA Discussions may include off-label uses.

Participants
Colleen H. Neal, MD, Ann Arbor, MI (Moderator) Nothing to Disclose
Debra S. Copit, MD, Wynnewood, PA (Moderator) Research funded, Hologic, Inc
Sub-Events
SST01-01 Comparative the Average Glandular Dose between Digital Breast Tomosynthesis (DBT) and Full-Field
Digital Mammography (FFDM): Correlation with Breast Thickness and Density
Friday, Nov. 30 10:30AM - 10:40AM Room: E353B

Participants
Chanjuan Wen, Guangzhou, China (Presenter) Nothing to Disclose
Weimin Xu, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Hui Zeng, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Zilong He, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
Weiguo Chen, Guangzhou, China (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
182389433@qq.com
PURPOSE
To compare the average glandular dose (AGD) between single-view digital breast tomosynthesis (DBT) and single-view full-field
digital mammography (FFDM), and to evaluate the correlation of AGD with breast thickness and density.
METHOD AND MATERIALS
A total of 318 female patients who underwent both DBT and FFDM (DBT and FFDM were performed in the same compression
thickness in each breast) were included. 636 DBT images of unilateral breast mediolateral oblique (MLO) view and 636 FFDM images
of unilateral breast mediolateral oblique (MLO) view were analyzed. Mammographic breast density was determined according to BI-
RADS breast density grading, and breast thickness and AGD per exposure in MLO views retrieved from DICOM headers were
recorded. Breast thickness were divided into the following four groups: <= 30cm, 31 ~ 45cm, 46 ~ 60cm and > 60cm. The
statistical analyses used variance analysis and Pearson's correlation for parametric tests.
RESULTS
(1) The AGD of DBT had a weak negative correlation with breast density (correlation coefficient =-0.305, P<0.001), decreased as
the breast density increased. The AGD of FFDM did not change significantly with breast density increased (correlation coefficient
=-0.027, P=0.501). (2) Breast thickness was significantly associated with AGDs, and both AGDs of FFDM and DBT increased with
increased breast thickness (correlation coefficient =0.771 and 0.935, respectively, all P<0.001). (3) When breast density was >75%
and breast thickness was >60cm, the AGD of DBT was lower than that of FFDM, and the difference was statistical significant (P =
0.031).
CONCLUSION
The AGD of DBT increased with breast thickness increased and decreased with breast density. For thick and dense breast, the
radiation dose of DBT was lower than that of FFDM.
CLINICAL RELEVANCE/APPLICATION
In this study, we evaluated the AGD of MLO FFDM and DBT according to breast density and thickness.
SST01-02 Accuracy of Molecular Breast Imaging in Patients with Suspicious Calcifications
Friday, Nov. 30 10:40AM - 10:50AM Room: E353B

