Segmentação, análise e classificação de imagens de lesões de pele usando dimensão fractal

Abstract

Melanoma is a type of skin tumor that most affects the population of clear skin. Although not as common among skin tumors, melanoma is the most aggressive, so early diagnosis is of paramount importance. To distinguish a benign from a malignant melanoma can be used the rules ABCD, where A is related to the asymmetry of melanoma, B is related to melanoma edge C is associated with the colors melanoma and finally D which is associated with melanoma diameter. This work aims to segment, analyze and classify images of skin lesions acquired in a public image bank. Melanoma edge analysis was performed using the fractal dimension, and from the values obtained in the application of rules A, B and C lesions were classified as malignant or benign. With the use of the three rules (A, B and C) was reached at 80,76% accuracy in the skin lesions of image classification.