Assessment of anatomical heterogeneity effects on X-ray breast dosimetry using MCNP.

Mammography is an essential tool for breast cancer screening, yet the accuracy of Mean Glandular Dose (MGD) estimation is often limited by conventional dosimetric models that rely on homogeneous phantoms. These models oversimplify the complex breast anatomy and, as shown in this work, may significantly underestimate the glandular dose. This study addresses this limitation by establishing a novel computational workflow that integrates the FDA's open-source VICTRE pipeline with the advanced transport capabilities of the MCNP 6.3 code. Anatomically realistic heterogeneous breast models were generated using VICTRE Breast Phantom and VICTRE Breast Compress tools to simulate clinically accurate biomechanical compression. Normalized glandular dose coefficients (DgN) were computed and compared against standard homogeneous models (Dance; AAPM TG-195) and custom homogeneous equivalents. Simulations covered compressed breast thicknesses from 3.5 to 6.5 cm, glandularities from 10% to 80%, and beam qualities with half-value layers (HVLs) between 0.30 and 0.60 mmAl. Validation results for the homogeneous setups demonstrated excellent agreement with published data, with mean deviations below 2%. Crucially, the heterogeneous models yielded consistently higher DgN values compared to their homogeneous counterparts. The homogeneous models underestimated the dose significantly, with the DgN ratio falling as low as 0.582 (Model 3) for 6.5 cm breasts at 0.60 mmAl, indicating a discrepancy of approximately 42%. These findings underscore the fundamental importance of anatomical realism for improving MGD estimation accuracy, suggesting that current protocols may require refinement to better account for tissue heterogeneity in radiological risk assessment.