ML Detects Wrist Fractures on Radiographs with Accuracy Matching Experts

Machine learning detection of wrist fractures on radiographs achieved pooled AUCs of 0.88 (scaphoid), 0.97 (distal radius), and 0.93 (unspecified), comparable to expert AUCs of 0.88 and 0.98.

• Across 48 included algorithms, ML showed high accuracy for distal radius (AUC 0.97) and unspecified wrist fractures (AUC 0.93), and moderate accuracy for scaphoid fractures (AUC 0.88).
• Expert performance was similar, with AUC 0.98 for distal radius and 0.88 for scaphoid fractures.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.