A Comprehensive Evaluation of Deep Learning based Melanoma Detection and Classification Scheme

As the global population ages, skin cancer has emerged as one of the leading malignancies. The earlier skin cancer is detected, the better the prognosis. Increases in automated healthcare diagnosis systems can be directly attributed to the widespread adoption of deep learning algorithms in other sectors. In this study, deep learning is proposed as a means of accurate lesion zone extraction. While interventional treatments like surgery, radiation, and chemotherapy form the basis of cancer diagnosis, research suggests that emerging computer technologies, namely image processing mechanisms, can aid physicians in the crucial early stages of cancer treatment. In this study, we offer a technique for automatically identifying skin cancer in dermoscopy pictures. The suggested model, which made use of convolutional functions and depth wise separable convolutions, has been given the name Modified Learning based Melanoma Detection (MLMD). The framework's accuracy in classification is unquestionably higher than that of the network's initial learning phase and other possible dome topologies. Simulations comparing the suggested method to other similar modern techniques for malignancy detection show that the suggested technique achieves higher accuracy. To clearly evaluate the efficacy of the suggested model, it is cross-validated using the traditional learning method known as Convolutional-Neural-Network (CNN).