This copy is for personal use only. To order printed copies, contact reprints@rsna.

org
833

Imaging Physics
Pictorial Review of Digital Radiog-
raphy Artifacts1
Alisa I.Walz-Flannigan, PhD
Kimberly J. Brossoit, RT Visual familiarity with the variety of digital radiographic artifacts
Dayne J. Magnuson, RT is needed to identify, resolve, or prevent image artifacts from creat-
Beth A. Schueler, PhD ing issues with patient imaging. Because the mechanism for image
creation is different between flat-panel detectors and computed
Abbreviation: AAPM = American Association radiography, the causes and appearances of some artifacts can be
of Physicists in Medicine unique to these different modalities. Examples are provided of ar-
RadioGraphics 2018; 38:833–846 tifacts that were found on clinical images or during quality control
https://doi.org/10.1148/rg.2018170038
testing with flat-panel detectors. The examples are meant to serve
as learning tools for future identification and troubleshooting of
Content Codes:
artifacts and as a reminder for steps that can be taken for preven-
1
From the Department of Radiology, Mayo tion. The examples of artifacts provided are classified according to
Clinic, 200 First St SW, Rochester, MN 55901.
Presented as an education exhibit at the 2016 their causal connection in the imaging chain, including an equip-
RSNA Annual Meeting. Received March 6, ment defect as a result of an accident or mishandling, debris or
2017; revision requested October 13 and re-
ceived November 18; accepted December 6. For
gain calibration flaws, a problematic acquisition technique, signal
this journal-based SA-CME activity, the authors, transmission failures, and image processing issues. Specific artifacts
editor, and reviewers have disclosed no relevant include those that are due to flat-panel detector drops, backscatter,
relationships. Address correspondence to
A.I.W.F. (e-mail: walzflannigan.alisa@mayo.edu). debris in the x-ray field during calibration, detector saturation or
©
RSNA, 2018
underexposure, or collimation detection errors, as well as a variety
of artifacts that are processing induced.
SA-CME Learning Objectives ©
RSNA, 2018 • radiographics.rsna.org
After completing this journal-based SA-CME
activity, participants will be able to:
■■Identify digital radiographic artifacts
and the circumstances that resulted in Introduction
their appearance. An artifact on an image is a feature that does not correlate with
■■Discuss the cause of the artifact. the physical properties of the subject being imaged and may
■■Describe methods to eliminate the arti- confound or obscure interpretation of that image. In this ar-
fact and improve image quality.
ticle, examples of artifacts from flat-panel detector–based digital
See www.rsna.org/education/search/RG. radiographic systems are presented. The examples of artifacts are
paired with their cause and resolution as a means to aid in the
future identification, resolution, or prevention of digital radio-
graphic artifacts that may affect the quality of patient care. The
artifact examples provided were taken from clinical images as well
as routine quality control testing.
Flat-panel detector–based digital radiographic systems differ
in their image creation mechanism from computed radiography,
having an intrinsic pixel matrix and differences in signal detection
and processing. This difference can contribute to artifact types that
834  May-June 2018 radiographics.rsna.org

