Melanoma Detection on Dermoscopic Images using Superpixels Segmentation and Shape-Based Features

摘要

In this work, we present a shape-based approach to automatic skin lesion segmentation and classification indermoscopic images. We aim to differentiate three types of lesion 1) common nevi, 2) atypical nevi, and 3)melanomas by exploring the morphology features of segmented skin lesions. Our method is an attempt to designa computer-aided ABCDEs of melanoma, where the Asymmetry and Border components are estimated usingmorphological features. The lesions are first segmented using a super-pixel merging strategy with an RGBcriterion. Later, the segmentation method was evaluated on the PH~2 dataset, and compared with other state-ofthe-art skin segmentation methods. The classification was also conducted on the PH2 dataset through a 10-foldcross-validation set-up with a training and testing set partition of 90% and 10% respectively. We employedlogistic regression, SVM and a neural network as classification algorithms. The best performances was 86.5% onaverage with the neural network.