Medical Diagnosis Chatbots Using Mistral Decoder Model

The integration of Large Language Models (LLMs) in healthcare has facilitated the development of medical diagnosis chatbots can assist patients in understanding the preliminary reason for their symptoms. This proposed work developed a medical diagnosis chatbot using LLMs which can be used as an initial diagnostic assessments tool based on patient-provided symptoms and helps to understand the basic medical terms. The chatbot uses natural language techniques (NLP) to read queries, retrieve relevant medical data using the Mistral Decoder Model, and generate diagnostic suggestions. This proposed model integrates Retrieval-Augmented Generation (RAG) with FAISS (Facebook AI Similarity Search) for efficient information retrieval improved response relevance. The proposed chatbot is fine-tuned using The Gale Encyclopedia of Medicine 2, a comprehensive medical reference that ensures reliable and accurate responses. The suggested model is evaluated using accuracy and latency measures. our model attained 82.5% diagnostic accuracy, 1.6 sec latency and 2.0 sec response time, outperforming existing key driven and rule-based chatbots. Thus, the model shows reduced latency, improved response time. This model demonstrates enhanced conversational performance, making interactions more human-like and informative to the user.