Método chan-vese de segmentação aplicado na localização de lesões de câncer melanoma e não melanoma

Abstract

Skin cancer is the most common of all cancers and the increased incidence is due in part to the people´s behavior in relation to sun expose. In Brazil, non-melanoma skin cancer is the most frequent in most regions. Dermatoscopy is the main kind of exam for the diagnosis of dermatological skin diseases. The computer aided medical diagnosis – CAD, has become increasingly common, such as the diagnosis of skin lesions, where the techniques for automatic extraction of their contours becomes crucial. Usually, the system starts by pre-processing the image, in other words removing undesired artifacts such as hair, freckles or shading effects. Next, the system performs a segmentation step to identify the lesion boundaries. The idea of CAD can be applied to all kinds of obtaining image, including conventional radiography, computed tomography, magnetic resonance, ultrasound and nuclear medicine. The computer's response can be useful, since the diagnosis of the health professional is based on subjective assessment, subject to intraand inter-personal variation and low image quality, eye fatigue or distraction. This study aims to present an analysis of the Chan-Vese model for segmentation of dermatological images of melanoma and non-melanoma types. The Chan-Vese segmentation model is based on the region growing technique and assets boundary model. The model showed a good result of segmented images in relation to the number of images used. The procedure of segmentation analysis was performed according to the results that had already been collected from database in which the images were acquired