Date of Award

Summer 2019

Document Type


Degree Name

Doctor of Philosophy (PhD)


Electrical & Computer Engineering

Committee Director

Chung-Hao Chen

Committee Member

Gene Hou

Committee Member

Jiang Li

Committee Member

Chunsheng Xin


This dissertation presents a computer-aided diagnosis (CAD) system using deep learning approaches for lesion detection and classification on whole-slide images (WSIs) with breast cancer. The deep features being distinguishing in classification from the convolutional neural networks (CNN) are demonstrated in this study to provide comprehensive interpretability for the proposed CAD system using the domain knowledge in pathology. In the experiment, a total of 186 slides of WSIs were collected and classified into three categories: Non-Carcinoma, Ductal Carcinoma in Situ (DCIS), and Invasive Ductal Carcinoma (IDC). Instead of conducting pixel-wise classification (segmentation) into three classes directly, a hierarchical framework with the multi-view scheme was designed in the proposed system that performs lesion detection for region proposal at higher magnification first and then conducts lesion classification at lower magnification for each detected lesion. The majority voting scheme was adopted to improve the error tolerance of the system in lesion-wise prediction. For all collected 186 slides, the slide-wise prediction accuracy rate strikes to 95.16% (177/186) in binary classification to predict carcinoma (malignant) or non-carcinoma (benign), and the sensitivity for cases with carcinoma reaches 96.36% (106/110). In multi-classification, the accuracy rate is 92.47% (172/186) when predicting Non-Carcinoma, DCIS, and IDC for each slide. Most importantly, the interpretability for the mechanism of the proposed CAD system is provided from the pathological perspective. The experimental results show that the morphological characteristics and co-occurrence properties learned by the deep learning models for lesion detection and classification meet the clinical rules in diagnosis. Accordingly, the pathological interpretability of the deep features not only enhances the reliability of the proposed CAD system to gain acceptance from medical specialists, but also facilitates the development of deep learning frameworks for various tasks in pathology.


In Copyright. URI: This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).