Publication

Abstract : Skin cancer is an uncontrolled growth of abnormal cells in the human body. It is usually caused by the sun's harmful rays. Studies indicate that there is much research going on to identify if a skin lesion is malignant (cancerous) or benign (non-cancerous), but the most difficult task for a physician is to identify the type of skin cancer. Hence, it is very important to classify the type of tumor for a right treatment. The other focus is to propose an algorithm which suits the real time environment in terms of detection speed, learning capabilities and better quantitative scores. Hence, to effectively identify and classify the type of tumor, this paper proposes a Yolo deep neural network which can classify 9 different classes of skin cancer. In training phase, data augmentation process is carried out to increase the number of images for better classification efficiency. Then, a bounding box is drawn to capture the object of interest which preserves the salient features of the image. In the testing phase, the given image is basically divided into equally sized blocks and each block is compared with the trained labelled image blocks to classify the type of cancer. Both Yolo V3 and Yolo V4 are tested for classification. The experimental analysis shows that the proposed neural network achieves the mean average precision score of 88.03% and 86.52% for Yolo V3 and Yolo V4 respectively on a dataset of 4389 images, divided into nine distinct classes. Also, the trained neural network is statistically evaluated with well-known quantitative measures such as accuracy, precision, re-call and F1 score where the average attained scores are 98.06%, 92.75%, 91% and 92% respectively.