Note: This copy is for your personal non-commercial use only.

To order presentation-ready
BREAST IMAGING copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

316

Breast Density: Clinical Implications

and Assessment Methods1
Nicole S.Winkler, MD
Sughra Raza, MD
Breast density assessment is an important component of the screening
Meaghan Mackesy, MD mammography report and conveys information to referring clinicians
Robyn L. Birdwell, MD about mammographic sensitivity and the relative risk for developing
breast cancer. These topics have gained substantial attention because of
Abbreviations: ACRIN = American College of recent legislation in several states that requires patients to be informed
Radiology Imaging Network, BI-RADS = Breast of dense breast tissue and the potential for associated breast cancer risk
Imaging Reporting and Data System
and decreased mammographic sensitivity. Because of the considerable
RadioGraphics 2015; 35:316–324 implications of diagnosing a woman with dense breast tissue, radiolo-
Published online 10.1148/rg.352140134 gists should strive to be as consistent as possible when assessing breast
Content Codes:
density. Commonly used methods of breast density assessment range
from subjective visual estimation to quantitative calculations of area
1
From the Department of Radiology, Brigham
and Women’s Hospital, 75 Francis St, Boston, and volume density percentages made with complex computer algo-
MA 02115. Presented as an education exhibit at rithms. The basic principles of currently available commercial methods
the 2013 RSNA Annual Meeting. Received April
2, 2014; revision requested July 21 and received
of calculating fibroglandular density are described and illustrated.
August 20; accepted August 25. All authors have There is no criterion standard for determining breast density, but un-
disclosed no relevant relationships. Address derstanding the pros and cons of the various assessment methods will
correspondence to N.S.W. (e-mail: nicole
[email protected]). allow radiologists to make informed decisions. Radiologists should
See also the article by Freer (pp 302–315) and understand the basic factors involved in breast density assessment, the
the discussion by Butler (pp 324–326) in this changes related to density assessment described in the fifth edition of
issue.
the American College of Radiology Breast Imaging Reporting and Data
System (BI-RADS) lexicon, and the capabilities of currently available
software. Online supplemental material is available for this article.
©
RSNA, 2015 • radiographics.rsna.org

Introduction
Breast density refers to the amount of fibroglandular tissue rela-
tive to fat in the breast. The radiographically opaque fibroglandular
breast tissue is made up of epithelial glandular components, in-
cluding terminal ductal lobular units and ducts, as well as stromal
components, including the supportive fibrous connective tissue
within the inter- and intralobular stroma. The fifth edition of the
American College of Radiology Breast Imaging Reporting and
Data System (BI-RADS) lexicon has modified the previous den-
sity categorization system by eliminating the methods described in
earlier editions that were based on estimated density percentage
quartiles. Rather than using an overall estimate of the percentage
of fibroglandular density, the current BI-RADS revision suggests
using a more subjective four-category overall assessment of breast
composition to convey the potential that a dense area may obscure
a cancer. Breast imagers should assess “the volume of attenuat-
ing tissues in the breast, to help indicate the relative possibility
that a lesion could be obscured by normal tissue” and thus reduce
RG • Volume 35 Number 2 Winkler et al 317