Some artifacts are easy to spot on a clinical

Teaching Points image; their effect may be to obscure parts of an
■■ Some artifacts may be seen clearly at retrospective viewing, image or to compromise the desired image field
but because of inattention bias, they might not be identified
during normal quality assurance steps or diagnostic inter-
of view. Examples of these types of artifacts may
pretations. This type of artifact may also affect patient care. arise from the misidentification of the desired im-
For example, the residual image of a marker, carried over to age field of view or from detector damage caused
subsequent images, could lead to misinterpretation of patient by dropping (“detector drop”). In these situa-
laterality. It is important to understand the conditions under tions, the need for response is clear, but it may
which these artifacts may be created, to either take steps to
prevent their clinical appearance or foster a targeted vigilance
be unclear how to prevent or address the issue
in quality assurance during situations in which the artifact is without understanding how it was created.
more likely to appear. Some artifacts may be seen clearly at retro-
■■ Providing technologists with a safe strategy and a demonstra- spective viewing, but because of inattention bias
tion of proper detector bagging, detector cleaning, or hold- (2), they might not be identified during normal
ing wireless flat-panel detectors during battery changes may quality assurance steps or diagnostic interpreta-
provide additional reduction in detector drop–related failures.
tions. This type of artifact may also affect pa-
For any practice, good communication detailing when detec-
tor drops occur and the situations in which they occur can tient care. For example, the residual image of a
help to identify targets for reducing the number of detector marker, carried over to subsequent images, could
drop–related artifacts. lead to misinterpretation of patient laterality (3).
■■ If projections of detector electronics are visible on a patient It is important to understand the conditions un-
image, options for a repeated image include using tighter col- der which these artifacts may be created, to either
limation (decreasing the field of view), verifying good imag- take steps to prevent their clinical appearance or
ing geometry, or placing additional shielding behind the de-
tector, such as a lead apron or plate to block the backscatter
foster a targeted vigilance in quality assurance
from striking the detector. during situations in which the artifact is more
■■ The most basic tool for avoidance of image saturation is to likely to appear.
ensure that the designated size-based techniques have been Other artifacts are more subtle and not as
followed. For manual techniques, the source-to-detector dis- easily seen on a clinical image. These artifacts are
tances and the thickness of the anatomy need to be mea- most likely to be found during routine quality
sured.
control testing. Ideally, any artifacts that cre-
■■ Temporary nonuniformities that are present in the x-ray beam ate deviation from ideal detector performance
during calibration (including lag from previous images) will
persist into subsequent imaging as an inverse image. For this
would be resolved; however, it is more practical
reason, it is important to visually validate the quality of a cali- to have a threshold for artifact identification to
bration with a flat-field image. help determine when an effort should be put into
resolution. This threshold for an artifact affect-
ing diagnostic interpretation may be rightfully
debated and may differ on the basis of the exami-
are unique to flat-panel detector–based digital nation, the specific detector exposure, and image
radiographic systems. More extensive details on processing. The threshold for responding to an
how flat-panel detectors are used in digital radio- artifact may also differ on the basis of whether it
graphic image creation, compared with computed can be eliminated through service on the detec-
radiography, are provided by Lança and Silva (1) tor or whether elimination of the artifact requires
and will not be reviewed here. However, each ar- purchase or warranty replacement of a detector.
tifact example provided will be classified accord- Examples of these more subtle artifacts include
ing to a type of failure mechanism in the digital flaws in gain calibration, debris interfering with
radiographic imaging chain. For organizational the x-ray beam, or visible detector tiling.
simplicity, artifacts are grouped according to their For routine quality control testing with a
causal connection. The types of artifact examples uniform phantom, we have established a strategy
presented include those related to equipment that uses a qualitative visual assessment with
defects or disruption from an accident or mishan- a viewing window that provides the identifica-
dling, debris or gain calibration flaws, acquisition tion threshold. Our method is similar to what
technique, and image processing issues. Although is described in the report from the American
certain artifact types may appear uniquely with Association of Physicists in Medicine (AAPM)
flat-panel detector–based digital radiographic sys- Imaging Physics Committee Task Group 151
tems, including any artifact related to a detector (4), which uses a qualitative visual evaluation of
gain calibration or disruption to readout circuitry, flat-field images in which anatomic image pro-
other artifacts related to image processing can be cessing has not been applied (ie, “original data,”
found in common with both flat-panel detector– as described in international standard IEC
based and computed radiography–based digital 62220-1) (5). Depending on the pixel value win-
radiographic systems. dow that is used for viewing, qualitative visual
RG  •  Volume 38  Number 3 Walz-Flannigan et al  835

Artifacts from Detector Accidents

or Mishandling

Detector Drops
Wireless flat-panel detectors get dropped.
Drops can damage the detector system in
several ways through fracturing of the flat-
panel detector, through disruption of readout
electronics, or through shifting gain and offsets.
Some vendor systems may have integrated
diagnostics software to alert the user when
detector drops are of sufficient magnitude for
a risk to image quality and may provide follow-
up instructions for how to respond. Following
Figure 1.  Detector drop. Chest radiograph obtained with instructions that direct a user to seek service
a portable radiographic unit shows a white band (arrow) at
or discontinue use of a damaged detector can
the top of the image. The white band appeared on images
obtained after a detector drop. prevent unnecessary repeated images. However,
because drops occur that can be withstood by
detectors without apparent damage and be-
cause some systems do not provide user feed-
back about the effect of a detector drop, the
truth of the drop’s effect is sometimes found in
an image itself. We provide examples of images
affected subsequent to detector drops.

Description of Artifact Appearance.—The

artifact shown as an example in Figure 1 is from
a cesium iodide flat-panel detector. Subsequent
to the detector being dropped, a narrow white
band appeared at the top of images. Review of
the images taken with the detector after the drop
showed that the white band grew progressively
Figure 2.  Detector drop. Chest radiograph obtained with larger with time, until the detector was pulled
a portable radiographic unit shows two types of artifacts: A from service. Figure 2 shows another example of
large white smoothly edged defect (horizontal arrow) ap- a detector that was used subsequent to a drop.
pears at top left, and straight lines (vertical arrow) are seen
across the middle of the image. These artifacts appeared on
In this case, a large area of signal dropout with
images obtained after multiple detector drops. a rounded perimeter was seen, in addition to
signal dropout along lines of signal readout. The
artifact in Figure 3 was found with a gadolinium
assessments may be effective but overly sensitive oxysulfide flat-panel detector and appears as
in identifying artifacts that may not have obvi- white speckles that do not correspond to any
ous clinical effects. Our threshold for identifica- dirt or surface debris.
tion and response is given by the appearance of
an artifact relative to pixel noise for an image Cause.—Although the exact mechanism for the
acquired with a detector exposure value in the progression from a thin line of signal dropout
middle of its exposure response curve. If any to what is shown in Figure 1 is not well under-
unexpected structure is depicted on a diagnos- stood, the known precipitating incident was a
tic workstation at a window width that is pro- detector drop. The appearance of the white band
portional to the magnitude of the pixel noise is presumed to be due to the disruption of the
(30 times the standard deviation of the central readout circuitry but could also be related to a
region of a flat-field), the unexpected structure loss of adhesion between the scintillator layer
is then followed up to determine if it can be and the thin-film transistor layer of the detector.
resolved or accepted. Vendor-supplied quality Because the lines of signal dropout in Figure
control software may also provide automated 2 appear parallel to detector readout, the most
tools that can identify artifacts; however, should likely explanation is the mechanical disruption
a test fail, visualization of a flat-field would likely in the readout circuitry. The large rounded area
still be needed to identify the source of failure of signal dropout in Figure 2 likely represents
and the steps needed to resolve it. a region of compromised adhesion between the
836  May-June 2018 radiographics.rsna.org

