Intelligent Framework for Skin Cancer Detection and Classification Using Deep Learning

Skin cancer is one of the most common types of cancer worldwide, and catching it early can make a big difference in treatment success. In this paper, This study proposes a deep learning-based framework for skin cancer classification. Our approach involves carefully preprocessing images, augmenting data to improve the model, and then training CNN models through transfer learning. We tested our method on a dataset of 2,300 labeled skin lesion images, categorized into cancer and non-cancer groups. To see how well it works, we looked at various metrics like accuracy, sensitivity, specificity, precision, F1-score, and the area under the ROC curve (AUC). The results are promising: ResNet 18 achieved an impressive average accuracy of 96.8%, with sensitivity at 96.4%, specificity at 96.5%, precision at 96.7%, and an F1-score of 96.3% across different validation tests. It also scored an average AUC of 0.99, showing its strong ability to tell these two classes apart. These encouraging findings suggest that deep learning image classification can be a helpful tool for automated skin cancer analysis and could support doctors in making more informed diagnoses.