Federated LearningBased Skin Disease Detection with Secure Teledermatology for Privacy Preservation

The increasing prevalence of skin diseases, including eczema, melanoma, atopic dermatitis, basal cell carcinoma, psoriatic conditions, fungal infections, and viral lesions, demands accessible, accurate, and privacy-preserving diagnostic solutions. This paper introduces a federated learning-based skin disease detection system integrated with secure teledermatology to enable remote consultations while safeguarding patient data privacy. The proposed framework employs the MERN stack (MongoDB, Express.js, React, Node.js) for a robust web application supporting two primary user roles: patients and doctors. Patients can register, log in, upload skin images for automated prediction using deep learning models, view available doctors, book consultations, track request status, process payments, engage in real-time remote consultations, note prescribed medications, and submit feedback. Doctors handle registration/login, review and accept/reject consultation requests, conduct virtual examinations, provide prescriptions, view payments and feedback, and communicate securely.

For skin disease prediction, the system leverages a multi-class classification approach trained on a comprehensive image dataset covering diverse dermatological conditions (e.g., eczema, melanoma, psoriasis, tinea, warts). Multiple convolutional neural network architectures—CNN, ResNet, and VGG16—are evaluated and trained via federated learning to enable collaborative model improvement across distributed devices/hospitals without sharing raw sensitive patient images, ensuring compliance with privacy regulations. To enhance trust and clinical interpretability, Explainable AI (XAI) techniques are incorporated to provide transparent reasoning behind predictions. This integrated approach combines accurate AI-driven preliminary diagnosis, privacy-preserving model training, and seamless teledermatology services, offering an efficient, secure, and user-centric solution for early skin disease detection and management.

Keywords: Federated Learning, Skin Disease Detection, Teledermatology, Privacy Preservation, CNN, ResNet, VGG16, Explainable AI (XAI), MERN Stack, Remote Consultation, Deep Learning, Dermatological Diagnosis, Multi-class Classification