Speaker
Description
To support clinical translation of propagation-based phase-contrast breast CT (BCT) at the Imaging and Medical Beamline (IMBL) of the Australian Synchrotron, we developed and compared two patient-specific digital breast phantom modelling approaches for Monte Carlo–based mean glandular dose (MGD) estimation. The approaches included: (1) fully voxelated phantoms generated directly from reconstructed experimental BCT scans of whole mastectomy samples, where each voxel preserved local tissue composition and was mapped to tissue types (adipose, glandular, skin) or air, and (2) segmentation-based phantoms, where each tissue type in the BCT images was segmented into non-overlapping masks and assigned a uniform mean density per tissue class. MGD values and air kerma-to-MGD conversion coefficients (DgN) were calculated using the EGSnrc MC code and systematically compared. To assess the impact of phantom spatial resolution and partial-volume effects on MGD, CT image datasets were downsampled by averaging multiple slices to generate new phantoms, introducing tissue mixing and partially simulating segmentation inaccuracies. MGD values from these downsampled phantoms were then compared with those from the original high-resolution models. In addition, both types of patient-specific phantoms were compared with homogeneous models to evaluate the validity of commonly used simplified approaches.
MGD was strongly influenced by breast anatomy and X-ray energy, with higher glandular density decreasing MGD and larger breast volumes increasing it; both digital phantom types consistently captured these trends. The segmentation-based model overestimated DgN by up to 6% compared with the fully voxelated model, while downsampling had negligible impact. In contrast, homogeneous phantoms underestimated MGD by up to 35%, due to neglect of glandular tissue distribution.
This study demonstrates that segmentation-based phantoms overestimated DgN, while moderate segmentation inaccuracies and partial-volume effects have minimal impact on dose calculations. Homogeneous phantoms introduce significant underestimation, highlighting the importance of accounting for patient-specific anatomical heterogeneity for accurate breast CT dosimetry.