Assessment of a deep-learning system for fracture detection in musculoskeletal radiographs

Missed fractures are the most common diagnostic error in emergency departments and can lead to treatment delays and long-term disability. Here we show through a multi-site study that a deep-learning system can accurately identify fractures throughout the adult musculoskeletal system. This approach may have the potential to reduce future diagnostic errors in radiograph interpretation.