Skin Cancer Classification Using Machine Learning Techniques

Abstract

Skin cancer is the most common form of malignancy that affects human populations. The Basal cell carcinoma, Squamous cell carcinoma and Melanoma are the most commonly occurring skin cancers on the human skin that occurs due to the abnormal growth of skin cells. All these three types of skin cancers are the most dangerous form of skin cancers whose treatment is possible only if it is detected in early stage. Therefore, the classification of these three cancers types are really challenging. In this paper, these three commonly occurring skin cancers on human skins are classified using machine learning techniques. The proposed technique consists of three stages, namely, preprocessing, feature extraction, and classification. The database consists of 150 images of the three cancer types which are taken from the Dermnet public database. Thus, each cancer types having 50 images are used in this paper. In the first stage, the images were preprocessed using two methods, the edge enhancement method and then the histogram equalization method. In the second stage, two different features vectors were used such as texture and statistical features. In the classification stage, the support vector machine classifier is used to classify the three cancer types as it uses a decision plane to separate a data set having different classes. The performance measures were analyzed for the texture and statistical feature vectors individually as well as combined. The overall classifier performance measures used are accuracy, sensitivity and specificity. A classificiation with a success of 97% accuracy has been obtained by using SVM classifier.