This project introduces an AI-powered virtual Medical Chatbot to assist in the diagnosis of skin diseases and provide consultations. It applies deep learning-based image analysis and sophisticated natural language processing (NLP) for interactive support. Users can upload images of skin lesions, and the system processes them with a Convolutional Neural Network (CNN) using EfficientNetB3 to accurately label them into one of ten prevalent skin conditions such as eczema, melanoma, or psoriasis. To achieve good performance even with imbalanced datasets, the system applies various preprocessing techniques, including oversampling, under sampling, and data augmentation, before splitting the data into 70% training, 15% validation, and 15% for testing. By applying transfer learning and fine-tuning, the image classifier achieves high accuracy in its output. Following diagnosis, a BERT-based NLP chatbot engages in dialogue with the user, interpreting their needs and giving appropriate feedback to questions on symptoms, treatment, and prevention. The whole solution is packaged into an easy-to-use web application, providing a complete and automated system for skin disease diagnosis and giving personalized medical advice. This new framework demonstrates how well computer vision and transformer-based NLP models can be mixed, paving the way for scalable and affordable dermatological care using a smart healthcare assistant.
Likitha et al. (Sat,) studied this question.