Detection of Melanoma with Multiple Machine Learning Classifiers in Dermoscopy Images

Skin cancer cases, in recent years, has become increasingly widespread because of increasing the effect of ultraviolet radiation as a result of thinning and perforation of the ozone layer in the atmosphere. The fact that melanoma, one of the most lethal types of skin cancer, can be treated at a high rate in early diagnosis has increased the interest in the studies in this field. In this study is focused on machine learning approaches that can be used in the diagnosis of melanoma in dermoscopic images. In the first step, color and texture features of images accessed from the dermoscopic image database were extracted with image processing techniques. In the second step, by using these features machine learning classifiers in different program environments have been trained and tested. Results from the proposed method indicated that melanoma can be detected with 97 % accuracy.