Computer-aided automated Gleason grading of prostate cancer histopathology images

The most common scoring system utilized for prostate cancer aggressiveness is the Gleason Grading system. The scoring system is performed by pathologists and can be variable, even by the same pathologists. Various factors affect the pathologist's reports such as training, pathologist's skill, tiredness, variation in interpretation, and applying Gleason grading to only small portions of prostate biopsy tissue sample. Several Computer aided diagnostics and grading systems utilizing the features based on different tissue components have been proposed in the past, of which only handful of systems have been recorded to utilize the nuclei information of the tissue. The Nuclei information plays a key role in diagnosing prostate cancer. The cancer nuclei differ in shape and size as compared to the normal cell. The cancerous cells have uneven shape and may be larger or smaller in size.

The goal of this thesis is to develop a fully automatic computer-aided detection and Gleason grading (CAD-CG) system using the present nuclei information in the histopathology images. Particularly we propose, (i) a new nuclei segmentation method, (ii) sub-system to classify the prostate histopathology images as cancerous and non-cancerous (iii) sub-system to classify the prostate cancer histopathology images into different grades of cancer, (iv), a novel hierarchical classification system to combine the two proposed sub-systems, and (v) analysis of the system performance using different evaluation parameters.