AN AI-POWERED MEDICAL CHATBOT FOR PARKINSON'S DISEASE TREATMENT ASSISTANCE USING LARGE LANGUAGE MODELS

Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting motor function in millions of
individuals worldwide. Patients and caregivers routinely encounter challenges in understanding the complex
pharmacological regimens used in PD management, including drug side effects, contraindications, and treatment
strategies. Existing digital health platforms either provide static, non-interactive content or general-purpose
chatbots lacking PD-specific domain depth. This paper presents the design, implementation, and evaluation of an
Artificial Intelligence-based Medical Chatbot tailored to Parkinson's disease treatment assistance. The system
integrates a curated pharmacological knowledge base covering six core PD medications — Levodopa,
Pramipexole, Ropinirole, Amantadine, Selegiline, and Entacapone — with the LLaMA 3.3-70B large language
model (LLM) accessed via the Groq high-speed inference API. A Flask-based web application delivers an
interactive chat interface alongside RESTful API endpoints for programmatic drug data access. A response
enhancement module automatically enriches LLM outputs with drug-specific authoritative resources and clinical
case-study references. Systematic functional and non-functional evaluation across 17 test cases demonstrates
average API response latencies of 2.3–3.8 seconds, a chatbot gesture recognition success rate of 97.8%, and high
user satisfaction in usability testing across diverse participant groups. The system also supports multilingual query
responses in Hindi, Tamil, Malayalam, Japanese, Chinese, and Spanish, as evidenced by prototype testing. The
proposed architecture establishes a reusable, modular template for domain-specific AI health chatbots targeting
other chronic neurological conditions.