Enhanced CNN and Federated Learning Algorithm for Secure and Precise Dermatological Diagnosis

The diagnosis of skin diseases, a prevalent global health concern, often begins with visual observation. However, the complex formations, diverse colors, and data security concerns make accurate classification challenging. This project proposes the development of an enhanced Convolutional Neural Network (CNN) model integrated with a Federated Learning approach to ensure secure and precise dermatological diagnosis. A custom image dataset encompassing five skin disease classes was created for this purpose. The CNN model was compared with various benchmark algorithms, demonstrating significant improvements in precision and recall for diseases such as acne, eczema, psoriasis, melanoma, and lichen planus. Federated Learning was employed to address data privacy issues by distributing data across multiple clients while collaboratively updating a central model. The results show that the integration of CNN-based classification with Federated Learning not only enhances accuracy but also prioritizes data security, making it a promising approach for advancing skin disease detection.