MO‐AB‐BRA‐05: Monte Carlo Simulation of Glandular Breast Dose in Mammography Using Breast CT‐Derived Glandular Distributions

Abstract

Purpose:To determine the effect of breast CT‐derived fibroglandular distributions on glandular dose in mammography.Methods:Previously‐reported 3D glandular distributions, obtained from a large cohort of breast CT (bCT) data sets (N=219), were fit to bi‐Gaussian functions and used as probability density maps to assign glandular distributions within compressed breast models. Using a conventional mammography geometry, MCNPX was used to simulate monoenergetic normalized glandular dose “DgN(E)” values in compressed phantoms composed of either a homogeneous mixture of glandular and adipose tissue or heterogeneously distributed glandular tissue. The DgN(E) values were weighted by mammographic x‐ray spectra to produce polyenergetic DgN (pDgN) values for the homogeneous (pDgNhomo) and heterogeneous (pDgNhetero) tissue compositions. This work quantified the effect of permutations in glandular fraction, phantom size, x‐ray technique, and location of glandular distributions on pDgN values.Results:Averaged across all phantom sizes and glandular fractions, pDgNhetero values were 30.1% and 21.6% lower than the pDgNhomo values for the Mo and W x‐ray spectra, respectively, when the heterogeneous distributions were centered within the breast phantom. Displacement of the glandular distributions in the superior and inferior directions resulted in pDgNhetero values 5.7% and 49.4% lower than pDgNhomo values, respectively, for Mo x‐ray spectra. Lateral displacement of the glandular distributions resulted in no notable difference in pDgN values.Conclusion:This study demonstrates that glandular dose values in mammography are 20 to 30% lower than assumed using the homogeneous approximation resulting from overestimation of glandular tissue at the breast entrance where the dose deposition is higher. The sensitivity analysis also demonstrates that the differences depend on likely perturbations in the glandular distributions. While these results are robust because the glandular distributions are modeled from a large cohort of bCT data sets, future work will focus on improved quantification of the glandular distributions based on breast size and glandular fraction.