Augmenting Medical Diagnostics with AI: A Dual Approach Using RAG-Based Chatbots and Nano GPT Models

A pioneering dual approach is proposed to enhance medical diagnostics through the integration of RAG-based chatbots and Nano GPT models. Leveraging AI advancements, our methodology aims to revolutionize medical practice by improving diagnostic accuracy, patient care, and healthcare efficiency. The research encompasses a comprehensive exploration of hyperparameters critical for optimizing the Nano Generative Pre-trained Transformer model, focusing on parameters such as learning rate, batch size, and layer configuration. Through systematic experimentation, we evaluated various hyperparameter configurations to minimize both training and validation loss of 0.4889 and 1.4031 respectively for 1.34 million parameters, ensuring the model’s ability to learn from medical data while effectively generalizing to unseen patient cases. Our findings demonstrate the intricate interplay between model complexity, regularization techniques, and generalization capabilities. We further propose a sampling mechanism tailored for generating medical text from the GPT-like language model, facilitating model setup and fine-tuning generation details. The proposed dual approach, integrating interactive patient consultations with advanced diagnostic analysis, showcases promising results, marking a significant step toward the future of AI-driven healthcare. Future work will focus on conducting clinical trials to further validate the chatbot’s performance in real-world scenarios.