Sjögren's disease & overlap diseases
Machine learning to classify the focus score and Sjögren's disease using digitalised salivary gland biopsies: a retrospective cohort study
Authors' conclusion
This study showed that deep learning can reliably classify the focus score and Sjögren's disease using minor salivary gland biopsy exclusively. The study identified that CD8+ T-cell infiltration in acini was associated with Sjögren's disease.
Study context
- Population
- Participants with sicca without SjD and patients with SjD (ACR-EULAR 2016 criteria) with focus score ≥1 or <1, six expert centres (UK, Greece, Portugal, France), European H2020 NECESSITY consortium490 (90%) female
- Participants
- 545 participants
- Outcome
- AUROC to classify the focus score and SjD
Methodological detailsmulticentre (six expert centres), deep learning model trained on five centres, externally validated on the sixth, Shapley values for explainability, Conducted Oct 13, 2021, to Sept 5, 2024
- Multicentre (six expert centres)
- Deep learning model trained on five centres, externally validated on the sixth
- Shapley values for explainability
- Conducted Oct 13, 2021, to Sept 5, 2024
Main finding
As reported in the abstract
Area under the receiver operator curve (AUROC), focus score classification
0.88
95% CI 0.82–0.94
AUROC, Sjögren's disease classification
0.89
0.82–0.94
Observational data: associations, not effects.
Further findings
Sjögren's disease classification in anti-Sjögren's syndrome-related antigen A negative patients (subgroup)
- Performance of Sjögren's disease classification
- 0.92 (0.87-1.00)
Histological patterns identified by the model
- New pattern of CD8+ T cells around acinar epithelial cells associated with Sjögren's disease diagnosis
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →