Patient-Specific Normalized Glandular Dose Range Estimate For Mammography

Breast cancer is the most common cancer among women in the United States, and early detection is essential for reducing mortality. Screening mammography plays an important role in early detection by enabling identification of malignancies at earlier, more treatable stages. However, currently accepted dose estimation methods rely on simplified assumptions, typically modeling the breast as a homogeneous 50/50 mixture of adipose and fibroglandular tissue. Such assumptions neglect the substantial spatial heterogeneity of glandular tissue within individual breasts, most of which contain well below 50% FG tissue, and may therefore lead to inaccurate estimates of normalized glandular dose (DgN). In this work, we first show via Monte-Carlo Simulations the range of the DgN possible by just placing the fibroglandular tissue in the top, center or bottom for a range of breast models with GF varying sizes and glandular fraction percent. Then we propose a patient-specific framework for estimating a range of normalized glandular dose (DgN) from a (single) mammographic projection and its corresponding glandular fraction (GF) map, derived by our prior method on GF estimation (Smith, Dey et al) from a single image. We demonstrate that DgN can vary by up to a factor of three depending solely on differences in depth distribution of fibroglandular tissue, even when resultant projection images are indistinguishable from one another. Using simulated projections and GF maps, a Siddon ray-tracing-based back projection algorithm was applied to generate volumetric breast reconstructions that yielded the minimum and maximum achievable DgN while preserving both the projection image and GF distribution. We analytically demonstrate that the dose to these back projected volumes can be minimized or maximized by altering the placement of fibroglandular tissue within the volume. Monte Carlo simulations including Compton scatter were then performed, and the resulting doses were normalized by entrance air kerma to obtain DgN estimates for each reconstructed configuration. The results demonstrate that realistic bounds on patient dose can be derived from limited imaging information, supporting more personalized and transparent mammographic dose assessment.