AAPM/EFOMP TG282 versus dance dosimetry for digital breast tomosynthesis in the assessment pathway of a breast...

Within breast screening programs, digital breast tomosynthesis (DBT) can play a key role in the assessment pathway for screen-recalled abnormalities, where women may undergo multiple or additional imaging views, making accurate characterization of mean glandular dose (MGD) particularly important. Breast dosimetry has traditionally been based on the Dance model, with the more recent AAPM/EFOMP TG282 framework offering a more anatomically realistic alternative.

This study aimed to compare MGD estimates derived using the Dance and TG282 methodologies across a large dataset of DBT examinations performed in a breast screening service. In addition, the study compared breast density metrics from the DBT-imaged population with the breast density assumptions of the TG282 model and evaluated the relationship between breast density and MGD estimates.

This retrospective study analyzed 7,142 DBT images acquired between 2022 and 2025 from Siemens and Hologic mammography systems installed at assessment sites within the breast screening program in Queensland, Australia. MGD was estimated for each DBT exposure using the Dance and TG282 methodologies and evaluated as a function of compressed breast thickness (CBT). Statistical comparisons were performed using the Wilcoxon signed-rank test. Results were additionally stratified by vendor model to explore potential system-dependent effects. Volumetric breast density (VBD) was estimated using Volpara, and median values were evaluated as a function of CBT. The relationship between VBD and MGD was assessed within a representative CBT range using Spearman correlation analysis. TG282 MGD estimates were additionally recalculated using exposure-specific VBD values and compared with estimates derived using model-defined median VBD.

Across the full dataset, TG282 yielded MGD values that were, on average, 8.98% lower than those derived using the Dance methodology for craniocaudal (CC) views and 7.91% lower for mediolateral oblique (MLO) views (p < 0.001). The magnitude of difference between MGD estimates showed dependence on CBT and system-specific beam quality. In addition, the study population exhibited higher median VBD values than those assumed by the TG282 reference breast model. The influence of VBD on MGD estimates varied by system, with stronger dependence observed for systems with softer beam quality. Recalculation of TG282 MGD using exposure-specific VBD resulted in slightly lower estimates, further increasing the relative difference compared with Dance-derived values to approximately -10%.

Differences between TG282 and Dance MGD estimates depend on CBT and system beam quality, with TG282 producing lower estimates overall. The higher breast density observed in the DBT assessment population and its system-dependent influence on MGD estimates highlight the importance of accounting for both population breast density characteristics and mammography system mix when comparing DBT MGD estimates across populations.