Radiation Protection Dosimetry (2005), Vol. 114, Nos 1-3, pp.
359–363
doi:10.1093/rpd/nch510
BREAST DOSIMETRY USING HIGH-RESOLUTION VOXEL
PHANTOMS
D. R. Dance1,, R. A. Hunt1, P. R. Bakic2, A. D. A. Maidment2, M. Sandborg3, G. Ullman3
and G. Alm Carlsson3
1
Department of Physics, The Royal Marsden NHS Foundation Trust, Fulham Road,
London SW3 6JJ, UK
2
Department of Radiology, Hospital of the University of Pennsylvania, Philadelphia, PA 19104, USA
3
Department of Radiation Physics, IMV, Faculty of Health Sciences, Linköping University,
SE-581 85 Linköping, Sweden
A computer model of X-ray mammography has been developed, which uses quasi-realistic high-resolution voxel phantoms to
Downloaded from http://rpd.oxfordjournals.org/ at New York University on May 30, 2015
simulate the breast. The phantoms have 400 mm voxels and simulate the three-dimensional distributions of adipose and fibro-
glandular tissues, Cooper’s ligaments, ducts and skin and allow the estimation of dose to individual tissues. Calculations of the
incident air kerma to mean glandular dose conversion factor, g, were made using a Mo/Mo spectrum at 28 kV for eight
phantoms in the thickness range 40–80 mm and of varying glandularity. The values differed from standard tabulations used
for breast dosimetry by up to 43%, because of the different spatial distribution of glandular tissue within the breast. To study
this further, additional voxel phantoms were constructed, which gave variations of between 9 and 59% compared with
standard values. For accurate breast dosimetry, it is therefore very important to take the distribution of glandular tissues
into account.
INTRODUCTION average MGD for a population. In view of the
worldwide use of mammography for both screening
A Monte Carlo computer program has been devel-
and the examination of symptomatic women, it is
oped to realistically model mammographic X-ray
important to study the limitation of these simple
imaging systems. The model uses a voxelised phan-
models. The results presented here are based on
tom to simulate the breast and takes account of the
two series of voxel phantoms, which allow changes
various components of the imaging system including
in the distribution of glandular tissue to be made and
the X-ray spectrum, compression plate, anti-scatter
the effect of these changes on MGD to be studied.
device and image receptor. Anatomical details can
be included in the voxel phantom and the program
can be used to estimate measures of image quality(1). METHODS
The program also calculates the doses to the differ-
ent tissues simulated by the breast voxel phantom Monte Carlo model
and in particular to the glandular tissues, so that The Monte Carlo computer program is based on
the mean glandular dose (MGD) may be estimated. programs developed previously by our group(4,7),
The MGD is believed to be related to the risk of which have been extended by the addition of a
radiation-induced carcinogenesis, and is the quantity voxel phantom. The program follows photons from
normally used for breast dosimetry. When the MGD the focal spot of the X-ray tube, through the
is calculated in combination with measures of image compression plate and into the breast, simulating
quality, the model can be used for optimisation. photoelelectric interactions and coherent and inco-
This paper focuses on the use of the program and herent scattering. All energy deposited within each
a series of voxel phantoms for breast dosimetry. At breast tissue is recorded so that the dose to indivi-
present, the European protocol for breast dosime- dual tissues and the MGD can be estimated. The
try(2) is based on the calculations of Dance(3). Later incident air kerma at the upper surface of the breast
work by the same group(4) is used in the United (without backscatter) is also calculated.
Kingdom to facilitate calculations for breasts of
varying glandularity. The computer code used for
these calculations and that used in the United States Breast dosimetry
for the same purpose(5,6) were based on very simple In the European and United Kingdom Mammogra-
models of the breast. The use of such calculations phy protocols, the MGD, DG, for individual patients
will therefore result in a systematic error in the esti- is estimated from experimentally determined values
mation of the MGD for any particular breast, or the of the incident air kerma at the upper surface of the
breast, Ki, with the help of conversion coefficients
estimated from Monte Carlo calculation by using a
Corresponding author: david.dance@rmh.nthames.nhs.uk simple model of the breast (Figure 1). This simple
ª The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oupjournals.org
DANCE ET AL.
model assumes that the compressed breast is a fraction by weight for the whole breast. This defini-
cylinder of semi-circular cross section with a central tion is consistent with that used previously for the
region, which is a uniform mixture of adipose and calculations for simple breast models, although it is
glandular tissues surrounded by an adipose shield not well suited to some of the situations simulated.
