An Accelerated System for Melanoma Diagnosis Based on Subset Feature Selection

In this paper we present an optimised system fordiagnosing skin lesions based on digitized dermatoscopic color images.This system is composed mainly of three levels : lesion detection, lesion description (featuresselection) and decision.The preprocessing of the lesion imageis used to remove the undesired objects from the original imageand the extraction of the lesion is done by separating it from the healthy surrounding skin.The classification scheme is based on the extraction of a set of features modeling clinical signs of malignancy. The produced vector of features scores is used as input toa multi-layer perceptron classifier in order to assign the lesionto the class of benign lesions or to the one of malignant melanomas.We focus particularly in this paper on the critical step ofthe features selectionallowing to select a reasonablereduced number of useful features while removingredundant information and approximating the properties of melanoma recognition.This permits to reduce the dimension of the lesion\u27s vector, and consequently the calculation time,without a significant loss of information.In fact, a large set of features was investigatedby the application of relevant features selection techniques.Then, the number of features forclassification was optimized and only five well-selected features were usedto cover the discriminatory information about lesions malignancy.With this approach, for reasonably balanced training/test sets, we recorda good classification rate of 77.7% in a very promising cpu time