Figure 3.  Detector drop. Chest

radiograph obtained with a por-
table radiographic unit shows the
area of the inset (rectangle). Inset:
Magnified view shows multiple
white specks (arrows) toward the
top of the image. The specks ap-
peared on this image that was
obtained after a detector drop.
The speckled appearance of the
artifact relates to the gadolinium
oxysulfide phosphor structure of
the detector.

Figure 4.  Liquid contamination.

Chest radiograph obtained with a
portable radiographic unit shows
an artifact that appears as repeated
vertical banding. This image was
obtained after water contaminated
a bagged detector.

scintillator layer and the thin-film transistor provide for greater ease of handling to avoid
layer. The image shown in Figure 3 is presumed drops, such as incorporation of a handgrip. In
to be the result of minute shifts between the addition, protective cases are available, which
scintillator layer and the thin-film transistor some practices may find useful. Providing
layer of a gadolinium oxysulfide flat-panel detec- technologists with a safe strategy and a dem-
tor; such shifts affect the effectiveness of the onstration of proper detector bagging, detector
flat-field calibration. cleaning, or holding wireless flat-panel detectors
during battery changes may provide additional
Avoidance or Remedy.—Some detector drops reduction in detector drop–related failures. For
might result in shifts in signal gain or offset any practice, good communication detailing
that can be compensated for with recalibration when detector drops occur and the situations in
of the detector. The appearance of a speckle which they occur can help to identify targets for
pattern in Figure 3 subsequent to a detector reducing the number of detector drop–related
drop was resolved by a detector recalibration. artifacts.
Likewise, we have seen detector artifacts that
initially look like dead lines but were able to be Liquid Contamination
addressed through recalibration, indicating that Although it is not a novel concept that liquids do
signal readout was affected but not broken. not mix well with sensitive electronic equipment,
Other detector drops may result in mechani- unless such an incident is noticed, reported, and
cal failures in signal readout, creating artifacts addressed, it may result in imaging failures dur-
like those seen in Figures 1 and 2. In these ing examination of a patient.
circumstances, it is not possible to address the
failures through detector calibration. Instead, Description of Artifact Appearance.—After a
the detector needs to be replaced. leak in a patient’s water mattress during imag-
Some vendors advertise detectors that have ing, an image was acquired that shows banding
greater robustness with drops or detectors that throughout (Fig 4).
RG  •  Volume 38  Number 3 Walz-Flannigan et al  837

Figure 5.  Backscatter. (a) Radiograph obtained with a portable radiographic unit shows a shadow of detector electronics, created
by backscattered x-rays. (b) Image obtained when the detector was exposed upside down shows the detector electronics. (c) High-
lighted radiograph shows the matching features (yellow) that appear on the image of the detector electronics (b) and on the patient
radiograph with the backscatter artifacts (a), to highlight the similarities.

