Integrated Variable Marker Controlled Watershed Method with Level Sets for Semisupervised Classification
U. Rajyalakshmi, S. Koteswara Rao, Satya Prasad K · International Journal of Advanced Science and Technology · 2017
Symptomatic awareness of the Breast Cancer (BC) in the early stage is needful for treatment and also to support the radiologists during their diagnosis.In the present module, nuclei detections of BC biopsy images stained with Hematoxylin and Eosin, are done using Hough Transform and their Segmentation with Proposed Variable -Marker Controlled Watershed Method (VMCWM).Fixed size Structuring Element (SE) wipes out the dark and bright particulars while closing and opening morphology.So using weighted variance VMCWM, an altered SE map is generated to protect all image details.Total of 24 features (10 shape, 12 texture, and 2 intensity) are generated for classification using K-Nearest-Neighbor, Decision Trees, and Multi-Class Support-Vector Machine (MC-SVM) classifiers.The provided results integrated with level sets are compared with traditional methodologies and proved to be more accurate to the bench mark results.The entire module used 90 images for testing and 20 images for training obtained from database.Also cross validation is done using Leave One Out technique for all the samples.