AJR 2004; 182:713-717
© American Roentgen Ray Society
Mammographic Breast Glandularity in Malaysian Women: Data Derived from Radiography
Noriah Jamal1,
Kwan-Hoong Ng1,
Donald McLean2,3,
Lai-Meng Looi4 and
Fatimah Moosa1
1 Department of Radiology, University of Malaya, Kuala Lumpur 59100,
Malaysia.
2 School of Medical Radiation Sciences, University of Sydney, Sydney, NSW 1825,
Australia.
3 Medical Physics Department, Westmead Hospital, Sydney, NSW 2145,
Australia.
4 Department of Pathology, University of Malaya, Kuala Lumpur 59100,
Malaysia.
Received May 2, 2003;
accepted after revision September 22, 2003.
Address correspondence to N. Jamal
(noriahj{at}mint.gov.my).
Supported by research grant F Vote F0199/2003A from University of Malaya
and the Malaysian Institute for Nuclear Technology Research (MINT).
Presented at the 2003 World Congress on Medical Physics and Biomedical
Engineering, Sydney, Australia.
Abstract
OBJECTIVE. This study was undertaken to estimate mammographic breast
glandularity in Malaysian women from radiographic data.
MATERIALS AND METHODS. A mammography X-ray unit was used to expose
different thicknesses of phantom material of varying glandular and adipose
composition at 27 kV. A least squares method was then used to fit the combined
data of phantom glandularity, thickness, and milliampere-seconds. The
subsequent fitted equation was then applied to calculate breast glandularity
for 705 women who underwent diagnostic mammography, who were drawn equally
from the three major ethnic groups of Malaysia: Malay, Chinese, and Indian.
The difference in breast glandularity among ethnic groups was tested for
significance using the nonparametric Kruskal-Wallis test.
RESULTS. The fitted equation gave an absolute error of less than or
equal to ± 8% when applied to the data from phantom exposure. The
average breast glandularity of the study sample was 48.9% ± 18.7%.
Breast glandularity was found to decrease with breast thickness and age.
CONCLUSION. No significant difference was seen in breast
glandularity among the ethnic groups (p > 0.05, Kruskal-Wallis
test).
Introduction
At present, it is generally assumed that glandular tissue, which is a
common site for breast cancer, is the most vulnerable among the tissues
(adipose, skin, and areolar tissues) making up the breast
[1]. The amount of glandular
tissue is linked to breast cancer risk, so an objective quantitative analysis
of glandular tissue can aid in risk estimation
[2]. In a previous study
[3], we estimated the mean
glandular dose during diagnostic mammography in Malaysian women using the
assumption that all breasts were composed of 50% glandular and 50% adipose
tissue. However, in the course of calculating mean glandular radiation dose of
a patient, knowledge of the breast glandularity and compressed breast
thickness for each breast is neccessary to choose mean glandular dose
conversion factors
[4–6].
This information will permit mean glandular dose calculations to be extended
from breasts of average composition (50% glandular and 50% adipose) to breasts
of individually determined composition.
Various approaches have been suggested to estimate breast glandularity, and
each of them has its own limitations. A subjective approach is to trace the
fibroglandular parts of the breast and measure their percentage of the whole
breast area [7,
8]. Some authors have proposed
automated approaches to measure breast glandularity, which do make the measure
objective [9,
10]. However, these
measurements do not correspond to the anatomy of the breast or to the imaging
physics because the X-ray beam usually passes through a mixture of both
adipose and dense tissue before striking the detector. Thus, the map of
fibroglandular tissue does not correspond to reality
[2]. More recently, Highnam et
al. [11] measured breast
glandularity using the thickness of glandular tissue between the pixel and the
X-ray source. However, those authors argue that their work should be used only
for relative measurements unless a careful calibration has been performed.
Kaufhold et al. [2] measured
breast glandularity on a pixelwise basis, with a calibration approach after
that of Highnam et al. Bloomquist et al.
