Open Collections

Automatic pathology of prostate cancer in whole mount slides incorporating individual gland classification

University of British Columbia

This thesis presents an automatic pathology (AutoPath) approach to detect prostatic adenocarcinoma based on the morphological analysis of high resolution whole mount histopathology images of the prostate. We are proposing a novel technique of labeling individual glands as benign or malignant exploiting only gland specific features. Two new features, the Number of Nuclei Layers and the Epithelial Layer density are proposed here to label individual glands. To extract the features, individual gland and nuclei units are segmented automatically. The nuclei units are segmented by employing a marker-controlled watershed algorithm. The gland units are segmented by consolidating their lumina with the surrounding layers of epithelium and nuclei. The main advantage of this approach is that it can detect individual malignant gland units, irrespective of neighboring histology and/or the spatial extent of the cancer. Therefore, a more sensitive annotation of cancer can be achieved by the proposed AutoPath technique, in comparison to the current clinical protocol, where the cancer annotation is performed at the regional macro level instead of glandular level technique.We have also combined the proposed gland-based approach with a regional approach to perform automatic cancer annotation of the whole mount images. The proposed algorithm performs the task of cancer detection in two stages: at first with pre-screening of the whole mount images in a low resolution (5x), and then ii) a finer annotation of the cancerous regions by labeling individual glands at a higher magnification (20x). In the first stage, the probable cancerous regions are classified using a random forest classifier that exploits the regional features of the tissue. In the second stage, gland specific features are used to label individual gland units as benign or malignant. The strong agreement between the experimental results and the pathologist's annotation corroborates the effectiveness of the proposed technique. The algorithm has been tested on 70 images. In a 10-fold cross validation experiment it achieved average sensitivity of 88%, specificity of 94% and accuracy of 93%. This surpasses the accuracy of other methods reported to date.

Text

2014-03-28

Attribution-NonCommercial-NoDerivs 2.5 Canada

http://hdl.handle.net/2429/46290

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/2.5/ca/