Participants
Carrie B. Hruska, PhD, Rochester, MN (Abstract Co-Author) Institutional license agreement, CMR Naviscan Corporation
Katie N. Hunt, MD, Rochester, MN (Presenter) Nothing to Disclose
Matthew Johnson, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Amy Lynn Conners, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Michael K. O'Connor, PhD, Rochester, MN (Abstract Co-Author) Royalties, Gamma Medica, Inc
Deborah J. Rhodes, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
Dietlind Wahner-Roedler, MD, Rochester, MN (Abstract Co-Author) Nothing to Disclose
PURPOSE
Molecular breast imaging (MBI), which shows uptake of Tc-99m sestamibi in metabolically-active tissue, has been proposed as a
tool for determining whether biopsy of mammographically-detected lesions is necessary. Here, our goal was to evaluate the
diagnostic accuracy of MBI in patients with suspicious calcifications on mammography.
METHOD AND MATERIALS
Women scheduled to undergo stereotactic biopsy of calcifications detected on 2D mammography were prospectively enrolled to
undergo MBI prior to biopsy. MBI was performed with injection of Tc-99m sestamibi and a dedicated gamma camera. A breast
radiologist interpreted MBI in conjunction with mammography.
RESULTS
In 71 women studied, 76 discrete areas of calcifications were identified for biopsy, of which pre-biopsy MBI was positive in 17/76
(22%). Of 76 calcification lesions, 24 (32%) were malignant, including 20 DCIS and 4 invasive ductal cancer; MBI was positive in
10/20 (50%) DCIS and 2/4 (50%) invasive cancers. In 21 cancers with calcification morphology of amorphous, coarse
heterogeneous, or fine pleomorphic (BI-RADS 4B), MBI was positive in 12/21 (57%), while in three cancers with fine linear or fine
linear branching calcifications (BI-RADS 4C), MBI was negative in all 3 (p=0.06). Calcification distribution was more varied for the
MBI-positive cancers (0 regional, 7 grouped, 1 linear, 4 segmental) than for the MBI-negative cancers (1 regional, 10 grouped, 1
linear, 0 segmental) (p=0.14). The median pathologic size for MBI-positive cancers was 1.5 cm (range=0.5-3.2 cm) compared to
0.9 cm (range=0.1-2.0 cm) for MBI-negative cancers (p=0.09). Beyond calcification lesions, detection of non-mass focal areas of
uptake on MBI led to additional biopsies of 6 sites, of which 2 were malignant (DCIS). The overall positive and negative predictive
values of MBI were 61% (14/23) and 81% (48/59), respectively.
CONCLUSION

MBI has insufficient negative predictive value to be used for identifying calcifications in which biopsy could be avoided. However,
MBI can reveal additional sites of mammographically-occult disease.
CLINICAL RELEVANCE/APPLICATION
Negative findings on MBI should not be used to avoid biopsy of suspicious calcifications on mammography.
SST01-03 Staging Early Breast Cancer with Simultaneous PET/ MRI: Impact on Management
Friday, Nov. 30 10:50AM - 11:00AM Room: E353B

Participants
Sangeeta Taneja, MD, New Delhi, India (Presenter) Nothing to Disclose
Ramesh Sarin, MS, New Delhi, India (Abstract Co-Author) Nothing to Disclose
Amarnath Jena, MD, New Delhi, India (Abstract Co-Author) Nothing to Disclose
Aru Singh, PhD, New Delhi, India (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

s_tanejaa@yahoo.com
PURPOSE

Breast cancer is a biologically heterogenous disease with certain clinical subtypes having a greater propensity to develop
metastasis even at an early stage. In its present state, NCCN recommends 18-F FDG PET CT in patients above clinical stage III
disease. There have been reports on PET CT in altering clinical stage in early breast cancer. DCE (Dynamic Contrast-Enhanced) MRI
has been shown to detect additional disease in the ipsilateral & contralateral breast. Simultaneous PET/MRI combines 18F-FDG PET
& MRI of the whole body & DCE-MRI of the breast in a single examination. This retrospective study evaluates the impact of
simultaneous 18F-FDG PET/MRI in pretreatment staging of early breast cancer (Stage I-IIIA).
METHOD AND MATERIALS
The study was approved by institutional ethics committee. 101 patients with histologically proven breast cancer (clinical stage I,
IIA, IIB, IIIA) who underwent simultaneous PET/MRI (including DCE MRI breast) were included. Breast lesions, nodes & metastases
were evaluated on PET, MRI & PET-MRI for lesion count & diagnostic confidence (DC).
RESULTS

101 index breast lesions were identified on MRI, PET/MRI (Mean DC 4.96) & 99 on PET (Mean DC >=4). MRI detected multifocality in
15 (14.8%), multicentricity in 10 (9.90%) & contralateral unsuspected cancer in 2 patients. PET detected axillary nodal metastases
in 12/18 (DC>=4), MRI in 15/18 (DC>=4) and PET/MRI in 15/18 patients. Distant metastases were found in 18 /101 (18 %) on PET
(Mean DC score 4.1), MRI (Mean DC score=3.7) & PET/MRI (Mean DC score =4.7) with bone (n =11), lymph nodes (mediastinal;
n=2), liver (n=4), brain (n=1) & lung (n=5). The mean metastatic lesion size on MRI was 1.52 ± 0.25 cm (Range: 0.5-5.7cm). PET
MRI changed the overall stage in 66 patients (65%, upstaging: 62, downstaging:4) & overall change in management in 29 % of
early stage breast cancer patients including 18 patients who were upstaged to stage IV.
CONCLUSION