Cause.—Although the detector had been placed Cause.—Backscatter artifacts are more likely to
into a plastic bag, the bag was not one that could appear in situations with more scatter (ie, with
be effectively sealed. A substantial leak from a large patients or wide-open collimation) or if the
patient’s water mattress resulted in water getting x-ray beam is not fully intercepted by the de-
inside the bag. It is presumed that water infiltra- tector. This backscatter-related artifact is more
tion into the detector caused damage or disrup- possible with certain imaging geometries that
tion in the readout electronics that resulted in a use a wireless detector outside a table or upright
lack of signal, or false signal, that appears along holder, a situation in which it may be more chal-
the readout lines. lenging to properly angle the detector or use tight
collimation (as in some cross-table imaging).
Avoidance or Remedy.—For both infection
control and prevention of liquid infiltration, wire- Avoidance or Remedy.—If projections of detector
less flat-panel detectors are often bagged. Using electronics are visible on a patient image, op-
watertight protection, such as a sealable plastic tions for a repeated image include using tighter
bag, can better protect a detector from liquid ex- collimation (decreasing the field of view), verify-
posure such as that resulting in the failure shown ing good imaging geometry, or placing additional
in Figure 4. Vendors of newer models of flat-panel shielding behind the detector, such as a lead
detectors also advertise better water sealing to apron or plate to block the backscatter from strik-
prevent damage. The detector cannot recover ing the detector.
from the damage demonstrated in Figure 4, and
the detector needed to be replaced. Image Saturation
Information may be lost (“clipped”) from an
Artifacts Related to Acquisition image when an exposure exceeds the dynamic
Technique range imposed by an image processing algorithm
or in some circumstances in which the exposure
Backscatter exceeds the dynamic range of the image receptor.
Image artifacts can be created when backscat- It can be difficult to tell which is the source of the
tered radiation reaches detector elements through problem unless one is able to access an unpro-
the back of the detector (3). cessed (“for-processing”) image. Although either
computed radiography or digital radiography
Description of Artifact Appearance.—The provides greater exposure latitude than film, thus
shadow image of detector electronics appears reducing the likelihood of image receptor satura-
superimposed on a patient image in Figure 5a. tion within the anatomy of a patient image, such
Corresponding features between an image of the saturation may be more likely with digital radi-
detector electronics and the backscatter projec- ography than with computed radiography. The
tion of the electronics on the patient image are saturation point for digital radiography is typically
shown in Figure 5b and 5c. much lower than that for computed radiography
838  May-June 2018 radiographics.rsna.org

Figure 6.  Image saturation. (a) Abdominal radiograph obtained with a portable radiographic unit shows the
effect of excessive image exposure that resulted in a patient’s femurs and surrounding soft tissue being lost from
the image. (b) Upper abdominal radiograph shows image saturation (arrow) at the diaphragm, a location in
which image saturation is more commonly seen. For both images, a simulated-grid acquisition was used. Both
acquisitions used manual techniques for one size above what would have ideally been recommended on the
basis of patient thickness, as listed on the technique charts. The exposure was approximately double what was
recommended for the patient in a and was 40% greater than what was recommended for the patient in b.

(eg, a maximum image receptor air kerma for simulated-grid imaging, which does not use a
one flat-panel detector is approximately 90 μGy, grid but employs a scatter-suppression image
compared with approximately 440 μGy for a com- processing algorithm.
puted radiographic plate). (Dynamic range limita-
tions are as reported by a vendor representative.) Cause.—The manual technique used to ob-
Data from the report of the AAPM Imaging tain the image in Figure 6a was approximately
Physics Committee Task Group 232 provide twice what was recommended for the measured
some perspective on the likelihood of observing abdominal thickness but was only one size step
detector saturation (6). For imaging an abdo- larger on the manual technique chart that was
men, if the target image receptor air kerma is provided to the technologists. The image in
3.4 μGy and the image has a pixel value dy- Figure 6b also used a technique one size step
namic range of 50 (including tissues up to the larger than might have been recommended on
skin line), then with a detector saturation point the basis of the patient thickness, which resulted
of 90 μGy, one would anticipate seeing detec- in using a setting of 40% more milliampere-
tor saturation for a median exposure of 12 μGy. seconds than would have been desired.
This value for a median exposure would not be We saw an increase in image saturation arti-
anticipated in a typical acquisition. However, in facts with the initial use of simulated-grid acqui-
situations with higher image receptor air kerma sition, most frequently along the diaphragm at
targets or for patient images that may have a abdominal imaging, as shown in Figure 6b. The
greater pixel value dynamic range (greater dif- increased occurrence of image saturation arti-
ference between the thickest and thinnest ana- facts with simulated-grid images, compared with
tomic features), there could be a higher prob- gridded images, can be logically related to a
ability of detector saturation. higher recommended detector air kerma relative
to gridded techniques. (A higher recommended
Description of Artifact Appearance.—The target detector air kerma is inferred from
abdominal radiograph in Figure 6a represents a the default technique guides provided by the
dramatic case of image saturation (“clipping”) manufacturer, which show a 0%–35% reduc-
in which the lower part of the patient’s anatomy tion in the milliampere-second setting between
is completely missing from the image. Figure gridded and simulated-grid chest and abdomen
6b is an example of saturation-related clip- techniques for the same detector for a medium
ping at the diaphragm in an upper abdominal patient.) Technique recommendations, as pro-
radiograph. Attempts to reasonably recover the vided by the vendor, suggested between a 0%
image content and quality with image process- and 35% reduction in the milliampere-second
ing were unsuccessful for each example. Both setting between grid use and simulated-grid use
images in Figure 6 were acquired by using in acquisitions. However, maintaining the same
RG  •  Volume 38  Number 3 Walz-Flannigan et al  839

Figure 7.  Detector calibration limita-

tions. Clavicle radiograph shows that
vertical lines appear in the background
in the raw-radiation portion of the image
(rectangle), where the exposure exceeds
the ability of gain and offset corrections
to create a flat-field image.

