This study is a cross-sectional, observational investigation designed to develop and validate a noninvasive ocular imaging-based predictive model for major inflammatory skin diseases, including psoriasis, atopic dermatitis, and urticaria. Grounded in Traditional Chinese Medicine (TCM) ocular diagnostic theory and integrated with advanced computer vision techniques, the study aims to establish objective, quantifiable biomarkers derived from scleral and bulbar conjunctival images. Approximately 950 participants (patients with psoriasis, atopic dermatitis, urticaria, and healthy controls) will be recruited from a tertiary academic hospital. Standardized high-resolution ocular images will be collected alongside comprehensive clinical, demographic, and comorbidity data. Disease severity will be assessed using validated clinical scoring systems (e.g., PASI, EASI, UAS7, DLQI). Image analysis will combine traditional radiomics feature extraction with deep learning architectures, including a Vision State-Space (VMamba) module for global-local feature representation and a Multi-Gate Mixture-of-Experts (MMoE) framework for multi-task learning. The model is designed to simultaneously perform disease classification and predict comorbidity risks (such as metabolic syndrome, hepatic insulin resistance, and recurrence risk). Primary outcomes include characterization of ocular feature patterns associated with inflammatory skin diseases. Secondary outcomes include correlations between ocular image-derived features and clinical severity indices. Model performance will be evaluated using ROC curves, AUC, calibration analysis, and decision curve analysis. This study aims to provide an objective, digitalized, and clinically applicable ocular biomarker framework to support early diagnosis, risk stratification, and comorbidity screening in inflammatory skin diseases.
See this in plain English?
AI-rewrites the medical criteria so a patient or caregiver can understand them. Always confirm with the trial site.
A TCM ocular diagnostic feature spectrum
Timeframe: 2026.03.05-2028.03.05