AI-supported Approaches for Mammography Single and Double Reading: A Controlled Multireader Study

Key Finding: AI support significantly improved sensitivity across all mammography reading approaches, especially benefiting radiologists with lower baseline sensitivity, which rose from 56.7% in the human single reading to 89.5% (p<0.001) in the human-AI double reading. In simulated double reading, AI as an independent second reader further improved sensitivity without affecting specificity. Importance: This finding emphasizes the potential of AI to improve mammography reading diagnostic performance, particularly supporting radiologists with comparatively worse baseline sensitivity.