Automated Skin Cancer Classification Using VGG16-Based Deep Learning Model

Melanoma is the most fatal type of skin cancer due to its high potential to metastasize and its early-stage similarity to benign skin lesions, such as common moles. This resemblance often leads to delayed diagnosis and treatment. This study proposes a skin cancer classification model using the VGG-16 architecture through a transfer learning approach. Utilizing the ISIC 2017 dataset, which includes three skin lesion categories such as Melanoma, Nevus, and Seborrheic Keratosis—this research applies preprocessing, segmentation, and feature extraction. The classification stage uses a modified VGG-16 model, achieving the best performance at a 70:30 train-test split with 100 epochs and batch size of 16, resulting in an accuracy of 73.09% and F1-score of 0.71. Evaluation with the ROC curve indicates challenges in distinguishing Melanoma from other lesions due to overlapping patterns. Additionally, the study presents a prototype mobile application for real-time classification, demonstrating the practical implementation of the proposed model.