AI-Infused Chatbots in Healthcare: A New Frontier for Chronic Disease Management

Abstract

This study introduces an AI-based telemedicine system for long-term mental health care that integrates wearable sensors, facial emotion analysis, conversational AI, and predictive deep learning models into a privacy-preserving system. The presented framework continuously tracks patients' physiological and behavioral parameters through IoMT-enabled wearable devices, which capture heart rate variability (HRV) and activity measures via Bluetooth synchronization. A convolutional neural network (CNN) extracts facial images for real-time emotion recognition, while a recurrent neural network (RNN) examines physiological temporal patterns to make inferences about emotional state. The patient communicates using an AI-based chatbot with natural language processing (NLP) through a Transformer-based model (Distilbert, GPT-2) for empathic interaction, intent recognition, and symptom tracking. A multi-layer perceptron (MLP) and LSTM-based prediction model creates personalized treatment plans with doctor-validated assurance through a secure web-based portal. The system is built using TensorFlow, PyTorch, Flask, React.js and PostgreSQL, with all AI modules installed locally to preserve data privacy. Experimental results prove 78.9% accuracy in multi-modal emotion recognition, 89.3% accuracy in chatbot intent classification, and 76.4% agreement between AI-derived treatment plans and physician recommendations. The integrated platform demonstrates the feasibility of AI-enabled telemedicine for ongoing, customized and ethical mental health care, presenting a scalable model to be used for comprehensive chronic disease management purposes.