Machine learning-based multiclass model for autoimmune disease diagnosis and classification through nailfold videocapillaroscopy features
Authors' conclusion
This study developed a robust multiclassification model for differentiating autoimmune diseases using NVC features.
Study context
- Population
- Controls and patients with rheumatoid arthritis (RA) or systemic lupus erythematosus (SLE) with nailfold videocapillaroscopy (NVC) images
- Participants
- 600 NVC images
- Comparison
- NVC-feature model distinguishing controls, RA and SLE
- Controls (Ctr)n = 117reference
- RAn = 337
- SLEn = 146
- Outcome
- Discrimination of Ctr, RA and SLE by NVC features (macro AUC)
Methodological detailsmachine learning, eXtreme Gradient Boosting multiclassification model, SHAP analysis, training and test sets at a 7:3 ratio
- Machine learning
- eXtreme Gradient Boosting multiclassification model
- SHAP analysis
- Training and test sets at a 7:3 ratio
Performance
As reported in the abstract
Macro AUC of the multiclass model
Macro AUC, training set
0.96
Macro AUC, test set
0.80
Further findings
NVC features differing among groups
- NVC features with significant differences among the groups (of 9)
- 7
- Key features by SHAP: papilla shape, RBC aggregation, number of capillary loops, number of crossed capillary loops, subpapillary venous plexus (SVP)
Group-specific NVC patterns
- Ctr: papilla shape key feature, correlated with neutrophils, white blood cells and monocytes
- RA: number of crossed capillary loops most prominent, correlated with ESR, C3 and C4 and inversely with IgG
- SLE: SVP dominant feature, effectively distinguished SLE from Ctr and RA
Original publication
Machine learning-based multiclass model for autoimmune disease diagnosis and classification through nailfold videocapillaroscopy features. RMD Open. Published 4 March 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →