Analysis of complexity features of dermatological images, effective tool for automated diagnosis of melanoma

In recent years, several diagnostic methods have been proposed aiming at early detection of malignant melanoma tumour which is among the most frequent types of skin cancer. In this paper we discuss a new approach based on complexity analysis for classification of pigmented skin lesions using dermatological images. Features that describe the structure and colour of lesions, and show high discriminative characteristics are extracted using Approximate Entropy and the novel approach GeoEntropy. These features are used to construct a classification module based on support vector machines (SVM) for recognition of malignant melanoma from benign nevus. Experimental results showed that combination of proposed nonlinear features led to a classification accuracy of 94%.