Abstract
Cutaneous malignancies, or skin cancers, are a frequently diagnosed type of cancer worldwide. A late-stage diagnosis of cancer causes it to spread and metastasize to other nearby organs. As the disease progresses, it can decline the patient’s immune system, eventually to death. The main challenge in skin cancer diagnosis is classifying Melanoma from benign pigmented skin lesions and treating Melanoma early to save the patient’s life. There have been growing appeals for using convolutional neural networks (CNN) in image recognition and classification. Recent studies of skin cancers detection via CNN methods have stimulated related studies, including using the advanced Region-CNN (R-CNN) algorithm. This article examines papers on detecting skin cancers/lesions using classic CNN methods and Faster R-CNN and evaluates their performance and classification results in several currently available data sets.