Development of automated cancer Diagnostic system for breast and Prostate histopathology images

Abstract

Cancer, a dreadful disease, causes loss of lives and also has tremendous socioeconomic impact. Merely investing in cancer treatment is not an economically viable option. Early detection is one of the key solutions that need focus. Histopathology Images (HI) is the gold standard for cancer detection and evaluation. However, the analysis of such images is challenging even for experienced pathologists, resulting in problems of inter and intra observer variability. Besides that, the analysis is time and resource consuming. One of the ways to accelerate such an analysis is by using Computer Assisted Diagnosis systems. CAD systems are developed using intelligent algorithms from the areas of image processing and pattern recognition. The primary goal of this research is to develop a comprehensive cancer diagnosis system for hematoxylin and eosin (HandE)-stained breast and prostate histopathology images (HI). The work presented in this thesis focuses on, (1) designing a more suitable colour normalization approach, (2) building an integrated model for nuclei segmentation and multiple object overlap resolution, (3) introducing and evaluating multi-level features for extracting relevant histomorphometric and interpretable information, (4) designing classification algorithms for performing the tasks of cancer diagnosis :tumour identification newline