Skin Lesion Images Segmentation: A Survey of the State-of-the-Art

This paper presents a detailed and robust survey of the state-of-the-art algorithms and techniques for performing skin lesion segmentation. The approach used is the comparative analysis of the existing methods for skin lesion analysis, critical review of the performance evaluation of some recently developed algorithms for skin lesion images segmentation, and the study of current evaluating metrics used for performance analysis. The study highlights merits and demerits of the algorithms examined, observing the strength and weakness of each algorithm. An inference can thus be made from the analysis about the best performing algorithms. It is observed that the advancement of technology and availability of a large and voluminous data set for training the machine learning algorithms encourage the application of machine learning techniques such as deep learning for performing skin lesion images segmentation. This work shows that most deep learning techniques outperform some existing state-of-the arts algorithm for skin lesion images segmentation.