AI-Based Predictive Models in Ayurveda

Abstract

Machine-learning predictive modelling has matured substantially in general healthcare applications over the past five years, with predictive models now contributing to clinical decision-support across risk stratification, treatment-response prediction, progression forecasting, and resource-allocation applications. For Ayurvedic clinical practice, predictive modelling has distinctive opportunities and challenges. Opportunities include constitutional-type prediction through structured phenotypic and biomarker data, treatment-response prediction that operationalises classical prakriti-specific therapeutic selection as a precision-medicine direction, progression prediction for chronic conditions under integrative management, and integrated predictive decision-support for European integrative-medicine clinical workflow. Challenges include data quality and availability for Ayurvedic-specific features, interpretability requirements for clinical trust, integration with existing clinical decision-making frameworks, and appropriate validation across European patient populations where most model training data originates from Indian cohorts. This study reviews 50 primary computational and translational studies published between 2015 and 2022, grouped under five model-application frames: PM1 (prakriti and constitutional-type prediction), PM2 (treatment response prediction and personalisation), PM3 (disease progression prediction for chronic integrative-management conditions), PM4 (integrated EHR + machine-learning clinical decision support), and PM5 (deployment validation in European integrative-medicine settings). Each frame was scored with the Ayurvedic Predictive Model Index (APMI). Deployment validation (PM5) scored highest at APMI 0.802, with prakriti prediction (PM1) at 0.784 and treatment response prediction (PM2) at 0.766.