Simultaneous18F-FDG PET/MRI has the potential to impact the initial staging in early breast cancer for an overall improved patient
management.
CLINICAL RELEVANCE/APPLICATION

Simultaneous PET/MRI has the potential to alter the stage and hence the clinical management in patients with early breast cancer
thus reducing the morbidity and cost due to inappropriate therapies.
SST01-04 The Gambler's Fallacy in Screening Mammography
 Friday, Nov. 30 11:00AM - 11:10AM Room: E353B

Participants
Andrew L. Callen, MD, San Francisco, CA (Presenter) Nothing to Disclose
Omar Mesina, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Sivan G. Marcus, BS, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Iryna Lobach, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Bonnie N. Joe, MD, PhD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Edward A. Sickles, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
Heather I. Greenwood, MD, San Francisco, CA (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
andrew.callen@ucsf.edu
PURPOSE
It has been documented in multiple settings that the sequence of decisions already made affects future decision-making. One
component of this phenomenon is known as the gambler's fallacy: the tendency to underestimate the likelihood of 'streaks'
(successive identical decisions) occurring by chance. The purpose of this study was to determine if the gambler's fallacy affects
radiologists interpreting screening mammography.
METHOD AND MATERIALS

This was a retrospective, HIPPA compliant IRB approved study. Patients who underwent routine screening mammography in 2014 at
our institution were included, with 8,543 total exams, which was sufficient to detect a 1% change in recall rate based on an 80%
power calculation with a two-sided 0.05 significance level, for our recall rate of 7%. Data were collected from log books containing
the BI-RADS assessments of routine screening mammograms, the order in which the examinations were interpreted, and the number
of preceding BI-RADS 0 and BI-RADS 1 or 2 assessments for each exam. If recalled (BI-RADS-0), subsequent diagnostic exam BI-
RADS assessment also was recorded. Analysis was performed using Fishers exact test to evaluate whether an increasing number of
preceding decisions to not recall (BI-RADS-1 or 2) resulted in an increased number of recalls that did not lead to a cancer
diagnosis. False positive was defined as a BI-RADS 0 assessment at screening, followed by a BI-RADS 1, 2 or 3 at diagnostic breast
imaging. A true positive was defined as a BI-RADS 0 assessment at screening, followed by a BI-RADS 4 or 5 assessment at
diagnostic breast imaging.
RESULTS
Data on 8,543 routine screening exams was collected for the year 2014. An average of 20.9 exams were batch read in each
session. 700 exams (8%) were assessed as BI-RADS 0. Of those, 231 (33%) were assigned either BIRADS-4 or 5 at the time of
diagnostic imaging. True and false positives were compared, stratified by the number of preceding BI-RADS 1 or 2 assessments in
that batch-read session. Exams with a higher number of preceding negative assessments did not have a higher false positive rate.
CONCLUSION

At our academic institution, we did not observe a statistically significant effect of the gambler's fallacy in one year's worth of
screening mammography.
CLINICAL RELEVANCE/APPLICATION

We did not detect an effect of the gambler's fallacy in one year's worth of screening mammography at an academic institution.
SST01-05 Outcomes of Ductal Carcinoma in Situ According to Detection Modality: A Multicenter Study
Comparing Recurrences Between Mammography and Breast US
Friday, Nov. 30 11:10AM - 11:20AM Room: E353B

