AI-Guided Histopathological and Molecular Phenotyping of Salivary Glands in Sjögren’s Disease: Subtyping, Mechanistic Insights, and Translational Implications—A Systematic Review and Structured Functional Synthesis

This is the authors' abstract. We don't add key points for this paper.

Background/Objectives: Salivary-gland biopsy is central to Sjögren’s disease (SjD) assessment, but conventional measures incompletely capture glandular heterogeneity. This systematic review evaluated artificial intelligence (AI)/machine learning (ML) applications to SjD salivary-gland histopathology and molecular profiling, including biological and translational validation. Methods: PubMed, Scopus, and Embase were searched. Eligible studies used salivary-gland tissue as the primary analytical substrate or provided substantive glandular validation of systemic AI/ML findings. Evidence was synthesized functionally because of methodological heterogeneity. Results: Twenty-five studies were included. AI/ML identified histopathological and molecular phenotypes spanning immune/interferon activation, tissue architecture and immune-cell organization, epithelial stress and metabolic/cell-death remodeling, and translational response. Only a limited subset progressed beyond computational or tissue-level confirmation to functional perturbation, pharmacological testing, in vivo validation, or human treatment-response assessment. Recurrent public-dataset reuse, small cohorts, heterogeneous pipelines, and limited independent validation constrained reproducibility. Conclusions: AI/ML supports multidimensional SjD glandular phenotyping and shows emerging translational potential, but demonstrated clinical or therapeutic utility remains unestablished. Prospective multicenter validation is required before AI-derived glandular phenotypes can inform routine patient stratification or therapeutic decision-making.