Related Experiment Video
Updated: Sep 28, 2025
A Method to Estimate Cadaveric Femur Cortical Strains During Fracture Testing Using Digital Image Correlation
Published on: September 14, 2017
Artificial Intelligence in Fracture Detection: A Systematic Review and Meta-Analysis
Rachel Y L Kuo1, Conrad Harrison1, Terry-Ann Curran1
1From the Nuffield Department of Orthopedics, Rheumatology and Musculoskeletal Sciences, Botnar Research Centre, Old Road Headington, Oxford OX3 7LD, UK (R.Y.L.K., C.H., M.S., G.S.C., D.F.); Department of Plastic Surgery, John Radcliffe Hospital, Oxford, UK (T.A.C., A.F.); Department of Vascular Surgery, Royal Berkshire Hospital, Reading, UK (B.J.); Department of Plastic Surgery, Stoke Mandeville Hospital, Aylesbury, Buckinghamshire UK (D.C.); and UK EQUATOR Center, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford Centre for Statistics in Medicine, Oxford UK (G.S.C.).
Artificial intelligence (AI) and clinicians show comparable diagnostic performance in fracture detection. This systematic review suggests AI holds promise as a valuable adjunct in clinical practice for identifying fractures.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Fracture misdiagnosis is a concern in emergency radiology.
- Artificial intelligence (AI) is increasingly explored for fracture detection support.
- Existing studies require systematic evaluation of AI's diagnostic performance.
Purpose of the Study:
- To systematically review and meta-analyze the diagnostic performance of AI versus clinicians for fracture detection.
- To include both peer-reviewed literature and gray literature in the comparison.
- To assess AI's potential as an adjunct to human diagnosis.
Main Methods:
- Systematic search of electronic databases (Jan 2018-June 2021).
- Inclusion of studies developing/validating AI for fracture detection across imaging modalities (excluding segmentation).
- Meta-analysis using a hierarchical model; risk of bias assessed with PROBAST checklist.
Main Results:
- 42 studies (115 tables, 55,061 images) analyzed; 37 on radiographs, 5 on CT.
- Pooled sensitivity: AI 92%, Clinicians 91% (internal); AI 91%, Clinicians 94% (external).
- Pooled specificity: AI 91%, Clinicians 92% (internal); AI 91%, Clinicians 94% (external). No significant differences found.
Conclusions:
- Artificial intelligence and clinicians demonstrate comparable diagnostic performance in fracture detection.
- AI shows promise as a diagnostic adjunct in clinical settings.
- Risk of bias and fracture type were identified as sources of heterogeneity.
