Interpretable machine learning integrates multi-source biomarkers for osteoarthritis diagnosis and mechanistic insights: A temporomandibular joint model
Authors' conclusion
This study presents a clinically useful, explainable AI model for OA diagnosis. Key predictors and cross-domain interactions improved accuracy and clarified early disease mechanisms.
Study context
- Population
- Participants with temporomandibular joint osteoarthritis (OA) and age- and sex-matched controlscase-control
- Participants
- 162 participants
- Comparison
- Temporomandibular joint (TMJ) OA vs age- and sex-matched controls
- TMJ OAn = 81
- Controlsn = 81reference
- Outcome
- Diagnostic performance (AUC) of an ML model integrating clinical, high-resolution imaging (radiomics, trabecular architecture, joint space) and serum and saliva biomarkers
Methodological detailsage- and sex-matched, nested 10-fold cross-validation, interpretable machine learning (SHAP)
- Age- and sex-matched
- Nested 10-fold cross-validation
- Interpretable machine learning (SHAP)
Main finding
As reported in the abstract
Diagnostic performance of final ensemble model
| Group | AUC | 95% CI |
|---|---|---|
| Final ensemble model | 0.828 | 0.757-0.892 |
Final ensemble model
AUC0.828
95% CI0.757-0.892
Observational data: associations, not effects.
Further findings
Top SHAP predictors
- Top predictors included headache severity, trabecular thickening, restless sleep, muscle soreness, limited mouth opening and joint space narrowing
Mechanistic interactions (SHAP)
- Radiomics-cartilage degradation links (e.g., condyle grey level nonuniformity with saliva CXCL-16)
- Clinical-molecular associations (e.g., headaches with saliva VE-cadherin)
- Subchondral microstructure correlations (e.g., grey level nonuniformity with run length nonuniformity)
Original publication
Interpretable machine learning integrates multi-source biomarkers for osteoarthritis diagnosis and mechanistic insights: A temporomandibular joint model. Osteoarthritis Cartilage. Published 14 August 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →