TOPLINE:

A neural-network artificial intelligence (AI) model trained on digitized minor salivary gland biopsy images showed high performance in classifying focus scores and Sjögren disease and identified a new T‑cell infiltration pattern associated with the disease.

METHODOLOGY:

• Researchers conducted a retrospective cohort study between October 2021 and September 2024 to assess whether a machine learning model could classify focus scores and Sjögren disease and to identify new histologic patterns linked to the disease.
• They included 545 participants (mean age, 54.2 years; 90% women) from six European expert centers and analyzed digitized minor salivary gland biopsy slides stained with hematoxylin and eosin, with saffron added for samples from France.
• Focus score was defined as the number of lymphocytic aggregates of ≥ 50 cells infiltrating the gland tissue per 4 mm² of surface. Among the participants, 243 had Sjögren disease with a focus score ≥ 1, 113 had Sjögren disease with a focus score < 1, and 189 control participants had non-Sjögren sicca with no autoimmunity.
• The AI model converted selected biopsy patches into digital summaries to create patient‑level profiles; a neural-network classifier was trained on profiles from five centers and externally validated using slides from the sixth center.
• The primary outcome was the area under the receiver operating characteristic curve (AUROC) for classifying focus score and Sjögren disease.

TAKEAWAY:

• After external validation, the model classified focus score positivity with an AUROC of 0.88 (95% CI, 0.82-0.94), with a specificity of 0.82 (95% CI, 0.72-0.91) and a sensitivity of 0.74 (95% CI, 0.63-0.85).
• For Sjögren disease classification, the model achieved an AUROC of 0.89 (95% CI, 0.82-0.94), with a specificity of 0.91 (95% CI, 0.82-0.98) and a sensitivity of 0.66 (95% CI, 0.55-0.76).
• In patients who tested negative for antibodies to Sjögren syndrome-related antigen A, the model predicted Sjögren disease with an AUROC of 0.92 (95% CI, 0.87-1.00).
• Researchers identified a novel pattern of CD8-positive T cells localized around acinar epithelial cells that was associated with the diagnosis of Sjögren disease.

IN PRACTICE:

“By sophisticatedly implementing the use of AI into Sjogren disease, this work marks a pivotal advancement, redefining the future landscape of diagnostics and healthcare in Sjögren disease,” experts wrote in an accompanying comment.

SOURCE:

This study was led by Julien Duquesne, MSc, Scienta Lab, Paris, France. It was published online on September 29, 2025, in The Lancet Rheumatology.

LIMITATIONS:

Biopsy staining practices varied across centers. The focus score was evaluated by a single expert pathologist at each center, with no central review. The study was conducted only at European centers, limiting generalizability to non‑European populations.

DISCLOSURES:

The study was supported by grants from Société Francaise de Rhumatologie and the European Alliance of Associations for Rheumatology. One author reported receiving support from the National Institute for Health and Care Research (NIHR), Birmingham Biomedical Research Centre, and NIHR-Wellcome Trust Birmingham Clinical Research facility. Several authors disclosed grants, contracts, honoraria, or other industry ties. 

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.