Classification of Malignant Melanoma and Benign Nevi from Skin Lesions Based on Support Vector Machine

This paper proposes an automated system for discrimination between melanocytic nevi and malignantmelanoma. The proposed system used a number of featuresextracted from histo-pathological images of skin lesionsthrough image processing techniques which consisted of aspatially adaptive color median filter for filtering and a Kmeansclustering for segmentation. The extracted featureswere reduced by using sequential feature selection and thenclassified by using support vector machine (SVM) to diagnoseskin biopsies from patients as either malignant melanoma orbenign nevi. The proposed system was able to achieve a goodresult with classification accuracy of 88.9%, sensitivity of87.5% and specificity of 100%.