An enhanced framework for automatic diagnosis of breast cancer and robust tumour proliferation scoring model for digitized histopathology images

Abstract

Breast cancer is the leading cause of mortality in women. Early diagnosis newlineof breast cancer can reduce the mortality rate. The diagnosis of breast cancer newlinehistology images with hematoxylin and eosin stained is non-trivial, labor-intensive newlineand often leads to a disagreement between pathologists. Computer-assisted newlinediagnosis (CAD) systems contribute to help pathologists improve diagnostic newlineconsistency and efficiency. Nowadays, Computer Assisted Diagnosis (CAD) newlinesystem are widely used to assist practitioners in detecting, and classifying various newlineabnormalities present in medical images. newlineWith the recent advances in machine learning and deep learning newlinealgorithms, they have been successfully used for histopathological image newlineanalysis. However, automated nuclei detection is problematic in unevenly shaped, newlineoverlapping, touching nuclei, classifying the benign and malignant cells, grading newlinethe cells and mitosis detection. The important challenge for the histopathological newlineimage analysis is the system evaluation. Due to the limitations in the availability of newlinedata, there are chances of substantial amount of bias if the evaluation of the system newlineis not done properly. newlineTherefore, this research is performed to automatically detect, classify, newlinecategorize and count the cells by overcoming these issues. This work is focused on newlinevarious algorithms to deal with the problems of setting an automatic saliency nuclei newlinedetection, classification of benign and malignant cells, rough set based feature newlineselection process, and counting the mitosis cells. newline