11 - Neutrosophic sets in dermoscopic medical image segmentation

Abstract

Medical image segmentation plays an important role in a disease's pattern recognition for further diagnosis and treatment. Recently, the neutrosophic set has been applied to segmentation to reduce/remove the uncertainty in the medical images. This chapter proposes an effective dermoscopic skin lesion segmentation procedure using a neutrosophic set-based kernel graph cut (NKGC) and segments the dermoscopic images in the neutrosophic set (NS) domain. Initially, a histogram-based clustering estimation (HBCE) procedure was applied to determine the number of clusters and their centroids. Afterward, the dermoscopic images are transformed into the NS domain using the neutrosophic c-means (NCM) algorithm. Lastly, a kernel graph cut (KGC) procedure was employed for the segmentation process. For performance assessment, comparative and quantitative studies were carried out over the International Skin Imaging Collaboration (ISIC) 2016 skin lesion dataset. The results proved the dominance of the proposed NKGC compared to the neutrosophic set-based graph cut (NGC) and the traditional graph cut (GC). The NKGC achieved the best average accuracy value of 97.41% to segment the skin lesion regions compared to the other methods in this chapter.