[12] published a similar work
on estimating breast glandularity using a volumetric technique. However, the
error arises because of compressed breast thickness estimation, residual
scattered radiation, quantum noise, and beam hardening
[2].
In the mid 1990s, studies by Cross
[13] suggested that breast
glandularity could be determined from radiographic data (tube potential [kV]),
tube loading [mAs], and compressed breast thickness). Since then, a few
studies have looked at breast glandularity estimation from radiographic
data.
The purpose of our study was to estimate mammographic breast glandularity
in Malaysian women from radiographic data. The present work extends the prior
approach of Heggie [14] to
include the 0.5-cm-thick adipose tissue as an outer layer, following a
definition of breast glandularity by Dance et al.
[4] and Beckett and Kotre
[6] and a model proposed by
Stanton et al. [15]. This work
is important in two aspects: from a fundamental point of view, it explains the
basis of breast glandularity estimation from radiographic data; and from a
practical point of view, an estimate of mammographic breast glandularity in
Malaysian women helps in choosing mean glandular dose conversion factors
[4] to calculate the mean
glandular dose to the breast. Subsequently, these mean glandular doses will be
used to estimate benefits and risks associated with radiation doses and to
predict the likely prevalence of radiation-induced cancer arising from
mammography examinations. To our knowledge, ours is a first attempt to
estimate mammographic breast glandularity from radiographic data for a defined
population in Southeast Asia.
Materials and Methods
Phantom Study
A mammography X-ray unit (Nova 3000, Siemens, Munich, Germany) operated
using a molybdenum target and filter was used in this study. An antiscatter
grid with a grid ratio of 5:1 was used with a nominal focal spot size of 0.3
mm and a focus-to-film distance of 66 cm. This unit has undergone an extensive
quality assurance procedure according to the recommendations of the American
Collage of Radiology [16].
Table 1 shows breast models
currently in use for mammography dosimetry and measurement of breast
glandularity.
For the purpose of this study, we chose to define breast glandularity as a
percentage by mass of glandular tissue after allowing for the 0.5-cm surface
layer of adipose tissue [4] and
leading to a relation [1,
17] that the total of
glandular and adipose tissues is unity. The breast model adopted in this study
is shown in Figure 1.
CIRS (Computerized Imaging Reference Systems, Norfolk, VA) mammography
phantom material (total thickness, 3, 4, 5, and 6 cm) of varying composition
(100% adipose to 100% glandular) was exposed at 27 kV, which is the most
commonly used kilovoltage. No attempt was made to verify the phantom
composition. All phantoms were rectangular in shape, with dimensions of 13.5
(along the side placed parallel to the chest wall edge of the cassette holder)
x 10 cm. Cassette and film were in place throughout the phantom
measurement. The automatic exposure control detector nearest the chest wall
was used to simulate the clinical situations. The setting for
milliampere-seconds was in auto mode. The experiment data— phantom
glandularity, milliampere-seconds, and phantom thickness—were recorded
for each exposure. The least squares method was used to fit the combined data
(see Appendix 1). We found that the fitted equation is:
where g is the breast glandularity and t is the breast
thickness.
The fitted equation was constrained to give glandularities of 0–100%
to ensure compliance with the model used
[15] and had little effect on
the fitted glandularity for older women
[6]. However, for younger
women, the assumption that there is a 0.5-cm-thick adipose layer becomes
unrealistic for very thin breasts (
2.5 cm) and leads to an estimated
glandularity of more than 100%
[4]. The ability of the fitted
equation to correctly predict the percentage of breast glandularity from
recorded milliampere-seconds was examined.
