Abstract

Among all the types of skin cancer Malignant Melanoma (MM) is the most dangerous skin cancer. Skin cancer is commonly called as melanoma. There are two types in melanoma namely, Benign Melanoma and Malignant Melanoma. Both benign and malignant melanoma appears similar at the initial stages. So that it is difficult to differentiate both the melanomas, which is the main problem in the detection of skin cancer. Only an expert dermatologist will be able to provide an accurate classification as to which is benign and which is malignant. The standard approach in automatic dermoscopic image analysis consists of three stages: 1) image segmentation 2) feature extraction 3) lesion classification. Main advantage of this Computer Aided Diagnosis (CAD) is that only the patient confirmed with malignant melanoma need to undergo various painful diagnoses like Biopsy and others with benign melanoma need not. In this paper three segmentation techniques were applied to segment the lesion boundary. Accurate segmented output can be taken out by comparing three performance metrics namely sensitivity, accuracy and border error. Two classifiers are used to classify the types of melanoma, namely Neural Network(NN) and Support Vector Machine (SVM).