Computer-Aided Classification of Skin Cancer based on the YOLO Algorithm

Skin cancer is considered to be the most common type of cancer worldwide. Nonetheless, the corresponding death rate can be considerably reduced with early detection and classification. Massive efforts have been made in recent years to build machine learning algorithms that can aid in the early identification of skin cancer. The three most prevalent forms of skin lesions — melanoma (MEL), squamous cell carcinoma (SCC), and basal cell carcinoma (BCC) — are the subject of our paper’s effort on the accurate classification of these types of cancer. To achieve this, YOLO, version 7 (v7), a convolution neural network (CNN) architecture, is implemented through transfer learning. After completing data augmentation, the results obtained by YOLO, with a total of 2792 training samples, demonstrate superior performance in comparison to previously published research works in the literature. In terms of accuracy, sensitivity, and specificity, the average values are 89.65 %, 85 %, and 91.90 %, respectively.