TEACHING POINTS Table 1: Breast Composition Categories (BI-

■■ Rather than using an overall estimate of the percentage of RADS Fifth Edition)
fibroglandular density, the current BI-RADS revision suggests Category A: The breasts are almost entirely fatty
using a more subjective four-category overall assessment of Category B: There are scattered areas of fibroglan-
breast composition to convey the potential that a dense area
dular density (an optional description of a few
may obscure a cancer.
or moderate scattered areas of density can be
■■ Heterogeneously dense breasts and breasts with extremely included in a second sentence)
dense fibroglandular tissue (more than 50% dense tissue,
Category C: The breasts are heterogeneously dense,
according to the BI-RADS fourth edition) are categorized as
“dense,” while breasts with scattered fibroglandular tissue
which may obscure small masses (if the dense
and largely fatty breasts are considered “nondense.” tissue is localized to one area of the breast, the
location of the dense tissue can be included in
■■ Mammography is recommended for all women of screening
age, independent of breast density and breast cancer risk.
a second sentence: “The dense tissue is located
anteriorly in both breasts, and the posterior por-
■■ At mammography, subjective visual assessment is used to
tions are mostly fatty” or “Primarily dense tissue
assign an overall breast composition rating on the basis of
the densest tissue area to convey the likelihood of lesion ob-
is located in the upper outer quadrants of both
scuration. breasts; scattered areas of fibroglandular tissue
are present in the remainder of the breasts”)
■■ If quantitative assessment methods are shown to be reliable
and more accurate, the assignment of breast density, and
Category D: The breasts are extremely dense,
thus part of a woman’s risk profile, would be better served by which lowers the sensitivity of mammography
using quantitative volume density calculations.
Source.—Adapted, with permission, from refer-
ence 1.

mammographic sensitivity (1). The four breast

composition categories described in the fifth with each patient, women with dense breast tis-
edition of the BI-RADS lexicon are listed in sue may be unaware that mammography may
Table 1.The BI-RADS manual also suggests the be less sensitive in their case and that their risk
optional use of additional descriptions of the for breast cancer may be higher than in women
breast density pattern, with sample descriptors with predominantly or almost completely fatty
included in the manual. breasts (3–7). It is further confounding for
patients to understand that these potentially
Clinical serious and alarming implications result from
Implications of Breast Density subjective density assessments and somewhat
By expert consensus, heterogeneously dense arbitrary classifications. The topic of breast den-
breasts and breasts with extremely dense fibro- sity has been propelled to the forefront by recent
glandular tissue (more than 50% dense, according legislation in several states that requires patients
to the BI-RADS fourth edition) are categorized as to be informed of breast density and the po-
“dense,” while breasts with scattered fibroglandu- tential for decreased mammographic sensitivity
lar tissue and largely fatty breasts are considered and increased cancer risk. This legislation is de-
“nondense.” Dense tissue is associated with an signed to empower the patient, foster discussion
elevated risk for breast cancer, although the level between the provider and patient, and possibly
of risk is largely unknown because many factors prompt supplemental screening.
influence tissue density, such as age, endogenous Supplemental screening modalities include
and exogenous hormones, chemotherapy and ra- whole-breast ultrasonography (US) and magnetic
diation therapy, and lactation. Various population- resonance (MR) imaging. In the American Col-
based studies have demonstrated the following lege of Radiology Imaging Network (ACRIN)
approximate distributions of breast densities: 10% 6666 trial, investigators evaluated the added
in density category A, 40% in category B, 40% in cancer detection rate with use of US and MR im-
category C, and 10% in category D. Therefore, aging in women with dense breast tissue and an
about 50% of the screening population has dense elevated (intermediate or high) risk for breast can-
breast tissue (2). cer and reported a supplemental cancer detection
Breast density assessment has been an im- rate of 14.7 cancers per 1000 women with use of
portant component of screening mammography MR imaging (8). However, although MR imaging
reports and conveys information to referring cli- improves the detection of early-stage breast can-
nicians about mammographic sensitivity and the cers, it is associated with an added biopsy rate of
relative risk for breast cancer. However, unless 7%. The additional cancer yield is less impressive
the referring clinician is invested in extracting with use of whole-breast US, which has a supple-
density assessments from mammography reports mental cancer detection rate of 3.7 per 1000 and
and discussing the implications of breast density a low positive predictive value for malignancy of
318 March-April 2015 radiographics.rsna.org