Clinical Study
This research did not require informed consent from subjects and was deemed
to be exempt from institutional review board approval. The formula was applied
to data recorded for 705 patients (Malay, 235; Chinese, 235; and Indian, 235)
between 30 and 79 years old (median age, 51 years) who underwent diagnostic
(referral) mammography during 2002 that was performed using the Siemens unit
at the University of Malaya Medical Centre. All mammograms were obtained using
27 kV. For each film, the breast glandularity was calculated using the fitted
equation given previously. Only craniocaudal images were included in this
study because less muscle is included, thereby simplifying the analysis of
glandularity. The milliampere-seconds, kilovoltage, breast thickness, and
views obtained were automatically recorded on the images at the time of
exposure. The difference between displayed and actual thickness at the chest
wall was evaluated by measuring the thickness of five breasts (craniocaudal
view). A correction was then applied to the displayed breast thickness.
The statistical significance of differences in breast glandularity among
ethnic groups was tested using the nonparametric Kruskal-Wallis test. The
relation of breast glandularity to breast thickness and age was investigated.
A total of 99 breast thickness and 112 age data points were assembled from the
whole set of data. Fitted lines were plotted using an inverse second-order
polynomial. The results were divided into four groups according to the quality
assurance manual of the American College of Radiology
[16].
Results
The fitted equation results in an absolute error in fitted glandularity
(expressed as a percentage) of no greater than ± 8% for the entire
range of values. Figure 2 shows
the ability of the fitted equation to correctly predict the percentage of
phantom glandularity on the basis of the recorded milliampere-seconds on the
CIRS phantom (3, 4, 5, and 6-cm thick) with a wide range of phantom
glandularity.
Table 2 shows distribution
of breast thickness, age, and breast glandularity of the study sample. Mean
breast glandularity of the study sample was 48.9% ± 18.7%. No
significant difference was seen for breast glandularity among the different
ethnic groups (p > 0.05, Kruskal-Wallis test).
The expected dependence of breast glandularity on breast thickness and age
is shown in Figures 3 and
4, respectively.
Results of our study are compared with those of similar recent studies in
Table 3 for four breast
thickness groups described in the quality assurance manual of the American
College of Radiology [16].
Discussion
Our study builds on the prior approach of Heggie
[14]. The breast model used in
this work also provides an extension of those in the American College of
Radiology [16] and Food and
Drug Administration [18]
recommendations, in accordance with the Institute of Physical Sciences in
Medicine [5] and the National
Council on Radiation Protection and Measurement
[19] for mammography
dosimetry. The introduction of 0.5-cm adipose tissue (top and bottom) provides
a simple approach that retains continuity with the actual clinical situation.
This model may be adequate as an aid in developing mammographic dosimetry
tables, particularly in calculating mean glandular dose conversion
factors.
Figure 3 shows that an
absolute difference of breast glandularity of roughly 9.6% exists between 3-
and 6-cm breast thicknesses. This finding is lower than the result reported by
Beckett and Kotre [6] of 88.12%
between 2- and 6-cm breast thickness. A possible explanation could be that our
study is limited to diagnostic (referral) mammography.
Figure 4 shows that breast
glandularity decreases with increasing age (20.1% reduction of breast
glandularity from 47 to 72 years). This decrease is due to an increase in the
proportion of adipose tissue in the breast
[20]. This trend is similar to
that reported by Klein et al.
[1], Beckett and Kotre
[6], and Soares et al.
[20] for German, British, and
Jamaican studies, respectively. Interestingly, we found that the greatest rate
of change occurs after the age of 50 years, whereas Dance et al.
[4] and Beckett and Kotre
reported that the greatest rate of change occurs between the ages of 45 and 55
years. A possible explanation for this difference could be that our study
involves only symptomatic patients (in whom breast carcinoma exists and breast
glandularity is relatively high) with a small study sample (112 data points).
When comparing Figures 3 and
4, we found that breast
thickness has the strongest modifying effect on breast glandularity. This
finding is similar to that reported by Dance et al. and Beckett and Kotre.
