A 3D-printed anthropomorphic breast phantom with realistic anatomy and attenuation properties for spatially resolved...

Objective.Accurate assessment of radiation dose in mammography requires physical phantoms that capture realistic breast anatomy and spatial dose variation. However, most commercially available phantoms are anatomically simplified, which limits their ability to reproduce heterogeneous tissue distributions and yields potential inaccuracies in internal dosimetry. This study presents a reproducible digital-to-physical workflow for generating 3D-printed anthropomorphic breast phantoms to enable spatially resolved dosimetry in mammography.Approach.We fabricated a dual-material phantom by fused deposition modeling with 0.15 mm layer height and 100% infill. The digital geometry was based on a detailed Chinese breast model with 50% glandularity and 50 mm compressed thickness. The effective linear attenuation coefficients of nine materials were measured. Black polyamide was assigned to fibroglandular and skin tissues, while gray acrylonitrile-butadiene-styrene was used for adipose tissue due to their similar attenuation properties. The phantom was irradiated on Hologic Selenia Dimensions digital mammography system with a W Rh^-1target/filter combination at 28 kVp. Three-dimensional dose distributions were measured and compared with Monte Carlo simulation results to assess dosimetric consistency. Additionally, dose distributions in a CIRS 50/50 homogeneous phantom with the same compressed thickness and nominal 50% glandularity equivalence were measured under identical conditions to isolate the influence of anatomical heterogeneity on dose estimation.Main results.The variation coefficients for material uniformity and reproducibility were less than 0.5% and 3%, respectively, demonstrating the potential of the materials for phantom fabrication. With a tube load of 360 mAs, point-specific glandular doses ranged from 0.915 to 18.6 mGy. Strong agreement was observed between measured and simulation dose distributions, with most points within 5% and a maximum deviation of 6.4%. The layer-averaged dose measurements from the anthropomorphic phantom differed by less than 3% from the Monte Carlo results, confirming accurate dose quantification. In contrast, the CIRS 50/50 homogeneous phantom systematically underestimated glandular dose, with deviations up to 15% at 50 mm depth.Significance.This work establishes a validated digital-to-physical workflow that links parameterized breast models with tissue-equivalent phantoms. It enables scalable and anatomically realistic phantom generation for spatially resolved mammographic dosimetry. The framework provides a standardized basis for developing diverse phantom libraries, optimizing dose assessment, and advancing individualized as well as population-based breast imaging protocols.