Participants
Jung Hyun Yoon, MD, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Kyunghwa Han, PhD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jieun Koh, MD, Seongnam, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ga Ram Kim, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hye Jung Kim, Daegu, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Young Mi Park, MD, PhD, Busan, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Ji Hyun Youk, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Jin Chung, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
In Hye Chae, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Eun Jung Choi, MD, PhD, Jeonju, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Hee Jung Moon, MD, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

lvjenny@yuhs.ac
PURPOSE
To determine whether if disease recurrence and intrinsic characteristics of ductal carcinoma in situ (DCIS) is associated to the
imaging method of detection in asymptomatic women who are diagnosed with DCIS.
METHOD AND MATERIALS
This retrospective, multicenter study was conducted at 8 institutions including 844 women who were treated for asymptomatic,
pure DCIS who had preoperative mammography and breast ultrasonography (US) available for review. Mean follow up interval after
treatment of the 844 women was 91.2 months (standard deviation: 533.3, range: 6.4-180.9 months). Medical records and breast
images were reviewed by 8 breast-imaging dedicated radiologists for clinicopathologic information and image analysis. Kaplan-Meier
analysis and univariable/multivariable Cox proportion hazard model was used to analyze the recurrence-free survival rates and
factors associated with recurrence after DCIS treatment.
RESULTS

Of the 844 women who were treated for DCIS, 25 (3.0%) had developed recurrences. Patients with US-detected DCIS had
significantly lower 5- and 10-year recurrence-free survival rates compared to patients with mammography-detected ones
(P=0.011). US-detected DCIS had significantly lower 5- and 10-year recurrence-free survival rates compared to mammography-
detected ones in patients <50years or with mammographically-dense breasts (P=0.002, and 0.002, respectively). Multivariable
analysis showed that US for detection modality (HR: 4.451, 95% CI: 1.530, 12.950, P=0.006) and HER2 positivity (HR: 4.036, 95%
CI: 1.438, 11.330, P=0.008) showed significant association to recurrences.
CONCLUSION

US for detection modality and HER2 positivity were factors significantly associated to recurrences in patients treated for
asymptomatic DCIS.
CLINICAL RELEVANCE/APPLICATION

Supplementary screening US may enable detection of clinically important DCIS, especially in younger women or mammographically-
dense breasts in which mammography has suboptimal performances in detection of DCIS or small invasive cancers.
SST01-06 Do Triple Negative Breast Cancers Have Characteristic Imaging Features According to Androgen
Receptor and Vimentin Status?
Friday, Nov. 30 11:20AM - 11:30AM Room: E353B

Participants
Rosalind P. Candelaria, MD, Houston, TX (Presenter) Nothing to Disclose
Beatriz E. Adrada, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Lumarie Santiago, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Deanna L. Lane, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Wei Wei, Houston, TX (Abstract Co-Author) Nothing to Disclose
Wei T. Yang, MD, Houston, TX (Abstract Co-Author) Consultant, General Electric Company; Medical Advisory Board, Seno Medical
Instruments, Inc
Monica L. Huang, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Elsa M. Arribas, MD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Gaiane M. Rauch, MD, PhD, Houston, TX (Abstract Co-Author) Nothing to Disclose
Alastair Thompson, Houston, TX (Abstract Co-Author) Nothing to Disclose
Stacy Moulder, MD, Houston, TX (Abstract Co-Author) Research funded, AstraZeneca PLC; Research funded, F. Hoffmann-La Roche
Ltd; Research funded, Oncothyreon; Research funded, Novartis AG; Research funded, Merck KGaA
For information about this presentation, contact:

rcandelaria@mdanderson.org
PURPOSE

Lehmann et al (Journal of Clinical Investigation, 2011) previously identified six molecular subtypes of triple negative breast cancer
(TNBC) through analysis of gene expression profiles; the luminal androgen receptor (LAR) subtype has been shown to have a higher
percentage of regional spread to lymph nodes and the mesenchymal (M) subtype, a lower percentage. The purpose of this study is
to determine if TNBCs have characteristic imaging features based on androgen receptor (AR) and vimentin (VM) status, which are
surrogate immunohistochemical markers for the LAR and M subtypes of TNBC, respectively.
METHOD AND MATERIALS

