Detection and Analysis of Skin Cancer from Skin Lesions

摘要

Skin cancers are the most common form of cancers in human, a physician faces many difficulties for accurate diagnose of lesion through its characteristics and in the naked eye. For that it is necessary to develop automatic methods in order to increase the accuracy of the diagnostic. In this paper, initially, skin images are filtered to remove unwanted particles, then a new method for automatic segmentation of lesion area is carried out based on Markov and Laplace filter to detect lesion edge, followed by convert image to YUV color space, U channel will be processed to remove thick hair and extract lesion area. Diagnosis of melanoma achieved by using ABCD rules with new method for determine asymmetry based on rotation of lesion and divide lesion to two parts horizontally and vertically then count the number of pixels mismatched between the two parts based on union and intersection between the two parts. New method to determine the number of colors based on suggestion of color regions for each color shade was suggested in this paper. The performance of the proposed method is tested on 220 different images. Accuracy for this method was encourage and reach up to 95.45%. The proposed method shows best accuracy when compared with other methods.