biopsied lesions (7.4%) (8). Supplemental MR percentage of the total breast volume that consists
imaging and whole-breast US are associated with of radiopaque fibroglandular components (Fig 1).
higher costs and higher biopsy rates, and their However, the recently published fifth edition of
utility in average-risk women remains unknown. the BI-RADS manual suggests the use of the same
In many states, adjunctive screening is not cov- category descriptors as part of an overall subjective
ered by health insurance and may result in sub- assessment that is intended to indicate the likeli-
stantial out-of-pocket expenses for patients. hood of lesion obscuration (1,10). For certain pat-
Despite the reduced sensitivity of mammogra- terns of scattered or heterogeneous fibroglandular
phy in women with dense fibroglandular tissue, tissue, the BI-RADS fifth edition suggests that
there is no replacement for mammography as a an additional optional description of the pattern
screening tool. Mammography is recommended of fibroglandular composition be included in the
for all women of screening age, independent of mammography report. Subjective visual assess-
breast density and breast cancer risk. ment is used to assign an overall breast composi-
The distinction between the BI-RADS catego- tion rating on the basis of the densest tissue area
ries of “scattered areas of fibroglandular density” to convey the likelihood of lesion obscuration. For
and “heterogeneously dense” breasts may not example, breasts with fibroglandular tissue that
be clinically relevant for an individual woman; is densely collected in the subareolar region, with
however, it is a branch point for conveying breast otherwise fatty tissue seen posteriorly, are consid-
cancer risk and reduced mammographic sensi- ered heterogeneously dense (even if the overall
tivity to the patient and referring clinician. This density percentage is less than 50%) to convey that
distinction may lead to a clinician-to-patient or this area may obscure a cancer. Figure 2 provides
radiologist-to-patient discussion about the pos- examples of breast composition assessment ac-
sible need for adjunctive imaging and/or risk cording to the BI-RADS fifth edition.
reduction. Most experts agree that breast density Women considered to have dense breasts are
is associated with increased cancer risk, but the those categorized as having heterogeneously dense
degree of this association is unclear because re- (BI-RADS category C) or extremely dense (cat-
search has been limited and is based on imperfect egory D) breasts. One advantage of visual assess-
density assessments without a criterion standard ment over software-based assessment is the lack of
(9). Given its association with a higher risk for additional cost, which is why it is the most widely
breast cancer, breast density may become a com- used method. However, interrater variability exists
ponent of overall risk-assessment models, despite among radiologists, with several studies reporting
limitations in the available methods of obtaining only moderate interobserver agreement (11–13).
objective, reliable, reproducible density mea- Therefore, much effort has been devoted to the
surements and the lack of data for assigning an development of computer-based quantitative and
accurate cancer risk metric to the percentage of objective methods of density measurement.
fibroglandular tissue density.
This article describes current methods of Quantitative Assessment
breast density assessment, including subjective, Quantitative assessment provides a reproducible
semiquantitative, and quantitative methods and calculated density percentage (either the area or
currently available automated software. Although the volume, depending on the method used). The
other methods of density assessment have been density percentage is determined by dividing the
described in the physics literature, discussion of calculated area (or volume) of the fibroglandular
these methods is beyond the scope of this article. tissue by the calculated total breast area (or vol-
ume). The basic difference between area and vol-
Assessment Methods ume density calculations is shown in the Movie.

Subjective and Calculation of Area Density Percentage.—Pla-

Semiquantitative Assessment nimetry is a two-dimensional method for directly
The earliest described subjective categorization measuring area. This method of calculating the
system was a qualitative assessment of fibroglan- area density percentage, described by Wolfe et al
dular breast tissue according to pattern recogni- (14) in 1987, was the first reported quantitative
tion, as described by Wolfe (5) in 1976. Since then, method of breast density assessment. The appli-
other subjective systems have been proposed, but cation of this manual technique to mammogram
by far the most widely used system in the United assessment is shown in Figure 3. The method is
States is the BI-RADS categorization. For well labor intensive, is highly time consuming, and
over a decade, breast imagers have been using does not account for inhomogeneous fibroglan-
the BI-RADS recommendations to assess overall dular density within the region of interest and,
breast tissue density by visually estimating the therefore, can be inaccurate.
RG • Volume 35 Number 2 Winkler et al 319

