Year of Publication

2017

Degree Name

Master of Science (MS)

Document Type

Master's Thesis

College

Engineering

Department

Computer Science

First Advisor

Dr. Jinze Liu

Abstract

Mammography is the most widely used method of screening for breast cancer. Traditional mammography produces two-dimensional X-ray images, while advanced tomosynthesis mammography produces reconstructed three-dimensional images. Due to high variability in tumor size and shape, and the low signal-to-noise ratio inherent to mammography, manual classification yields a significant number of false positives, thereby contributing to an unnecessarily large number of biopsies performed to reduce the risk of misdiagnosis. Achieving high diagnostic accuracy requires expertise acquired over many years of experience as a radiologist.

The convolutional neural network (CNN) is a popular deep-learning construct used in image classification. The convolutional process involves simplifying an image containing millions of pixels to a set of small feature maps, thereby reducing the input dimension while retaining the features that distinguish different classes of images. This technique has achieved significant advancements in large-set image-classification challenges in recent years.

In this study, high-quality original mammograms and tomosynthesis were obtained with approval from an institutional review board. Different classifiers based on convolutional neural networks were built to classify the 2-D mammograms and 3-D tomosynthesis, and each classifier was evaluated based on its performance relative to truth values generated by a board of expert radiologists. The results show that CNNs have great potential for automatic breast cancer detection using mammograms and tomosynthesis.

Digital Object Identifier (DOI)

https://doi.org/10.13023/ETD.2017.364

Share

COinS