Sjögren's disease & overlap diseases
Detecting Sjögren’s disease from parotid gland ultrasound radiomics
Authors' conclusion
Radiomics-based machine learning demonstrated high classification performance for distinguishing SjD from healthy controls using parotid gland ultrasonography.
Study context
- Population
- Patients fulfilling the 2016 ACR/EULAR criteria for Sjögren's disease (SjD), healthy controls, non-Sjögren sicca patients and incomplete SjD cases
- Participants
- 202 patients866 parotid US images
- Comparison
- Radiomics-based machine learning on parotid gland US vs conventional visual assessment by radiologists
- Outcome
- Classification performance for SjD vs healthy controls
Methodological detailsradiomics-based machine learning, 5-fold soft-voting SVM ensemble, held-out test set, SHAP analysis for interpretability
- Radiomics-based machine learning
- 5-fold soft-voting SVM ensemble
- Held-out test set
- SHAP analysis for interpretability
Performance
As reported in the abstract
Binary classification SjD vs healthy controls
| Test or model | AUC | Sensitivity | Specificity | Accuracy | Accuracy (two radiologists) |
|---|---|---|---|---|---|
| SVM ensemble | 0.99 | 0.86 | 0.96 | 0.94 | — not stated in the abstract |
| Radiologist assessment | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | 0.62 and 0.72 |
AUC
SVM ensemble0.99
Radiologist assessment— not stated in the abstract
Sensitivity
SVM ensemble0.86
Radiologist assessment— not stated in the abstract
Specificity
SVM ensemble0.96
Radiologist assessment— not stated in the abstract
Accuracy
SVM ensemble0.94
Radiologist assessment— not stated in the abstract
Accuracy (two radiologists)
SVM ensemble— not stated in the abstract
Radiologist assessment0.62 and 0.72
— not stated in the abstract
Further findings
Strongest predictors (SHAP)
- Intensity dispersion metrics, GLCM-based texture features and LBP micro-texture patterns
Non-Sjögren sicca vs confirmed SjD (PCA and feature-level analyses)
- Substantial overlap in radiomic features
Original publication
Detecting Sjögren’s disease from parotid gland ultrasound radiomics. Rheumatology (Oxford). Published 11 August 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →