Comprehensive Analysis of Deep Transfer Learning Models for Skin Disease Classification with Explainable AI

Skin diseases such as cellulitis, impetigo, fungal infections, chickenpox, and shingles are sometimes difficult to diagnose owing to their visual similarity, resulting in misdiagnosis or delayed treatment. This problem is especially important in areas with low resources and rural locations where dermatological knowledge is hard to come by. In order to solve this issue, the current work uses a dataset of 1,159 clinical images to assess how well deep transfer learning models—more especially, VGG16, ResNet152V2, InceptionResNetV2, and Xception—perform in the automated categorization of eight major skin diseases. ResNet152V2 reached the maximum classification accuracy of 96.57% under controlled train-test conditions; nevertheless, because of dataset limitations, this result should be interpreted cautiously. Explainable Artificial Intelligence (XAI) approaches, namely Local Interpretable Model-Agnostic Explanations (LIME), were used to improve the interpretability and clinical reliability of the model predictions. For teledermatology, especially in underprivileged regions, the combination of interpretable AI and high-performance deep learning models provides a reliable and scalable solution. This method offers a precise and clear way to classify skin diseases automatically, which might lower diagnostic mistakes and enhance patient outcomes.