An Agentic AI Framework for Adaptive Symptom Clarification and Knowledge-Augmented Disease Prediction in Healthcare Applications

Abstract

Artificial intelligence (AI) is transforming the healthcare domain, enabling agentic systems to provide accurate and open-access diagnostics to provide much-needed healthcare equity in the world. The article presents MEDICA (Medical Expert Diagnostic Intelligent Conversational Agent), a revolutionary AI-based platform that demonstrates a new level of validation accuracy of 98.85% and an F1-score of 98% on 24 different medical conditions. MEDICA, developed on a carefully fine-tuned DistilBERT architecture, and trained with 1305 balanced symptom descriptions, incorporates a novel Cognitive Self-Adaptation (CSS) module, which generates clarifying questions when the confidence falls below 0.7, making sure that it offers unparalleled diagnostic accuracy. It uses an advanced Retrieval-Augmented Generation (RAG) system to draw on a moderated library of 50 + authoritative documents to provide evidence-based and real-time medical guidance. Such outstanding outcomes, confirmed by extensive experimentation, make MEDICA a revolutionary solution in the area of early disease detection, and the extended options to multilingual support and mobile accessibility, make it a strong contender in terms of government financial contribution to the further development of innovations in the field of public health.

Ähnliche Arbeiten