DBT-DINO: Toward Foundation Model-Based Analysis of Digital Breast Tomosynthesis

The authors aimed to develop and evaluate a foundation model for digital breast tomosynthesis (DBT), named DBT-DINO, and to assess the impact of domain-specific pretraining on various clinical tasks. The study found that DBT-DINO outperformed the baseline model in breast density classification but showed no significant differences in 5-year breast cancer risk prediction or lesion detection, indicating the need for further refinement in domain-specific pretraining for these tasks.