detector exposure without a grid would require Avoidance or Remedy.—Some degree of limita-
a reduction of approximately 75% in technique. tion in flat-field correction may be anticipated
The net effect is that the recommended image when the detector receives exposures outside
receptor air kerma for simulated-grid imaging is its dynamic range. Because this type of artifact
higher than that for imaging with a grid. Aiming has caused some confusion in the past, it has
for a higher detector target in effect increases been important to educate technologists about
the likelihood of image saturation, creating a what this artifact looks like, so they understand
greater need for technique optimization and that it is not related to a failure in grid move-
careful technique selection. ment or suppression and so they learn to not
reject images unless the artifact extends into the
Avoidance or Remedy.—The most basic tool for anatomy. This type of artifact is also less likely
avoidance of image saturation is to ensure that when well-optimized target exposures are used
the designated size-based techniques have been for the images.
followed. For manual techniques, the source-
to-detector distances and the thickness of the Inverse Focal Spots and Calibration
anatomy need to be measured. In addition, we Artifacts
found that we needed to provide more size-spe- Debris in the x-ray beam creates an image on the
cific tailoring and further optimization to avoid detector, as might be expected of any x-ray–at-
image clipping with the simulated-grid acquisi- tenuating object found between the x-ray source
tion technique. Technique optimization involved and any image receptor. In particular, a small
extensive review of patient images, acquisition speck of x-ray–attenuating material near the head
techniques, and technologist-recorded patient of the x-ray tube can create an inverse image
sizes. With guidance from radiologists, techniques of the focal spot, acting as a pinhole camera as
that provided adequate noise properties with a described by Walker (8).
lower risk of artifacts were determined.
Description of Artifact Appearance.—A single
Artifacts from Limitations or Flaws in trapezoidal opacity can be seen in the flat-field im-
Detector Calibration age in Figure 8a, surrounded by the magnified and
offset appearance of its inverse (the area of lucency
Detector Calibration Limitations around the opacity). Figure 8b shows similarly ap-
pearing artifacts but of different sizes, locations, and
Description of Artifact Appearance.—The image orientations. Both images were acquired as part of
in Figure 7 shows vertical striping in the area of raw quality control testing of portable x-ray units, with
radiation. The lines are seen only in the background no added material in the x-ray beam, at 80 kVp,
on the image and do not extend into the anatomy. and with an image receptor air kerma of 17.5 μGy.

Cause.—The artifact lines demonstrate the Cause.—The cause of these artifacts was found
limits of gain and offset corrections of the detec- when the collimator was removed from the x-ray
tor calibration when the exposure is far from tube (Fig 9). Lead shavings seen near the head
the calibration conditions (as is the case when of the x-ray tube were created by the collimator
the detector is exposed to the unattenuated scraping against the beam-shaping port cylinder.
x‑ray beam). This evidence of detector structure The lead shavings created opaque-appearing
disappears at lower exposures. The threshold for pinhole images of the focal spot, with the size and
the appearance of detector structure will differ location depending on the location of the speck
for detectors that have different offset tolerances relative to the focal spot and the detector. The
between detector elements. darker trapezoids (Fig 8) represent the former
840  May-June 2018 radiographics.rsna.org

Figure 8.  Inverse focal spots and calibration artifacts. (a) Radiograph obtained with
flat-field acquisition shows a trapezoidal opacity. The trapezoidal opacity is the pinhole
image of the focal spot caused by lead debris near the head of the x-ray tube. Because
the image of the individual inverse focal spot was incorporated into the gain calibration
of the detector, its opposite is also seen (as the lucent area around the opacity), with
the differences in magnification and offset resulting from differences in the detector
position between gain calibration and subsequent image acquisition. (b) Radiograph
shows that as lead shavings build up near the head of the x-ray tube, a larger number
of inverse focal spots and their opposites are depicted.

Figure 9.  Lead debris was created near the head of the x-ray tube in a portable x-ray unit
when fully opened collimator blades rubbed against the lead beam-shaping port cylinder (ar-
row). This issue was discovered during the investigation of the cause of the image artifacts
shown in Figure 8.
RG  •  Volume 38  Number 3 Walz-Flannigan et al  841

Figure 10.  Flawed detector calibration.

Radiograph shows radiolucent lines near
the top of the image (rectangle). Inset:
Magnified view shows the inverse image
of a pillow edge as radiolucent wavy lines
(arrow). The pillow edge had been left in
the field of view during detector calibra-
tion, creating an artifact that remained
until the detector was recalibrated.

several incidents in which the artifact was visible on

clinical images led to a practice of further investiga-
tion to ensure that artifacts viewable in a quality
control viewing window were addressed.

Flawed Detector Calibration

Description of Artifact Appearance.—The image

in Figure 10 shows irregular lines that appear to be
radiolucent relative to the surrounding material.
These lines did not correspond to anything that was
being imaged. Subsequent images with this detec-
tor showed that the radiolucent wavy line artifact
remained in a fixed position on the detector when
the detector was rotated with respect to the tube.
Figure 11.  Evolving detector defect. Ra-
diograph shows that in addition to several
trapezoidal inverse focal spot artifacts on Cause.—The projection of a pillow edge on the
this flat-field image, many bloblike arti- detector during calibration became recorded in
facts (arrow) are depicted. These bloblike the flat-field correction. Although we adopt a
artifacts are differentiated from inverse fo-
quality control practice similar to that described
cal spot images because they are station-
ary when the detector is rotated and do in the AAPM Imaging Physics Committee Task
not change size with source-to-detector Group 151 report (4) to perform an artifact test
distance. The artifacts appeared within after detector calibration, this test had only been
6 months after a flat-field calibration in
done in the “free-detector” mode with the wire-
which the detector had been observed to
be artifact free. less detector outside the table. For this particular
digital radiographic system, detector flat-field
calibrations are done separately for imaging with
location of the inverse focal spot images that have the detector in the table or in the wall stand. As a
been recorded into the detector gain calibration. result, the artifact associated with the table cali-
bration was not found until clinical use.
Avoidance or Remedy.—The artifacts were
eliminated by removing the debris and recali- Avoidance or Remedy.—Temporary nonunifor-
brating the detector. Adjustment of the colli- mities that are present in the x-ray beam during
mator prevented further rubbing with the lead calibration (including lag from previous images)
port cylinder, reducing the appearance of these will persist into subsequent imaging as an inverse
artifacts. image. For this reason, it is important to visually
The gray-scale magnitude of the inverse focal validate the quality of a calibration with a flat-field
spot image created by the debris near the head of image. Since finding this artifact, we have used a
the x-ray tube can vary. Typically, these artifacts quality control strategy that tests all detector cali-
are minimally visible on patient images and instead bration configurations to ensure that no artifacts
are found during quality control testing. However, have been recorded into a flat-field correction.
842  May-June 2018 radiographics.rsna.org

Figure 12.  Failure of detector offset

correction. Abdominal radiograph
obtained with a portable x-ray unit
shows a superimposed inverse image
from a previous acquisition: the previ-
ous patient abdominal outline (black
arrow) and the previous laterality
marker (white arrow). This artifact oc-
curred when an offset calibration was
skipped after replacement of a detec-
tor battery.

Figure 13.  Electronic shutter failure.

(a) Cross-table lateral hip radiograph
appears to be an acquisition failure
but instead results from a failed detec-
tion of the collimator edges and poor
electronic shuttering. This artifact can
be fixed by reshuttering and repro-
cessing the image. (b) Repeat lateral
hip radiograph obtained after reshut-
tering and reprocessing shows a prop-
erly shuttered image of the same hip.

Evolving Detector Defect Failure of Detector Offset Correction

Description of Artifact Appearance.—A unique Description of Artifact Appearance.—Shown in

and subtle artifact was found regularly during qual- Figure 12, superimposed on an abdominal radio-
ity control testing with one particular type of detec- graph, are inverse images that show the outline of
tor. As shown in Figure 11, the artifacts appear as an abdomen and a laterality marker from a previ-
bloblike areas of opacity. The blobs are often visible ous image acquisition.
with our quality control window (30 times the noise
magnitude) but are generally not seen on clinical Cause.—After replacement of the battery in a
images. The artifacts remain stationary when the wireless digital radiographic detector, technolo-
detector is rotated with respect to the head of the gists are prompted to perform a detector offset
x-ray tube and hence are located on the detector calibration. In this case, when a detector bat-
itself. Although not seen immediately after detector tery needed replacement during the middle of
flat-field calibration, the artifacts have been seen to an examination, the technologists bypassed the
show up again 6 months later. calibration in “urgent” mode. When the offset
calibration is bypassed in urgent mode, the
Cause.—The presumed cause is related to detector residual signal on the detector becomes incor-
delamination “bubbles” that develop between the porated into the applied signal offsets until the
scintillator layer and the thin-film transistor layer of next offset calibration is done. Bypassing the
the detector. This change in coupling affects signal detector calibration resulted in an image pro-
transfer in the detector. portional to the inverse of this residual signal
appearing on a subsequent image (Fig 12).
Avoidance or Remedy.—The artifacts appearing as
bloblike areas of opacity seem to remain subclinical Avoidance or Remedy.—A new offset calibration
with regular detector recalibration. removed the residual image artifact. To avoid the
RG  •  Volume 38  Number 3 Walz-Flannigan et al  843

Figure 14.  Poor identification of values

of interest. (a) Patella radiograph ap-
pears washed out, giving the impression
that it was underexposed despite being
acquired at an exposure higher than that
recommended. (b) Repeat patella radio-
graph obtained after reshuttering to the
area within the dashed lines on a and
then reprocessing shows that the appear-
ance is greatly improved.

