Morphological and Statistical Operations Based on Segmentation and Classification of Skin Cancer Images

Cancer remains one of the primary causes of deaths worldwide. Skin cancer poses serious medical concerns for many people worldwide. Dermoscopy can help detect this preposterousness to determine whether visible areas on skin contain benign or malignant cancerous tumors. Melanoma or non-melanoma cancer of the skin are usually the two forms most likely to result from skin cancer development. Computer tools designed for skin cancer detection use a precise segmentation method to analyze clinical features that indicate tumour risk in order to calculate their probability. For this, relevant components in digital images must first be isolated using segmentation techniques that isolate them accurately from one another. Due to the complex nature of cells, automating their processing can be an extremely time-consuming and laborious task for pathologists. In this paper we offer an easy and innovative method to classify and segment cells of skin. As part of its preliminary processing stage, this procedure begins with de-nosing, de-correlation and color segregation operations being executed on images. Two new methods - Morphological as well as Statistical Edge Based Segmentation and Morphological and Statistical Region Based Segmentation Techniques are introduced and described here to separate skin images into individual nuclei of lesions. Once segmented color components have been segmented separately, they are combined for an ultimate segmentation outcome. After extracting nuclei from their cells and extracting their morphological features from them, two techniques previously discussed outperform standard segmentation methods in terms of performance. Morphological and Statistical Edge Segmentation based on Morphology is superior to Morphological and Statistical Region based Segmentation because nuclei can be identified based on their particular character of morphology; classification proves accuracy of segmentation is evident and overall, it stands at 96%.