Radiomics grading of hand osteoarthritis severity using standard radiographs: Results from the DIGICOD cohort
Authors' conclusion
The model showed good performance detecting structural damage, supporting radiomics as a potential tool to reduce reliance on subjective visual grading.
Study context
- Population
- Patients with hand OA, DIGICOD cohortradiographs analysed
- Participants
- 379 patients
- Comparison
- Radiomics analysis of standard hand radiographs vs Kellgren-Lawrence (KL) scores
- Outcome
- Detection of structural involvement (KL ≥2), severe disease (KL 3-4) and multiclass KL grades at the joint level
Methodological detailsretrospective, baseline data from the DIGICOD cohort, semi-automatic U-Net segmentation with manual correction, Random Forest classifiers, joint-level analysis, 80%/20% stratified training/test split
- Retrospective
- Baseline data from the DIGICOD cohort
- Semi-automatic U-Net segmentation with manual correction
- Random Forest classifiers
- Joint-level analysis
- 80%/20% stratified training/test split
Performance
As reported in the abstract
Detection of structural involvement (KL ≥2)
AUC
0.81
95% CI 0.79–0.83
Sensitivity (high)
89%
Specificity (low)
59%
Further findings
Severe disease detection (KL 3-4)
- AUC
- 0.83 [95% CI: 0.81-0.85]
- Sensitivity (very good)
- 80%
- Specificity (good)
- 70%
Multiclass KL prediction (0-1, 2, 3, 4)
- Macro-averaged AUC
- 0.76
- Accuracy
- 56%
- Lower performance, reflecting challenges in distinguishing intermediate (KL 2-3) and severe (KL 4) grades
Original publication
Radiomics grading of hand osteoarthritis severity using standard radiographs: Results from the DIGICOD cohort. Osteoarthritis Cartilage. Published 21 May 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →