Bilkent University
Department of Computer Engineering
MS THESIS PRESENTATION
Detection and Classification of Breast Cancer in Whole Slide Histopathology Images Using Deep Convolutional Networks
Barış Geçer
MS Student
(Supervisor: Assoc. Prof. Dr. Selim Aksoy)
Computer Engineering Department
Bilkent University
The most frequent non-skin cancer type is breast cancer which is also named one of the most deadliest disease where early and accurate diagnosis is critical for recovery. Recent medical image processing researchers yield promising results that may contribute to the analysis of biopsy images by enhancing the understanding or by revealing possible unhealthy tissues during diagnosis. However, they study on well-annotated and -cropped patches, whereas a fully automated computer-aided diagnosis (CAD) system requires whole slide histopathology image (WSI) processing which are, in fact, gigantic in size and, therefore, difficult to process with a scalable computational capacity and time. Moreover, those whole slide biopsies consist of healthy, benign and cancerous tissues at various stages and thus, the detection and classification of diagnostically relevant regions is challenging.
We propose an efficient end-to-end CAD system for localization and classification of region-of-interests in WSI by employing state-of-the-art deep learning techniques. The system is developed to resemble organized workflow of expert pathologists by means of progressive zooming-into-details, and it consists of two separate sequential steps: (1) detection of region-of-interests (ROI) in WSI (2) classification of the detected ROIs into five diagnostic classes. The novel saliency detection approach propose to mimic efficient search pattern of experts at multiple resolution by training a number of deep networks with the samples extracted from the tracking records of pathologists' viewing of WSIs. The detected relevant regions are fed to the classification step that includes a deeper network that produce probability maps for classes, followed by a post-processing step for final diagnosis.
According to our experiments, the proposed saliency detection approach outperforms the state-of-the-art method by means of both efficiency and effectiveness, and the final classification of our complete system obtains slightly lower accuracy than the mean of 45 pathologists' performance. According to the McNemar's statistical tests, we cannot reject that the accuracies of 30 out of 45 pathologists are not different from the proposed system. At the end, our trained deep model is visualized with several advanced techniques for better understanding of the learned features and the overall information captured by the network.
DATE: 27 July, 2016,Wednesday @ 13:30
PLACE: EA-409