DBT-DINO, a foundation model for digital breast tomosynthesis (DBT) pretrained on 487 975 volumes from 27 990 patients, demonstrated the potential of self-supervised pretraining in DBT analysis.