Automatic skin lesion segmentation

Abstract

In the modern world, cancer has increasingly become a health problem. It has been listed as the first three disease among the ‘cause-known deaths’ in our country. Malignant Melanoma, one of the skin cancer types, is the cause of 75% of all skin cancer related deaths even though it is 4% of all skin cancer cases. The examination of the diseases are diagnosed through visual inspections by the dermatologists. This brings the possibility of human error. In this study, a computer based segmentation system is developed to assist the expert dermatologists for determining whether the lesions on the skin are cancerous or not. The edges of lesions on the original images is drawn by enhancing and segmenting the lesions via image processing techniques, since lesions' shapes, color distributions and edges are important parameters for cancer determination process. In this way, the error rates are reduced by making it easier for the dermatologist to examine the lesions and make decisions.