Table 3 shows that the
average breast glandularity obtained from our study is higher than that
reported in the United States' studies
[21] and comparable to values
reported for Australia [14]
and Germany [1]. These
differences may be due to the categories of breast thickness studied. However,
Heggie [14] and Geise and
Palchevsky [21] estimated
breast glandularity without the adipose layer in the breast model used,
whereas Klein et al. [1] used
different types of phantoms with a tungsten anode. The breast glandularities
and breast thicknesses presented here do not include measurements in women
from other parts of Malaysia and might not be typical of women at other
geographic locations or with other minority ethnic distributions.
Our study had some limitations. The objective nature of the technique used
removes concerns about observer variability in a subjective study in
quantifying breast glandularity. Further, this method is relatively easy and
does not involve image processing techniques or quantitative interpretation of
fibroglandular and adipose tissues on mammograms
[7–11].
However, uncertainties in our results arose primarily from the accuracy of the
breast thickness display unit
[21], limitations in the
curve-fitting process, and variability of milliampere-second values from day
to day. Another limitation of this approach is that the automatic exposure
control does not cover the entire breast area, so the value that is calculated
is in reality the average glandularity of the volume above the automatic
exposure control detector
[6].
In conclusion, the average breast glandularity of the study sample was
48.9% ± 18.7%. No significant difference was seen for breast
glandularity among the different ethnic groups.
APPENDIX 1. Basis of Breast Glandularity Estimation from Radiographic Data
For a long time it has been known that X-rays attenuate exponentially.
Thus, a constant tube potential (kV) and scatter rejection condition lead to
the relationship:
 | (1) |
where a and b are fit parameters, and t is breast
thickness. In a modern mammography unit, the exposure is controlled by the
automatic exposure control system, which detects the amount of radiation
reaching a detector behind the imaging cassette to give uniform density on
successive film exposures. Thus, for a fixed kilovoltage, a very glandular
breast requires a higher milliampere-second setting than a very fatty breast
does. The relation of breast glandularity (g) to tube loading may
have a form [13]:
 | (2) |
In this method of calculation, g is represented by the use of two
arbitrary coefficients—A(t) and
B(t)—that need to be obtained. Therefore, if
experimental data pairs of g and ln(mAs) are plotted, a
straight line graph of the form y = mx + c can be
obtained, where y = g, m = A(t), and
c = B(t). We used the least squares method
[22] to choose the best values
for the arbitrary coefficients A(t) and
B(t), where the quantity of sum of squares of the errors
(SEE) as shown below:
 | (3) |
about the regression line is minimum to achieve the closest agreement between
g and ln(mAs). These requirements were achieved by making
each of the partial derivatives
and
equal
zero.
If data pairs of A(t) and inverse values of thickness are
plotted, a straight line graph of the form y = mx +
c can be obtained, where y = A(t),
c = a1, and m = a2.
Again, if data pairs of B(t) and inverse values of the
thickness are plotted, a straight line graph of the form y =
mx + c can be obtained, where y =
B(t), c = a3, and m =
a4. Thus, the calculated g involves four fitted
parameters in the form:
 | (4) |
Again, we have used the least squares methods to chose the best values for
a1 and a2, which are the fitted
parameters corresponding to the arbitrary coefficient A(t),
and a3 and a4 are the fitted
parameters corresponding to the arbitrary coefficient B(t).
In this case, the partial derivatives
,
,
,
must be zero.
Acknowledgments
We thank J. C. P. Heggie, St. Vincent's Hospital, Melbourne, Australia, for
his useful input to the application of the breast model and least squares
analysis.
References
- Klein R, Aichinger H, Dierker J, et al. Determination of average
glandular dose with modern mammography units for two large groups of patients.
Phys Med Biol1997; 42:651
–671[Medline]
- Kaufhold J, Thomas JA, Eberhard JW, Galbo CE, Trotter DE. A
calibration approach to glandular tissue composition estimation in digital
mammography. Med Phys2002; 29:1867
–1880[Medline]
- Jamal N, Ng K-H, McLean D. A Study of mean glandular dose during
diagnostic mammography in Malaysia and some of the factors affecting it.
