Enhanced fracture detection on radiographs with AI assistance for clinicians: a systematic review and meta-analysis.
Han Qin, Yunxia Ding, Jiangyi Ju et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis assessed whether AI assistance improves clinicians’ detection of fractures on radiographs. It included 26 diagnostic accuracy studies published between January 1, 2010, and October 10, 2025, covering extremity and trunk fractures. With AI assistance, pooled clinician sensitivity increased from 77% (95% CI: 72-81) to 87% (95% CI: 83-90), and specificity increased from 88% (95% CI: 85-90) to 92% (95% CI: 89-94). The corresponding AUC rose from 0.90 (95% CI: 0.87-0.92) to 0.95 (95% CI: 0.93-0.97).
Key findings
- Pooled sensitivity increased from 77% (95% CI: 72-81) without AI to 87% (95% CI: 83-90) with AI assistance.
- Pooled specificity increased from 88% (95% CI: 85-90) to 92% (95% CI: 89-94) with AI assistance.
- AUC increased from 0.90 (95% CI: 0.87-0.92) before AI assistance to 0.95 (95% CI: 0.93-0.97) after assistance.
- Eight studies were rated at high risk of bias; heterogeneity was associated with factors including fracture location, AI model type, and reference standards.
Why it matters
Missed or incorrect fracture interpretations can affect subsequent assessment and care. Across the included diagnostic accuracy studies, AI assistance was associated with improved pooled sensitivity, specificity, and overall discrimination, suggesting potential value as an aid to radiograph interpretation. However, variation between studies and risk of bias mean the pooled estimates should be interpreted in context.
What to keep in mind
Eight studies were rated as high risk of bias, and results showed heterogeneity related to fracture location, AI model type, and reference standards.