Meta-analysis finds X-ray deep learning models show high accuracy for ankle and foot fracture detection
A systematic review and meta-analysis evaluated the diagnostic performance of X-ray-based deep learning models for detecting ankle and foot fractures. The analysis included 14 studies from 506 reviewed, though specific population characteristics and clinical settings were not reported. The intervention involved deep learning models analyzing X-ray images, with no specific comparator detailed.
The primary outcome was diagnostic accuracy. The cumulative pooled sensitivity was 93.2% (95% CI 88.8-95.9%) and specificity was 94.5% (95% CI 90.1-97.0%), with low heterogeneity (I² 6.6%). The pooled F1 score was 0.94 (95% CI 0.88-0.97). Subgroup analyses revealed models detecting calcaneal fractures performed significantly better, with sensitivity of 95.1% (95% CI 93.0-96.6%), specificity of 98.3% (95% CI 97.00-99.0%), and a diagnostic odds ratio of 1751.8 (95% CI 445.9-6882.5). Performance did not differ significantly between models using multi-view versus single-view X-rays or based on dataset type.
Safety and tolerability data were not reported. Key limitations include the need for larger sample studies, external validation, and clinical implementation research. The analysis did not compare model performance to a specific clinical standard or radiologist performance, and the population and clinical setting were not described. The results are based on aggregated study data rather than direct clinical application.
For practice, this meta-analysis suggests deep learning models can be considered a notably accurate method for detecting ankle and foot fractures, particularly calcaneal fractures. However, clinicians should interpret these findings cautiously as they represent diagnostic accuracy associations from aggregated research, not evidence of clinical effectiveness or superiority to standard care. Further validation in real-world clinical settings is required before considering implementation.
View Original Abstract ↓
Meta-analysis finds 87% implant survivorship after 10 years in patients under 50 undergoing TKA
Young Knee Pain Has a New Solution
Related in Orthopedics & Sports Medicine
From Other Specialties