of 5 mm thick. The MGD is given by: The eight structured phantoms only enabled a
limited study of the effect of the distribution of the
DG ¼ Ki g, ð1Þ
glandular tissue on the conversion coefficient, g, and
where the conversion coefficient g is the ratio of the it was necessary to develop a further series of phan-
MGD for the particular breast under consideration toms in order to study the effect of greater changes
to the incident air kerma. This coefficient depends on in the distribution. The new phantoms, referred to
the breast model used and the incident X-ray spec- as ‘unstructured phantoms’ were based on those of
trum. All data presented here are for a Mo/Mo Bakic et al.(8), and had central, surround and skin
X-ray spectrum at 28 kV with a half-value layer regions. However, the central and surround regions
(HVL) of 0.357 mm aluminium. were filled voxel-by-voxel rather than structure-
by-structure. The composition of each voxel was
Downloaded from http://rpd.oxfordjournals.org/ at New York University on May 30, 2015
Voxel phantoms chosen at random from the tissues present in the
region. For the surround region, the average compo-
Two series of voxel phantoms and calculations have sition used by Bakic et al.(8) was maintained. For the
been used to study the variation of the conversion central region, the glandularity could be varied to
coefficient, g, with the distribution of the glandular generate different phantoms. Figure 2 shows vertical
tissue within the breast. The first series of calcula- slices though a 50 mm structured phantom and the
tions used the breast voxel phantoms developed by unstructured phantom which simulates it. Varying
Bakic et al.(8). Here, these phantoms are referred to distributions of glandular tissue were obtained by
as ‘structured phantoms’. They simulate the uncom- moving the boundary between the central and
pressed breast with three regions. The central region surround regions of the phantoms, maintaining con-
is connected to the nipple and contains glandular nectivity with the nipple. Two series of calculations
tissue, adipose tissue and a ductal tree. The second were performed. In the first series, both the upper
region surrounds the central region and contains and lower boundaries between the central and
adipose tissues and Cooper’s ligaments. Here, it is surround regions were raised in such a way that the
referred to as the ‘surround region’. The third region volume of the central region and connectivity with
contains skin. The breast is modelled using random the nipple were maintained. In the second series, the
numbers to select the position and size of structures upper and lower boundaries were moved symmetri-
within each region according to chosen distribution cally about the horizontal plane through the nipple.
laws. A voxel size of 400 mm was used for the Figure 3a and b show vertical slices through phan-
present work. Eight phantoms constructed in this toms used for the two series of calculations. The first
way were used for the first series of calculations. and second series of phantoms were obtained by
They corresponded to breast thicknesses in the asymmetric and symmetric distortions, respectively.
range 40–80 mm and glandularities in the range of The validity of the above approach was checked
25–100%. The term ‘glandularity’ refers to the frac- by calculating the conversion coefficient, g, using
tion by weight of glandular and ductal tissues within unstructured and structured phantoms with the
the central region of the breast, rather than the same boundaries and average compositions.
Agreement within 1–5% was found.
RESULTS
Calculations for structured phantoms
Table 1 gives the values of the conversion coefficient,
g, calculated for the eight structured phantoms and
the 28 kV Mo/Mo spectrum. The corresponding
values of g deduced from the tabulations of data
based on earlier, simple models of the breast(3,4) are
also given. There are significant differences between
the two sets of values, which range from 10 to 43%.
Mixed adipose and Adipose The differences increase with increasing breast thick-
glandular tissue shield ness and decreasing glandularity. The reason for the
Figure 1. Simple geometrical model of the breast used difference between the two sets of values can be
previously for the calculation of the MGD. understood by reference to Figures 1 and 2. The
360
BREAST DOSIMETRY USING VOXEL PHANTOMS
(a) (b)
Figure 2. Vertical slices through 50 mm structured (a) and unstructured (b) breast voxel phantoms. Each has a
glandularity of 69% in the central region.
Downloaded from http://rpd.oxfordjournals.org/ at New York University on May 30, 2015
(a) (b)
Figure 3. Vertical slices through 50 mm unstructured breast voxel phantoms. Each has a glandularity of 69% in the
central region. (a) Left, asymmetric distortion; (b) right, symmetric distortion.
Table 1. Comparison of the conversion factor g calculated using a series of high-resolution voxel phantoms with the value
calculated using a simple phantom.
Breast Breast g-factor: g-factor: simple Relative
thickness glandularity structured voxel phantom difference
(mm) (%) phantom (mGy mGy1) (mGy mGy1) (%)
40 100 0.177 0.195 10
40 57 0.202 0.232 13
40 39 0.216 0.250 14
50 69 0.140 0.174 20
60 51 0.114 0.156 27
60 27 0.122 0.176 31
80 47 0.0667 0.116 42
80 25 0.0735 0.130 43
Note: 28 kV Mo/Mo spectrum, 0.357 mm Al HVL.
average distance of the boundary between the central plotted against the average value, d, of the distance
and surround regions from the breast surface in the from the breast upper surface to the boundary
structured phantom is greater than the distance of between the central and surround regions of the
5 mm used in the simple phantom. breast.
The results in Figure 4a show that, depending
upon the level of movements and distortion applied,
Calculations for unstructured phantoms
the conversion coefficient, g, can be greater than or
Figure 4a and b show the results of the first and less than that calculated from the simple model,
second series of calculations for the unstructured ranging from 52 to 122% of that value. The two
voxel phantoms. In both cases, the results are for a values are equal at a d-value of about 12 mm.
