Supervised Classification of Breast Cancer Malignancy Using Integrated Modified Marker Controlled Watershed Approach

Abstract

Worldwide statistics inform that breast cancer occupies second position causing mortality among women. Symptomatic detection of the disease in its early stage is important for treatment to help the internists and radiologists in their diagnosis. In the proposed module, nuclei locations are obtained using Hough Transform. Nuclei Segmentation of the pre-processed Hematoxylin and Eosin stained breast cancer histopathological images is done using Proposed Modified - Marker Controlled Watershed Approach (MMCWA). Small fixed Structuring Element (SE) size removes respective bright and dark details during opening and closing morphology & large SE size removes huge contour details of the input image. So, in the proposed MMCWA, by using weighted variance method, the adaptive Structuring Element size of the SE map is obtained to protect all details in the image. A total of 20 features, including 5 shape based features and 15 texture features were extracted for classification using Decision Trees, SVM and KNN classifiers. Algorithmic performance evaluation is accomplished and proved that the proposed integrated MMCWA provides better results than the traditional marker controlled watershed. The proposed module was trained with 96 images and tested over 24 images taken from the digital database.