Early Detection and Medication Suggestion for Parkinson's Disease Using Multimodal, Agentic AI, and RAG

Parkinson's disease (PD) is a progressive neurodegenerative disorder with no cure and limited early biomarkers. Early diagnosis is critical for intervention, enrollment in clinical trials, and optimizing symptomatic treatment. This paper proposes a multimodal deep learning framework that integrates clinical imaging, sensor data, speech, and biochemical biomarkers for early PD detection. To bridge detection with personalized medication suggestions, we embed an agentic AI system powered by Retrieval-Augmented Generation (RAG) to provide evidence-based clinical recommendations contextualized to patient profiles. Extensive experiments on publicly available datasets demonstrate that the proposed system outperforms unimodal baselines and yields actionable treatment options grounded in current clinical evidence.