Performance Evaluation of Breast Cancer Histopathological Images

Image segmentation plays a crucial task in image processing, as the segmentation output will influence all the successive processes during image analysis. Segmentation approaches developed in literature have their pros and cons. The book presents various segmentation approaches for Haematoxylin and Eosin stained Breast Cancer histopathological images to classify as Benign/Malignant. The book described an Adaptive Structuring Element¿s size Marker Controlled Watershed Approach. In this approach, Structuring Element map is constructed using the weighted variance method for detail components protection in the image. Two more hybrid integration Methodologies are proposed to handle critical image inhomogeneities. Local Clustering Image-Function from ASEMCWA is integrated with Novel Multi-phase Level Sets for segmentation of histopathological images. This hybrid approach solves the problem of images with blurred/weak edges at the cost of no reinitialization. To solve the gradient problem, an integrated approach of Non-Subsampled Contourlet transform and NMPLS is constructed. This book gives the Comparative Subjective performance of State-Of-Art Segmentation + Classification approaches.