ABSTRACT
Researchers in medical imaging and computational intelligence groups are keen to develop sophisticated and complex methods to meet the specialized needs of medical applications, including the associated medical data. One of the significant medical imaging applications is the accurate image segmentation of skin lesions, which is key to useful, early, and noninvasive diagnosis of coetaneous melanomas. Extracting the true border of a skin lesion is one of the most important features that reveals the global structure irregularity (indentations and protrusions), which may suggest excessive cell growth or
regression of a melanoma. In this chapter, we investigate the application of two developed approaches to the skin lesion segmentation problem: iterative segmentation (IS) and neural network edge detection (NNED). These approaches are compared for synthetic lesions at different image signal-to-noise ratios (SNRs) to test their ability to delineate the lesion border shape. The use of synthetic lesions is advantageous in initial analysis and verification because if we know the true position of the lesion border, the different methods can be quantitatively and more accurately compared.