This study is part of a clinical trial of stage I-III TNBC patients, which is being conducted at a single quaternary care center. A
total of 144 patients, who were randomized to the intervention arm of being informed of the results of their molecular
characterization including androgen receptor and vimentin status prior to receiving neoadjuvant chemotherapy, were included in
this interim imaging analysis. Androgen-receptor-positive tumors (AR+) were defined as having >=15% staining. Vimentin-positive
(VM+) tumors were defined as having >=50% staining. Two experienced, fellowship-trained breast radiologists used BIRADS (Breast
Imaging Reporting and Data System) lexicon to review and reach consensus on all imaging studies (i.e., mammogram, ultrasound,
and breast magnetic resonance imaging) while blinded to the immunohistochemical results. Fisher's exact test was used to assess
the association of AR or VM status with imaging features. P values less than 0.05 was considered statistically significant.
RESULTS

Androgen-receptor-positive TNBC was significantly associated with scattered and heterogeneous breast composition on
mammography (p=0.04), presenting as a mass with calcifications on mammography (p=0.04), having an irregular shape on
ultrasound (p=0.005), and having an irregular margin on MRI (p=0.04). However, vimentin expression in TNBC was not significantly
associated with any specific imaging features.
CONCLUSION
TNBCs have characteristic imaging features based on androgen receptor status but not based on vimentin status.
CLINICAL RELEVANCE/APPLICATION

Multimodality breast imaging may help identify LAR TNBC, which has been shown to be a subtype with a higher rate of regional
nodal disease and with decreased response to neoadjuvant chemotherapy.
SST01-07 Usefulness of Surveillance MR for Early and Late Recurrent Breast Cancer in Women after Breast-
Conservation Therapy
Friday, Nov. 30 11:30AM - 11:40AM Room: E353B

Participants
Jeong Min Lee, Seoul, Korea, Republic Of (Presenter) Nothing to Disclose
Bong Joo Kang, Seoul, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
Sung Hun Kim, Su Won, Korea, Republic Of (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

jmlee328@gmail.com
PURPOSE

To investigate the diagnostic performance of mammography, ultrasonography, and breast magnetic resonance imaging (MRI) for
early and late recurrences in patients who underwent breast-conservation therapy (BCT) for breast cancer.
METHOD AND MATERIALS

This retrospective study was approved by our institutional review board. Between January 2014 and February 2018, 1312 women
with 1951 surveillance breast MR examinations after BCT were studied. We assessed the cancer detection rate of each surveillance
MR, mammography and ultrasound.
RESULTS
Of 1951 cases of surveillance postoperative MRI, 59 cases were confirmed as cancer recurrence through biopsy. Nineteen cases of
recurrences within 12 months post-BCT were defined as early recurrences while other 40 cases of recurring after 13 months post-
BCT were defined as late recurrences. There were no statistically significant differences in patients' demographics between two
groups with p > 0.05; age at cancer diagnosis, age at recurrence, symptoms, laterality of recurred cancer and intense surveillance.
Among 19 patients with early recurrence, 7 cases were detected on mammography (36.8%), 10 on ultrasound (52.6%), and 17 on
MRI (89.5%). Of 40 patients with late recurrence, 24 were detected on mammography (60%), 29 on ultrasound (72.5%) and 39 on
MRI (97.5%). In both groups, MRI showed significantly higher cancer detection rate than mammography or ultrasound (p < 0.01).
CONCLUSION
In breast cancer patients with BCT, regardless of early or late, postoperative MR surveillance showed a significantly higher
detection rate for cancer recurrence than mammography or ultrasound.
CLINICAL RELEVANCE/APPLICATION
Postoperative surveillance MR is useful tool for screening early or late cancer recurrence in breast cancer patients with breast
conserving therapy.
SST01-08 The Role of Digital Breast Tomosynthesis (DBT) versrus Automated Breast Ultrasound (ABUS) in the
Detection and Characterization of the Different Breast Lesions
Friday, Nov. 30 11:40AM - 11:50AM Room: E353B

Participants
Maha H. Helal IV, MD, Cairo, Egypt (Presenter) Nothing to Disclose
Sahar Mansour, MD, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Lamia Bassam, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Reham Hussein, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
Emad Elgemeie, Cairo, Egypt (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:

dr.mahahelal@yahoo.com
PURPOSE
DBT and ABUS are advanced applications of digital mammography and breast ultrasound respectively. We aimed to evaluate the role
of DBT versus ABUS in the detection and characterization of breast lesions
METHOD AND MATERIALS

Institutional review board approval was obtained for this prospective study that included 80 patients with 87 breast lesions.
Methods of evaluation were digital breast tomosynthesis and automated breast ultrasound. For mammogram acquisition the system
acquires a traditional digital mammogram and a tomosynthesis scanning in the same compression in the MLO and CC views. 3D ABUS
were done for anteroposterior; lateral and medial acquisitions. Included breast lesions were analyzed regarding size, shape, margin,
extension, calcifications and multiplicity. Operative data was the gold standard reference
RESULTS
ABUS showed more accurate measurements of the size of the breast lesions as DBT overestimated 18.8% of masses, on the other
hand ABUS over estimated 25% of masses. No size under estimation by both modalities. ABUS was superior to DBT in estimation of
the shape of the lesions 87% versus 69.6% for DBT, but both displayed similar values in the evaluation of the margins
(65.2%).Tomosynthesis was far better in the detection of calcification in 40 lesions; while automated ultrasound was able to
detect calcifications in only 11 lesions of them. Multiplicity was better demonstrated by ABUS that showed an accuracy of 100%
compared to 80% by DBT. We found out that the sensitivity of tomosynthesis in detection and characterization of breast masses
was 100 %, the specificity was 81.25%, the positive predictive value was 87.5% and the negative predictive value was 100%. On
the other hand the sensitivity of automated ultrasound was 100%, the specificity was 75%, the positive predictive value was 84%
and the negative predictive value was 100%.
CONCLUSION
DBT and ABUS, both showed near estimation in the detection and characterization of breast lesions. DBT is the modality for
calcifications and ABUS is more accurate in the detection of multiplicity.
CLINICAL RELEVANCE/APPLICATION

DBT is considered as an adjunct to digital mammogram to increase the conspicuity of the different breast lesions. ABUS is a
revolution in the ultrasound scanning of the breast that can be used as a non-invasive, fast and easy tool of breast imaging in
early detection (i.e. screening) and differentiation of breast lesions.
SST01-09 First Description of Molecular Imaging Heterogeneity Profiles for Breast Tumors and Its Clinical Utility
Friday, Nov. 30 11:50AM - 12:00PM Room: E353B

Participants
MIchel Herranz, Santiago de Compostela, Spain (Presenter) Nothing to Disclose
Lucia Grana Lopez, MD, Lugo, Spain (Abstract Co-Author) Nothing to Disclose
Ines Dominguez, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Sonia Argibay, MD, PhD, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Manuel Vazquez-Caruncho, MD, Lugo, Spain (Abstract Co-Author) Nothing to Disclose
Roberto Garcia Figueiras, MD, PhD, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
Alvaro Ruibal, MD, PhD, Santiago de Compostela, Spain (Abstract Co-Author) Nothing to Disclose
For information about this presentation, contact:
michel.herranz.carnero@sergas.es
PURPOSE
The concept of tumor heterogeneity, also called in Radiology as Tumor Texture, is based on the different areas of tumor uptake,
which correspond to different levels of expression, cellularity, hypoxia or other parameters interested in being measured. We want
to know if the description of the tumor heterogeneity uses in Radiology has its relation with PET parameters and if any biological
characteristics of the breast tumors have a structure-function correlation.
METHOD AND MATERIALS

We have analyzed 1000 consecutive patients with breast cancer in a dedicated breast PET (dbPET). Different parameters have
been defined that allow us to find a pattern of Texture and Heterogeneity (TeHe), for this, and following the rules of the
radiological descriptions we defined a series of structural templates that cover practically all tumors, a mathematical formula has
been defined for this correlation, and a assisted software for tumor shape description has been used to perform 3D categories.
RESULTS

