Liver Fibrosis

Liver fibrosis requires early diagnosis to prevent cirrhosis, yet current methods like biopsy are risky, and non-invasive tools lack early-stage sensitivity. This study proposes a multi-modal logistic regression model combining blood biomarkers, coagulation profiles, and FibroScan, with adaptive thresholds optimized by machine learning. Using LASSO for feature selection and SHAP for interpretability, the model is validated in a multicenter cohort. Deliverables include two peer-reviewed articles and an open-access risk stratification tool. The model shows >0.85 AUC and 75% improved early fibrosis sensitivity, offering a scalable solution for MASLD diagnosis in resource-limited settings.