Related Experiment Video
Updated: Aug 13, 2026
07:25
Ultrasonographic Evaluation of Salivary Glands for Sjogren's Syndrome: Diagnostic and Monitoring Insights
Published on: October 13, 2023
Detecting Sjögren's Disease from Parotid Gland Ultrasound Radiomics
Gamze Akkuzu1, Omer Faruk Durugol2, Sena Tolu3
1Department of Rheumatology, Başakşehir Çam and Sakura City Hospital, Istanbul, Türkiye.
Rheumatology (Oxford, England)
|August 12, 2026
Summary
Radiomics machine learning accurately identifies Sjögren's disease (SjD) using parotid gland ultrasound, outperforming human assessment. This quantitative approach offers an objective tool for SjD diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Rheumatology
Background:
- Salivary gland ultrasonography is a non-invasive tool for Sjögren's disease (SjD) evaluation.
- Current diagnostic utility is limited by operator dependency.
Purpose of the Study:
- To evaluate radiomics-based machine learning for Sjögren's disease classification using parotid gland ultrasonography.
- To compare its performance against conventional visual assessment by radiologists.
Main Methods:
- 104 radiomic features were extracted from 866 parotid gland ultrasound images of SjD patients, healthy controls, and other sicca patients.
- A support vector machine (SVM) ensemble was trained and tested, with SHAP analysis for interpretability.
Main Results:
- The SVM ensemble achieved an AUC of 0.99, outperforming radiologist assessments (AUCs 0.62-0.72).
- Key predictors included intensity dispersion, GLCM texture, and LBP micro-texture features.
- Significant overlap in radiomic features was observed between non-Sjögren sicca and SjD patients.
Conclusions:
- Radiomics-based machine learning shows high performance in distinguishing SjD from healthy controls via ultrasound.
- Quantitative ultrasound analysis can serve as an objective tool for SjD assessment.
- Further validation in larger, multicenter cohorts is necessary.