need to change batteries during an examination, tion. It is important that the image be repro-
when a technologist may not feel that there is cessed after the correct shutter is applied because
time to wait for the offset calibration, it is advis- the values of interest for processing have an effect
able to have a procedure for either scheduled on the final image appearance.
detector battery changes or monitoring battery
life. In addition, it was helpful for technologists Poor Identification of Values of Interest
to understand the effect of choosing the urgent
mode and to avoid doing so, if possible. Description of Artifact Appearance.—The im-
age in Figure 14a appears washed out, giving the
Artifacts Related to Image Processing impression that the anatomy was poorly penetrated
A predominant variation between different despite the use of a higher (large-patient) technique
digital radiographic (or computed radiographic) for a small- to medium-sized patient. Another ver-
systems relates to their image processing algo- sion of the same image, reprocessed after shutter-
rithms and the image processing settings that ing to the area within the dashed lines, is shown in
are used. Failures in image processing or a sub- Figure 14b. Figure 14b is more consistent with the
stantial lack of robustness in image processing anticipated appearance of patella radiographs.
outcomes can create image artifacts that limit
or complicate viewers’ ability to see the imaged Cause.—Processing outcomes are dependent on
anatomy in a consistent way. what values of interest are used by the image
processing algorithm. In this case, there ap-
Electronic Shutter Failure pears to have been a failure of proper exposure
recognition or perhaps an improper use of pixels
Description of Artifact Appearance.—A cross- outside the anatomy in setting contrast enhance-
table lateral hip radiograph that was acquired ment (9). The processing outcome was seen to
according to the standard procedure is shown in vary depending on the amount of raw radiation
Figure 13a. Interpreting the image problem as an area in the image, indicating that raw radiation
acquisition failure, the technologist repeated the was being included in the values of interest used
acquisition, with the result obtained with proper for image processing.
shuttering shown in Figure 13b.
Avoidance or Remedy.—Generally, image pro-
Cause.—The image processing algorithm incor- cessing algorithms should not include large areas
rectly identified the collimator blades and incor- of raw radiation in the values of interest used for
rectly applied electronic shutters. processing. If a correctly acquired image appears
to have inappropriate contrast and/or opacity, it
Avoidance or Remedy.—An imaging failure of is wise to check that only appropriate pixels are
this type caused by electronic shutter failure can contributing to the values of interest. Some ven-
be fixed through reshuttering and reprocessing dor systems may display which pixels have been
the image, rather than requiring a new acquisi- included as relevant, or the user may be able to
844  May-June 2018 radiographics.rsna.org

Figure 15.  Midgray clipping artifacts. (a) Knee radiograph shows the homogeneous gray appearance of the
cement (arrow), which indicates a loss of detail. (b) Knee radiograph obtained with a larger field of view shows
the area of the inset (rectangle). Inset: Magnified view shows the area of the tibia in which there is a loss of detail
(arrow). (c) Knee radiograph obtained with different processing from that used in a shows recovery of image
detail in the area of the cement and the cement-implant boundary. (d) Knee radiograph obtained with different
processing from that used in b shows recovery of image detail in the tibia.

constrain what is considered relevant by selecting appreciated and was found after more patient im-
a particular image region for processing. ages demonstrated the problem. The outcome of
signal equalization also depends on the selection of
Midgray Clipping values of interest. For Figure 15a, it was seen that
the poor processing outcome related to a change
Description of Artifact Appearance.—The im- in collimation between the image shown and the
age examples in Figure 15a and 15b show a loss images for which the processing was optimized
of information in parts of the knee radiographs (knee images with a larger field of view).
where multiple pixel values and structures are
anticipated but not visible. This loss is seen in the Avoidance or Remedy.—Image processing set-
loss of detail in the cement and cement-implant tings were adjusted to avoid the loss of informa-
boundary in Figure 15a and a loss of detail along tion. This adjustment required adding another
the anterior edge of the tibia in Figure 15b. processing option for the images with a small
field of view, which resulted in the image in Fig-
Cause.—The artifacts in Figure 15a and 15b ure 15c, and modifying the default processing
appear to be related to poor optimization of the setting for lateral knees, as shown in Figure 15d.
signal equalization processing, also known as
contrast enhancement (9–11). The image process- Metal Interface Artifacts
ing that had previously been set for the lateral For certain image processing algorithms, an at-
knee, shown in Figure 15b, was chosen for overall tempt to emphasize bone detail in the presence
contrast and detail preference with a clinical image of metal may lead to an artifact in the image of
sample. The loss of image information with this the metal or the metal-bone interface (10,12).
particular processing setting was not immediately Given the importance of being able to visualize
RG  •  Volume 38  Number 3 Walz-Flannigan et al  845

Figure 16.  Metal interface artifacts. (a) Knee radiograph shows artifacts at the metal-bone interface (rectangle). Inset:
Magnified view shows artifacts at the metal-bone interface, creating a false lucency (arrow) as well as a stairstepped
edge that does not correspond to the structure of the object being imaged. (b) Knee radiograph was obtained with the
same source image data but with the use of an alternative processing setting to address the lucency artifact (rectangle).
Inset: Magnified view shows elimination of the lucency artifact depicted in a.