Figure 1. Basic subjective method

for visual assessment of fibroglan-
dular breast density by quartile.
Mediolateral oblique and cranio-
caudal mammograms are used to
estimate the volume percentage
of fibroglandular tissue (white
shading) within the spherical total
breast volume. In denser areas, the
superimposition of fibroglandular
tissue is also considered. Breasts
with less than 25% density (a),
25%–50% density (b), and 50%–
75% density (c) are shown.

Interactive thresholding is a more recent level thresholding values to digitized mammo-

method for two-dimensional calculation of the grams (ie, the Cumulus technique) and is shown
area density percentage that has been described in Figure 4. The method remains somewhat sub-
by Byng et al (15,16). This semiautomated jective because it requires user input, but it is less
method uses software to apply interactive gray- time consuming than planimetry.
320 March-April 2015 radiographics.rsna.org

Figure 2. Visual assessment of breast density on mediolateral oblique and craniocaudal mammograms, with density rated according
to the fifth-edition BI-RADS categories. (a) Category A, “The breasts are almost entirely fatty.” (b) Category B, “There are scattered
areas of fibroglandular density.” (c) Category B, with the optional descriptors “a few scattered areas of fibroglandular density tissue”
(top) and “moderate scattered areas of fibroglandular density tissue” (bottom). (d) Category C, “The breasts are heterogeneously
dense, which may obscure small masses.” (Fig 2 continues)
RG • Volume 35 Number 2 Winkler et al 321

Figure 2. (continued). (e) Category C, with the optional de-

scriptor “The dense tissue is located anteriorly in both breasts,
and the posterior portions are mostly fatty.” (f) Category C,
with the optional descriptor “Primarily dense tissue is located
in the upper outer quadrants of both breasts; scattered areas
of fibroglandular tissue are present in the remainder of the
breasts.” (g) Category D, “The breasts are extremely dense,
which lowers the sensitivity of mammography.”

tance (18–20). Currently, two U.S. Food and

Drug Administration (FDA)–approved software
programs (Quantra; Hologic, Bedford, Mass;
and Volpara; Matakina International, Wellington,
New Zealand) are available and provide fully au-
tomated volume density percentages calculated
with use of proprietary algorithms. The features
of these programs are compared in Table 2
(17,21). The programs convert the calculated
Calculation of Volume Density Percentage.— volume density percentage to the appropriate
Three-dimensional data obtained from digital BI-RADS density category so that no additional
breast tomosynthesis, MR imaging, com- time is required during mammographic inter-
puted tomography (CT), or US can be used pretation, but the software must be purchased
for straightforward calculation of the volume and may not reflect the recent modifications
density percentage; however, the methods used described in the BI-RADS fifth edition.
are not yet applicable to daily screening prac-
tices (17). Multiple methods for calculating the Method Selection
volume density percentage have been described No criterion standard exists for breast density as-
for use with two-dimensional mammograms. sessment. Radiologists’ visual assessment is known
These methods rely on known or estimated to be influenced by many factors, including per-
measurements of breast thickness at compres- sonal factors such as a tendency to over- or under-
sion, the energy of incident x-rays, the effective estimate, a bias toward defensive practice, fatigue,
x-ray coefficient, and measured x-ray transmit- contextual influences from the particular patient
322 March-April 2015 radiographics.rsna.org

Figure 3. Planimetric assessment

of breast density on a craniocaudal
mammogram. The fibroglandular
tissue is manually outlined on an
overlay (green), and the area of
the fibroglandular tissue is calcu-
lated in square centimeters. The
entire breast is then manually out-
lined on an overlay (orange), and
the area of the total breast is cal-
culated in square centimeters. The
density percentage is calculated
by dividing the area of the fibro-
glandular tissue by the area of the
entire breast.

