AI matches expert accuracy for traumatic femur fracture detection: meta-analysis of 37 studies

In a meta-analysis of 37 studies (95,148 femur images), AI detection of traumatic femur fractures had pooled sensitivity 0.94, specificity 0.94, AUC 0.98—similar to human experts (0.91, 0.93, 0.97). AI augmentation raised non-expert AUC from 0.95 to 0.97.

• Expert readers (orthopedic surgeons, radiologists) achieved an AUC of 0.989 unaided and 0.99 with AI augmentation; non-experts improved from 0.95 to 0.97 with AI assistance.

Automated summary

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