Figure 17.  Grid-line suppression failure. Chest radiograph obtained with a portable radio-
graphic unit shows a failure of the grid-line suppression software (square). Inset: Magnified
image shows detail of grid lines.

this boundary in arthroplasty imaging, it is an lucency; however, the adjustments did not satis-
important artifact to address. factorily resolve the false structure on the edge of
the hardware.
Description of Artifact Appearance.—In the
image in Figure 16a, the processing creates a Grid-Line Suppression Failure
false appearance of lucency between the bone
and metal, a finding that might be suggestive of a Description of Artifact Appearance.—Grid lines
loosening (Fig 16a, inset). In addition, the edge appeared on the chest radiograph obtained with a
of the implant appears falsely serrated. portable radiographic unit shown in Figure 17.

Cause.—Lucency artifacts (Fig 16a, arrow) are Cause.—The example in Figure 17 shows a failure
presumed to be related to excessive edge en- of a grid-line suppression algorithm related to using
hancement. The processing setting had worked a grid with a line frequency that was incompatible
well for knee images with a larger field of view with the grid-line suppression software algorithm.
but did not work well with the small field of view.
Avoidance or Remedy.—Proper functioning of
Avoidance or Remedy.—In Figure 16b, the image the grid-line suppression software may depend
processing was adjusted to eliminate the false on automatic recognition that a grid is present.
846  May-June 2018 radiographics.rsna.org

Moreover, the grid-line suppression algorithm References

may be designed for a specific grid-line fre- 1. Lança L, Silva A. Digital imaging systems for plain radiog-
quency. Ensure that when a fixed grid is used, the raphy. New York, NY: Springer, 2013.
2. Drew T, Võ ML, Wolfe JM. The invisible gorilla strikes
grid-line frequency matches that recommended again: sustained inattentional blindness in expert observers.
by the system manufacturer. Psychol Sci 2013;24(9):1848–1853.
3. Walz-Flannigan A, Magnuson D, Erickson D, Schueler
B. Artifacts in digital radiography. AJR Am J Roentgenol
Conclusion 2012;198(1):156–161.
It is extremely important, to the extent that it 4. Jones AK, Heintz P, Geiser W, et al. Ongoing quality control
is possible, that the radiologist is able to have in digital radiography: report of AAPM Imaging Physics Com-
mittee Task Group 151. Med Phys 2015;42(11):6658–6670.
confidence that image information relates to the 5. International Electrotechnical Commission. IEC 62220-1:
patient anatomy rather than being false infor- medical electrical equipment—characteristics of digital x-ray
mation pertaining to an image artifact. A visual imaging devices: Part 1-1. Determination of the detective
quantum efficiency—detectors used in radiographic imaging.
familiarity with radiographic artifacts can help International Electrotechnical Commission website. https://
in identifying when image information cannot be webstore.iec.ch/publication/21937. Published March 3, 2015.
trusted. In addition, understanding the mecha- Accessed March 29, 2018.
6. Dave JK, Jones AK, Fisher R, et al. Current state of practice
nisms for the formation of various artifacts can regarding digital radiography exposure indicators and devia-
help in being able to resolve or prevent their tion indices: report of AAPM Imaging Physics Committee
appearance, to improve the consistency and Task Group 232. Med Phys (in press).
7. Shepard SJ, Wang J, Flynn M, et al. An exposure indicator
quality of radiographic imaging. Certain types for digital radiography: AAPM Task Group 116 (executive
of artifacts may appear uniquely with flat-panel summary). Med Phys 2009;36(7):2898–2914. [Published
detector–based digital radiographic systems, and correction appears in Med Phys 2010;37(1):405.]
8. Walker J. The amateur scientist: the pleasure of the pin hole
these types include any artifact related to a detec- camera. 1981. Sci Am 1981;245(5). Wes Jones website.
tor gain calibration or to a disruption of readout https://www.wesjones.com/pinhole.htm. Accessed March
circuitry. Other artifacts related to image process- 3, 2017.
9. Flynn M. Digital image processing in radiography. Presented
ing can be found in common with both flat-panel at the 49th annual meeting of the American Association of
detector–based and computed radiography–based Physicists in Medicine, Minneapolis, Minn, July 25, 2007.
digital radiographic systems. American Association of Physicists in Medicine website.
https://www.aapm.org/meetings/amos2/pdf/29-7999-58461-
92.pdf. Accessed November 10, 2017.
10. Wang X, Foos DH. Digital image processing in radiography.
Presented at the 47th annual meeting of the American As-
sociation of Physicists in Medicine, Seattle, Wash, July 27,
2005. https://www.aapm.org/meetings/05AM/pdf/18-2725-
40546-665.pdf. Accessed November 10, 2017.
11. Carroll QB. Radiography in the digital age. Springfield, Ill:
Charles C Thomas, 2014.
12. Berry E. A practical approach to medical image processing.
Boca Raton, Fla: Taylor and Francis, 2008.

TM
This journal-based SA-CME activity has been approved for AMA PRA Category 1 Credit . See www.rsna.org/education/search/RG.