Figure 4. Interactive thresholding used to calculate breast

density on a craniocaudal mammogram. Images in the middle
columns have a higher threshold value, and images in the right
columns have a lower threshold value. (a) The user selects
threshold gray-level values for fibroglandular tissue (arrows).
Areas with pixels higher than the set threshold are identified as
fibroglandular density in automated area calculations that sum
the included pixels. (b) The software automatically identifies the
skin edge (orange outlines) and calculates the area of the entire
breast. (c) The area density percentage is calculated by dividing
the area of the fibroglandular tissue (white outlines) by the area
of the entire breast (orange outlines).

population in the screening pool, and reading

room conditions (eg, ambient light and worksta-
tion resolution). Therefore, volume density cal-
culations may provide the most accurate, reliable,
and reproducible method of density assessment
and may have more meaningful clinical implica-
tions than subjective density assessments. Further,
if quantitative density assessment methods are titative volume density calculations. However, re-
shown to be reliable and more accurate (given the search is needed to demonstrate the correlation of
known interobserver variability that limits the ac- calculated volume density percentages and breast
curacy of qualitative visual assessment), the assign- cancer risk.
ment of breast density, and thus part of a woman’s Subjective assessment methods will change
risk profile, would be better served by using quan- slightly with use of the updated information in
RG • Volume 35 Number 2 Winkler et al 323

Table 2: Comparison of Quantra and Volpara Software for Automated Quantitative Calculation of
Volume Density Percentages*
Common main components (completely automated): the total breast volume is defined (breast segmentation);
fat or fibroglandular tissue values are defined on the basis of mammographic image parameters to determine
the volume of the fibroglandular tissue
Quantra software: the fibroglandular density per pixel is estimated by using known image-acquisition param-
eters, including breast thickness; the pixel values are then added to determine the volume of the fibroglandular
tissue
Volpara software: a pixel value representing fat is identified automatically by the software to provide a reference
value; individual pixels in the breast are compared with the reference value to determine x-ray attenuation and
generate a density map; the volume of the fibroglandular tissue and the total breast volume are calculated by
adding the corresponding pixel values
* Volume density percentage = Vf/Vt × 100, where Vf is the fibroglandular tissue volume and Vt is the total breast
volume.

