Initial Explorations of Applying Human Factors Principles to a BI-RADS CAD System

Computer-aided detection (CAD) systems are widely implemented to improve radiologists’ sensitivity. As breast imaging advances, evaluating how CAD aids interact with digital breast tomosynthesis (DBT) and global classifications, such as the Breast Imaging Reporting and Database System (BI-RADS), to impact performance is essential. This study compared five CAD implementations: analog CAD (aCAD), original BI-RADS (oBI-RADS), oBI-RADS with aCAD, novel BI-RADS (nBI-RADS), and nBI-RADS with aCAD. The nBI-RADS CAD system incorporated intuitive colors, less clutter, and holistic guidance. Participants judged whether pseudo-DBT sets contained “cancer” with and without the assistance of their assigned CAD aid. Measures of trust in automation, usability, and workload were also assessed. Results suggest CAD aids may improve sensitivity, speed, and workload. Differences between implementations, including BI-RADS-integrated versions, were not statistically significant on any measure. These findings support CAD use with DBT while indicating more research is needed to determine more effective automated BI-RADS systems.