Mammographic breast density is a strong risk factor for breast cancer. Studies on imaging dose in mammography
have primarily focused on imaging quality and diagnostic accuracy, while little work has been done on
understanding its effect on the estimation of breast density. Studies on the effect of dose on mammographic density
estimation can be useful in dose reduction for the purpose of density estimation and monitoring. In this study, we
investigate the dependence of percent area (PD%) and volumetric (VD%) breast density estimation on imaging dose
using an anthropomorphic breast phantom (Rachel, Gammex). A set of digital mammograms were obtained with a
GE Senographe 2000D FFDM system, using 220 unique combinations of different imaging physics, namely
target/filter, kVp and mAs. Specifically, 8 different mAs settings were defined as corresponding to 10%, 20%, 40%,
70%, 100%, 150%, 200% and 300% of 1.8 mGy reference average glandular dose (AGD) for standard phototimed
exposure. Breast density was estimated using fully-automated FDA-cleared software (Quantra v.2.0, Hologic Inc.).
The obtained estimates were analyzed to study the effect of the imaging dose, using ANOVA and linear regression.
Results show that there is a statistically significant dependence of density estimation on x-ray imaging dose
(p-value=0.014 and <0.001 for PD% and VD%, respectively), while the actual variation of the estimation across the
different levels of dose is relatively low (standard deviation of 2.87% and 0.66% for PD% and VD% respectively),
the differences could be significant when breast density measures are used for risk estimation.