7 different patterns divided into 5 groups for TeHe are described, and classified as: 1: Homogeneous-diffuse, 2: Lobular, 3: Annular
and Spindle, 4: Eccentric and Focused; and 5: Speckled. A numerical value has been assigned between 1 and 5 for this
classification with 1 being the most homogeneous and 5 being the most heterogeneous. This value is achieved through a
mathematical relationship: medSUV/maxSUV: values close to 1 denote a high homogeneity and those close to 0 indicate a high
heterogeneity. Process is complicated when tumor geometry becomes part of this heterogeneity. In those cases, some geometric
patterns may explain similar values. We have analyzed the clinical utility of this classification and we have found two major uses: i)
in the description of the efficiency of neoadjuvant therapy, where changes in TeHe pattern define responders of non-responders
and ii) we have found, for the FIRST TIME, a correlation between TeHe patterns and the molecular subtype, crucial fact in the
future of imaging based breast cancer diagnosis.
CONCLUSION

Studies of tumor heterogeneity based on metabolism show us different patterns that correlate with molecular subtypes and predict
response to treatments.
CLINICAL RELEVANCE/APPLICATION
Tumor Texture and Heterogeneity are becoming, like in conventional radiology, in a new tool for prediction of response to the
treatment and in molecular subtype characterization.
 SPFR61

Friday Imaging Symposium: Screening with Imaging in 2018: Who Benefits?

Friday, Nov. 30 12:30PM - 3:00PM Room: E350

BR CH GI GU OI
AMA PRA Category 1 Credits ™: 2.50
ARRT Category A+ Credits: 3.00

FDA Discussions may include off-label uses.

