Enhanced Skin Cancer Classification with VGG-16 Using Dermoscopic Image Analysis
M Joly, M. Vadivel, S. Suguna Mallika, Kavinsandron Muthu, Ishwarya M. V, Cidambi Srinivasan
Abstract
One of the most often occurring cancers globally, skin cancer depends on early identification for efficient treatment. Often time-consuming and prone to human variability are traditional diagnostic techniques. This work improves skin cancer classification using dermoscopic image analysis using the VGG-16 deep learning (DL) model. 25,000 dermoscopic images from publicly accessible sources were utilized in a dataset; images were pre-processed using resizing, normalizing, and augmentation to enhance model generalizing. Transfer learning helped to refine the pre-trained VGG-16 model for binary classification benign against malignant. The experiment's results showed the proposed model's exceptional performance, which showed 100% accuracy. The technique also reduced false positives and negatives, providing ideal early detection. These results highlight the possibilities of DL in dermatology as they provide a quick and accurate AI-assisted diagnostic instrument. Real-time clinical validation and multimodal data integration will be the main emphasis of the next projects to improve model performance.