AI Healthcare Assistant: Integrating Symptom-Based Chatbot, Medical Image Analysis, And Real-Time Vitals Monitoring For Comprehensive Disease Prediction

Abstract

Healthcare systems worldwide face challenges in accessibility, early diagnosis, and patient monitoring. This paper presents an integrated AI Healthcare Assistant that combines natural language processing (NLP), deep learning for medical image analysis, and signal processing for real-time vitals de- tection. The system implements: (1) a symptom-based chatbot using TF-IDF vectorization and Naive Bayes classification for disease prediction, (2) convolutional neural networks for chest X-ray analysis supporting classification of Normal, Pneumonia, COVID-19, and Tuberculosis, and (3) photoplethysmography (PPG) signal processing for non-contact heart rate, respiratory rate, and stress level monitoring. Built on Streamlit with role- based access control for doctors, patients, and administrators, the platform provides automated treatment recommendations, confidence scoring, and comprehensive health analytics. Eval- uation on benchmark datasets demonstrates 91% accuracy for symptom classification, 93% for X-ray classification, and 95% for vitals detection. The system is deployable in resource-constrained settings and supports telemedicine workflows.