Br J Radiol2003; 76:238
–245[Abstract/Free Full Text]
- Dance DR, Skinner CL, Young KC, Beckett JR, Kotre CJ. Additional
factors for the estimation of mean glandular breast dose using the UK
mammography dosimetry protocol. Phys Med Biol2000; 45:3225
–3240[Medline]
- The Institute of Physical Sciences in Medicine. Patient
dosimetry techniques in diagnostic radiology. Report no. 53.
York, England: IPSM, 1988:43
–45
- Beckett JR, Kotre CJ. Dosimetric implications of age related
glandular changes in screening mammography. Phys Med
Biol 2000;45:801
–813[Medline]
- Byng JW, Yaffe MJ, Jong RA, et al. Analysis of mammographic density
and breast cancer risk from digitized mammograms.
RadioGraphics1998; 18:1587
–1598[Abstract]
- Boyd NF, Byng JW, Jong RA, et al. Quantitative classification of
mammographic densities and breast cancer risk: results from the Canadian
National Breast Screening Study. J Natl Cancer Inst1995; 87:670
–675[Abstract/Free Full Text]
- Byng JW, Boyd NF, Fishell E, Jong RA, Yaffe MJ. Automated analysis
of mammographic densities. Phys Med Biol1996; 41:909
–923[Medline]
- Chan Y-H, Wang X-H, Hardesty LA, et al. Computerized assessment of
tissue composition on digitized mammograms. Acad
Radiol 2002;9:899
–905[Medline]
- Highnam R, Brady M, Shepstone B. A representation for mammographic
image processing. Med Image Anal1996; 1:1
–18[Medline]
- Bloomquist A, Augustine B, Pawluczyk O, Mawdsley GE, Yaffe MJ.
Estimating breast density using a volumetric technique and full field digital
mammograms: preliminary results. In: Fifth International Workshop on
Digital Mammography. Sunnybrook & Women's College Health Sciences
Centre, Toronto, Canada, June 11–14, 2000:126. Available at
www.sunnybrook.utoronto.ca/~iwdm2000/iwdm2000_abstract_book.pdf.
Accessed July 7, 2003
- Cross P. Doses in mammography: from the phantom to the patient.
Australas Radiol1994; 38:20
–23[Medline]
- Heggie JCP. Survey of doses in screening mammography.
Australas Phys Eng Sci Med1996; 19:207
–216[Medline]
- Stanton L, Vallafana T, Day JL, Lightfoot DA. Dosage evaluation in
mammography. Radiology1984; 150:577
–584[Abstract/Free Full Text]
- American College of Radiology, Committee on Quality Assurance in
Mammography. Mammography quality control manual.
Reston, VA: American College of Radiology, 1999:144
–145
- McLean D. Breast composition and radiographic breast equivalence.
Australas Phys Eng Sci Med1997; 20:11
–19[Medline]
- Rosenstein M, Andersen LW, Warner GG. Handbook of
glandular tissue doses in mammography. HHS (FDA) publication
85-8239. Washington, DC: U. S. Government Printing Office, 1987:14
–15
- National Council on Radiation Protection and Measurements.
Mammography: a user's guide. Report no. 85. Bethesda,
MD: National Council on Radiation Protection and Measurements,1986
- Soares D, Reid M, James M. Age as a predictive factor of
mammographic breast density in Jamaican women. Clin
Radiol 2002;57:472
–476[Medline]
- Geise RA, Palchevsky A. Composition of mammographic phantom
materials. Radiology1996; 198:347
–350[Abstract/Free Full Text]
- Walpole RE, Myers RH, Myers SL. Probability and
statistics for engineers and scientists, 6th ed. Upper Saddle
River, NJ: Prentice Hall, 1998:362
–420
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
Top
Abstract
Introduction
= 3
cm, 
= 4 cm, 
= 5 cm, x = 6 cm.
