Classificação de assimetria em lesões de pele por meio de imagens usando máquina de vetor de suporte

Abstract

The increased occurrence of cancer cases over the years and the importance of prevention work motivated the development of this work. It aim is help the dermatologist in the diagnosis of skin lesions, providing information about the characteristics of asymmetry of ABCD rule (Asymmetry, Edge, Color and Diameter), widely used in the initial examination to determine if a lesion is malignant or no. To do so, are extracted from scanned images of the asymmetric features of the lesion, and classified as symmetrical or asymmetrical, through a Support Vector Machine (SVM). This process is used an anisotropic diffusion filter to soften the image and the model of active contour without edge (Chan-vese) to segment them. Thus, allows to define the contour of the lesion so that can be extracted their characteristics of asymmetry, used as input in the smart classifier.