To evaluate the existing data on the use of AI systems, specifically convolutional neural networks, for detecting scaphoid fractures on wrist radiographs and assess their diagnostic accuracy.
Key Findings:
Nine studies met the inclusion criteria, with sample sizes ranging from 356 to 11,838 images.
AI algorithms demonstrated improved sensitivity for detecting scaphoid fractures compared to traditional radiographic methods.
The AUC values indicated varying levels of diagnostic accuracy, with some models achieving excellent performance (AUC range: X to Y).
Interpretation:
AI has the potential to significantly enhance the detection of scaphoid fractures on wrist radiographs, addressing the limitations of traditional methods.
Limitations:
The included studies were all retrospective, which may introduce bias affecting the reliability of the findings.
Conclusion:
AI systems show promise in improving the detection rates of scaphoid fractures on X-rays, warranting further research, particularly prospective studies, and potential clinical integration.