28 kV Mo/Mo spectrum and a 50 mm compressed Equality for a d-value of 5 mm is not expected
breast of average glandularity 69%. The results are because of the asymmetrical distribution of
361
DANCE ET AL.
(a) of the breast provides a powerful tool for the study
0.25 of breast dosimetry. Significant differences have
Conversion coefficient g (mGy/mGy) been found between the incident air kerma to
0.20
MGD conversion coefficients tabulated in the litera-
ture and those obtained in this work. For the cases
considered, the differences can be as large as 48%
0.15 and are due to differences in the distribution of the
glandular tissue within the breast. These results
0.10 clearly demonstrate the limitations of the data that
are used currently for breast dosimetry.
At present, there are no data available on the
0.05 actual three-dimensional distributions of glandular
tissue within the breast for populations of women.
0.00 Therefore the methodology cannot yet be used at
present to provide better estimates of population
Downloaded from http://rpd.oxfordjournals.org/ at New York University on May 30, 2015
0 5 10 15 20 25
Distance to glandular region (mm)
dose for screening programmes; it is not suggested
(b) at present that the data in current use for the
0.25 estimation of breast dose be revised.
Conversion coefficient g (mGy/mGy)
Data for individual women can in principle be
obtained from volume imaging using CT or MR,
0.20
but the process is difficult and not well established.
However, the results of this work can be used to
0.15 provide better estimates of MGD for individual
cases where there is some knowledge of the distribu-
0.10 tion of glandular tissue within the breast.
0.05
ACKNOWLEDGEMENTS
0.00
This work has been funded by the Commission of
European Communities 5th Framework Programme
0 5 10 15 20 25
(grant CT200-0036 9, the ‘Radius Project’) and
Distance to glandular region (mm) the US Department of Defense (grant DAMD
Figure 4. Values of the conversion coefficient g calculated 17-98-1-1819).
using two sets of unstructured phantoms of 50 mm
thickness and glandularity 69% in the central region. (a)
Phantoms with asymmetric distortion of the central region
and (b) phantoms with symmetric distortion of the central REFERENCES
region. The horizontal dashed lines give the value of the
1. Hunt, R. A., Dance, D. R., Bakic, P. R.,
conversion coefficient deduced from the data and simple
Maidment, A. D. A., Sandborg, M., Ullman, G. and
breast model of Dance et al.(3,4).
Alm Carlsson, G. Calculation of the properties of digital
mammograms using a computer simulation. Radiat. Prot.
glandular tissue, with more tissue being above Dosim. 114(1-3), 395–398 (2005).
the midplane of the breast than below. For the 2. European Commission. European protocol on dosimetry
in mammography. European Commission Report EUR
symmetrical case, Figure 4b, there is a much smaller 16263 EN (Luxembourg: EC) (1996), ISBN 92-827-
variation in the conversion coefficient when com- 7289-6.
pared with the distance d, ranging from 75 to 92% 3. Dance, D. R. Monte Carlo calculation of conversion
of the value using the simple model. This is because factors for the estimation of mean glandular breast dose.
of the change in the amount of tissue above and Phys. Med. Biol. 35, 1211–1219 (1990).
below the midplane is the same. It is noted that the 4. Dance, D. R., Skinner, C. L., Young, K. C.,
value of the conversion coefficient for a distance d Beckett, J. R. and Kotre, C. J. Additional factors for
of 5 mm is similar to that obtained with the simple the estimation of mean glandular breast dose using the
model. UK mammography dosimetry protocol. Phys. Med. Biol.
45, 3225–3240 (2000).
5. Wu, X., Barnes, G. T. and Tucker, D. M. Spectral
DISCUSSION AND CONCLUSIONS dependence of glandular tissue dose in screen-film
mammography. Radiology 179, 143–148 (1991).
Our Monte Carlo computer model of mammogra- 6. Wu, X., Gingold, E. L., Barnes, G. T. and Tucker, D. M.
phy, incorporating quasi-realistic voxelised models Normalized average glandular dose in molybdenum
362
BREAST DOSIMETRY USING VOXEL PHANTOMS
target-rhodium filter and rhodium target-rhodium filter absorbed dose in mammography: a Monte Carlo study.
mammography. Radiology 193, 83–89 (1994). Br. J. Radiol. 73, 1056–1067 (2000).
7. Dance, D. R., Thilander Klang, A., Sandborg, M., 8. Bakic, P. R., Albert, M., Brzakovic, D. and Maidment,
Skinner, C. L., Castellano-Smith, I. A. and Alm A. D. A. Mammogram synthesis using a 3D simulation.
Carlsson, G. Influence of anode/filter material and tube 1. Breast tissue model and image acquisition simulation.
potential on contrast, signal-to-noise ratio and average Med. Phys. 29, 2131–2139 (2002).
Downloaded from http://rpd.oxfordjournals.org/ at New York University on May 30, 2015
363
