A multi-reader multi-case study comparing Breast Cancer Reading Efficiency and Accuracy in Screening with Tomosynthesis versus AI-assisted mammography

OBJECTIVES: Full-field digital mammography (FFDM) is the standard for breast cancer screening but has reduced sensitivity in dense breasts. Digital breast tomosynthesis (DBT) improves cancer detection compared to 2D mammography but is limited by cost and longer reading times. Artificial intelligence (AI)-enhanced mammography has demonstrated improved sensitivity compared to 2D mammography. This study compares AI-enhanced 2D mammography (DM+AI) with DBT in recalling interval breast cancers (IBCs). METHODS: : A multi-reader, multi-case crossover study was conducted using the Emory BrEast Imaging Dataset, which includes paired FFDM and DBT images with biopsy-proven outcomes. Six radiologists reviewed 53 IBC and 53 screen-detected cancer (SDC) cases, with random assignment to DM+AI or DBT, followed by crossover after a washout period. Primary outcomes included recall rate of IBCs. Secondary outcomes included recall rate of SDCs, interpretation time, and subgroup analysis by breast density. RESULTS: : For IBCs, DM+AI and DBT demonstrated similar recall rates (41.5% vs 40.9%) and detection rates (14.8% vs 11.9%). DM+AI identified more non-invasive (31.5% vs 13.0%) and mass-forming (61.0% vs 21.8%) cancers. For SDCs, recall and detection rates were comparable across modalities. In fatty breasts, recall was higher with DM+AI (91.7% vs 50.0%), while DBT performed better in extremely dense breasts. DM+AI enabled faster interpretation, averaging 24.1 seconds less per case. CONCLUSION: : AI-enhanced 2D mammography demonstrated comparable performance to DBT with shorter interpretation times, supporting a stratified screening approach. ADVANCES IN KNOWLEDGE: This study is the first direct comparison of AI-enhanced mammography and digital breast tomosynthesis for IBC detection.