5. Wolfe JN. Breast patterns as an index of risk for developing

the BI-RADS fifth edition. This change will likely breast cancer. AJR Am J Roentgenol 1976;126(6):1130–1137.
result in more mammograms being categorized as 6. Yaffe M, Boyd N. Mammographic breast density and cancer
heterogeneously dense when there is a relatively risk: the radiological view. Gynecol Endocrinol 2005;21(suppl
1):S6–S11.
dense collection of tissue that would have been 7. McCormack VA, dos Santos Silva I. Breast density and
considered “scattered areas of fibroglandular den- parenchymal patterns as markers of breast cancer risk: a
sity” according to the previous density percentage meta-analysis. Cancer Epidemiol Biomarkers Prev 2006;15
(6):1159–1169.
quartile system, but there likely will continue to 8. Berg WA, Zhang Z, Lehrer D, et al. Detection of breast can-
be significant interobserver variability. cer with addition of annual screening ultrasound or a single
screening MRI to mammography in women with elevated
Conclusion breast cancer risk. JAMA 2012;307(13):1394–1404.
9. Kopans DB. Basic physics and doubts about relationship
A basic knowledge of the factors involved in between mammographically determined tissue density and
breast composition and density assessment is breast cancer risk. Radiology 2008;246(2):348–353.
10. D’Orsi CJ, Mendelson EB, Ikeda DM, et al. Breast Imaging
important to understand the various assessment Reporting and Data System: ACR BI-RADS. 4th ed. Reston,
methods. Many methods can be used to assess Va: American College of Radiology, 2003.
breast density, ranging from subjective visual 11. Ciatto S, Houssami N, Apruzzese A, et al. Categorizing
breast mammographic density: intra- and interobserver re-
estimation to quantitative calculations made producibility of BI-RADS density categories. Breast 2005;14
according to complex algorithms. Radiologists (4):269–275.
must understand the considerable implications 12. Redondo A, Comas M, Macià F, et al. Inter- and intra­
radiologist variability in the BI-RADS assessment and breast
of diagnosing a woman with dense breast tissue density categories for screening mammograms. Br J Radiol
and should strive to be as consistent as possible 2012;85(1019):1465–1470.
when subjectively assigning a mammographic 13. Timmers JM, van Doorne-Nagtegaal HJ, Verbeek AL, den
Heeten GJ, Broeders MJ. A dedicated BI-RADS training
density category. Radiologists should consider programme: effect on the inter-observer variation among
using the available quantitative assessment screening radiologists. Eur J Radiol 2012;81(9):2184–2188.
methods if these methods are proved to provide 14. Wolfe JN, Saftlas AF, Salane M. Mammographic parenchymal
patterns and quantitative evaluation of mammographic densi-
consistent and reliable measurements of breast ties: a case-control study. AJR Am J Roentgenol 1987;148
density. A review of the BI-RADS breast com- (6):1087–1092.
position categories included in the fifth edition 15. Byng JW, Yaffe MJ, Jong RA, et al. Analysis of mammographic
density and breast cancer risk from digitized mammograms.
and the examples of breast density assessment RadioGraphics 1998;18(6):1587–1598.
included in this article will help improve consis- 16. Byng JW, Boyd NF, Fishell E, Jong RA, Yaffe MJ. The
tency among radiologists. quantitative analysis of mammographic densities. Phys Med
Biol 1994;39(10):1629–1638.
17. Yaffe MJ. Mammographic density: measurement of mam-
References mographic density. Breast Cancer Res 2008;10(3):209.
1. D’Orsi CJ, Mendelson EB, Ikeda DM, et al. Breast Imaging 18. Highnam R, Brady M, Yaffe MJ, Karssemeijer N, Harvey J.
Reporting and Data System: ACR BI-RADS. 5th ed. Reston, Robust breast composition measurement: Volpara. In: Martí
Va: American College of Radiology, 2013. J, Oliver A, Freixenet J, Martí R, eds. Digital mammogra-
2. Lee CI, Bassett LW, Lehman CD. Breast density legislation phy: 10th international workshop, IWDM 2010, Girona,
and opportunities for patient-centered outcomes research. Catalonia, Spain, June 16–18, 2010—proceedings. Berlin,
Radiology 2012;264(3):632–636. Germany: Springer, 2010; 342–349.
3. Boyd NF, Guo H, Martin LJ, et al. Mammographic density 19. Diffey J, Hufton A, Astley S. A new step-wedge for volumetric
and the risk and detection of breast cancer. N Engl J Med measurement of mammographic density. In: Astley SM,
2007;356(3):227–236. Brady M, Rose C, Zwiggelaar R, eds. Digital mammography:
4. Boyd NF, Martin LJ, Yaffe MJ, Minkin S. Mammographic 8th international workshop, IWDM 2006, Manchester, UK,
density and breast cancer risk: current understanding and June 18–21, 2006— proceedings. Berlin, Germany: Springer,
future prospects. Breast Cancer Res 2011;13(6):223. 2006; 1–9.
324 March-April 2015 radiographics.rsna.org

20. van Engeland S, Snoeren PR, Huisman H, Boetes C, Karsse- 21. Ciatto S, Bernardi D, Calabrese M, et al. A first evaluation
meijer N. Volumetric breast density estimation from full-field of breast radiological density assessment by QUANTRA
digital mammograms. IEEE Trans Med Imaging 2006;25 software as compared to visual classification. Breast 2012;21
(3):273–282. (4):503–506.