Participants
Hebert Alberto Vargas, MD, Cambridge, United Kingdom (Moderator) Nothing to Disclose
Dow-Mu Koh, MD,FRCR, Sutton, United Kingdom (Moderator) Nothing to Disclose
Sub-Events
SPFR61A Breast Cancer Screening: Lessons Learned from Where it all Started
Participants
Victoria L. Mango, MD, New York, NY (Presenter) Nothing to Disclose
For information about this presentation, contact:
mangov@mskcc.org
LEARNING OBJECTIVES
1) Discuss early mammography screening trials for breast cancer. 2) Analayze recent multi-modality breast cancer screening
literature. 3) Apply lessons learned from breast cancer screening to future screening for other diseases.
ABSTRACT
N/A
URL
N/A
SPFR61B Emerging CRC Screening Options
Participants
Perry J. Pickhardt, MD, Madison, WI (Presenter) Stockholder, SHINE Medical Technologies, Inc; Stockholder, Elucent Medical;
Advisor, Bracco Group;
LEARNING OBJECTIVES
1) To Understand the various CRC screening options, with emphasis on newer emerging strategies.
ABSTRACT
N/A
URL
N/A
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Perry J. Pickhardt, MD - 2014 Honored EducatorPerry J. Pickhardt, MD - 2018
Honored Educator
SPFR61C Liver Cancer Screening: Who Benefits?
Participants
Bachir Taouli, MD, New York, NY (Presenter) Research Grant, Guerbet SA; Research Grant, Bayer AG
For information about this presentation, contact:
bachir.taouli@mountsinai.org
LEARNING OBJECTIVES
1) Review current guidelines for liver cancer screening, including target population and methods used. 2) Review the limitations of
blood markers and ultrasound for liver cancer screening. 3) Review new methods such as abbreviated MRI for liver cancer
screening.
ABSTRACT
Hepatocellular carcinoma (HCC) is the 2nd leading cause of cancer-related death worldwide, and the fastest growing cause of
cancer death in the USA. The most important risk factor for HCC is cirrhosis. In this presentation, we will discuss the rationale of
HCC screening, the most recent AASLD guidelines for HCC screening and surveillance using ultrasound (US) with or without alpha-
fetoprotein (AFP). We will review the current results and limitations of this strategy. We will also review recent developments in the
use of abbreviated MRI protocols for HCC screening and surveillance.
SPFR61D Lung Cancer: Should We Be Screening Patients with Other Cancers?
Participants
Michelle S. Ginsberg, MD, New York, NY (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Review the approach to the inclusion of patients with a previous history of malignancy in lung cancer screening studies. 2) To
discuss the need for lung cancer screening in survivors of other cancers.
SPFR61E Prostate Cancer Screening: Will it Ever Happen?
Participants
Harriet C. Thoeny, MD, Bern, Switzerland (Presenter) Advisory Board, Guerbet SA
LEARNING OBJECTIVES
1) To identify the disadvantages of the current gold standard of prostate cancer detection. 2) To differentiate significant form
insignificant PCa and to understand its impact on management. 3) To assess the prerequisites of mpMRI as a screening tool of
prostate cancer detction.
ABSTRACT
Prostate cancer (PCa) is the most frequent malignant tumor in men in Europe and the USA. Up to date systematic transrectal
ultrasound guided- (TRUS) biopsy based on a rise in PSA and/or a suspicious digital rectal examinaition is the gold standard in PCa
detection. However, this approach is unsatisfactory as it leads to over-and underdiagnosis of PCa. MpMRI is now an integrated part
in the workup of PCa in many institutions and MR/TRUS-fuion guided instead of systematic blind biospies are more frequently used
leading to a higher detection rate of significant PCa on one hand and a lower detection rate of insignificant PCa on the other hand.
The NPV of mpMRI to detect significant PCa is reported between 63-98% depening on patient selection. mpMRI improves PCa
detection and might therefore be a valuable tool for PCa screening however, the prerequisities include excellent image quality, a
dedicated and experienced radiologist, availabilty of MRI and a short imaging protocol without contrast medium administration to
make the healthcare authoritites considering mpMRI as a cost effective screening tool. Furthermore, an improved NPV might reduce
the number of unnecessary biopsies in a high number of men and therefore decrease costs for the healthcare system.
SPFR61F Ovarian Cancer Screening: Have We Given Up Yet?
Participants
Andrea G. Rockall, FRCR,MRCP, London, United Kingdom (Presenter) Speaker, Guerbet SA
LEARNING OBJECTIVES

1) To know about the results of ovarian cancer screening studies. 2) To understand the possible reasons for failure. 3) To be aware
of screening studies in high risk patients.
Honored Educators
Presenters or authors on this event have been recognized as RSNA Honored Educators for participating in multiple qualifying
educational activities. Honored Educators are invested in furthering the profession of radiology by delivering high-quality
educational content in their field of study. Learn how you can become an honored educator by visiting the website at:
https://www.rsna.org/Honored-Educator-Award/ Andrea G. Rockall, FRCR,MRCP - 2017 Honored Educator
SPFR61G Non-cancer Screening: Should We Screen for Cardiovascular Diseases with Imaging?
Participants
Mathias Prokop, PhD, Nijmegen, Netherlands (Presenter) Speakers Bureau, Bracco Group; Speakers Bureau, Bayer AG; Research
Grant, Canon Medical Systems Corporation; Speakers Bureau, Canon Medical Systems Corporation; Research Grant, Siemens AG ;
Speakers Bureau, Siemens AG ; Departmental spinoff, Thirona; Departmental licence agreement, Varian Medical Systems, Inc; ;
SPFR61H Whole-body Screening for Multiple Cancers: Is a One-Stop-Shop Approach Feasible?
Participants
Giuseppe Petralia, MD, Milan, Italy (Presenter) Nothing to Disclose
LEARNING OBJECTIVES
1) Identify the most appropriate imaging technique for whole-body cancer screening. 2) Arrange a whole-body MRI scanning
protocol in their home Institutions. 3) Describe findings observed in a whole-body MRI performed for cancer screening in a Likert
scale. 4) Recommend the whole-body MRI for cancer screening to the appropriate population.