Dissertation/Thesis Abstract

Segmentation of Epicardial Adipose Tissue in Cardiac MRI Using Deep Learning
by Fulton, Miranda, M.S., Southern Illinois University at Edwardsville, 2020, 54; 27957747
Abstract (Summary)

Heart Disease is a major cause of death in the developed world. The term heart disease refers to any abnormality of the cardiovascular system: this can include stroke, heart failure, coronary artery disease and heart attack. A key aspect of treating the disease is a clear image of what is happening within the heart. This is achieved using a variety of imaging modalities such as CT, ultrasound and MRI. While these modalities each have their pros and cons, MRI has been shown to be a reliable way to identify and diagnose issues related to heart health. The advantages of MRI include no radiation or ionizing contrast exposure, while still providing a clear image of the heart and other organs over time. Thanks to advances in computer vision there has been a recent effort to develop segmentation algorithms to quantify the various tissue types and how they change over time. While the primary focus of these efforts have targeted the muscle of the heart, recent research has called new attention to the fat that surrounds the heart and its link to coronary heart disease. This fat has proven difficult to examine in a practical way. The structure of the fat is irregular and follows an unclear trend for how it is distributed around the heart muscle. The work presented here attempts to segment the cardiac fat using deep learning. More specifically, both a traditional Neural Network (NN) and a Fully Convolutional Network (FCN) were investigated. FCNs are a type of artificial intelligence (AI) that uses convolution to predict the location of an object within an image. The use of this AI is a newer development that has been shown to be valuable in object identification. This particularly applies in cases where the application requires a mask to extract the feature vector for classification. FCNs can produce the mask to then be used to generate the feature vector which then gives the necessary information for classification to occur through a variety of methods.

Indexing (document details)
Advisor: Klingensmith, Jon D.
Commitee: Umbaugh, Scott, Wang, Yadong
School: Southern Illinois University at Edwardsville
Department: Electrical and Computer Engineering
School Location: United States -- Illinois
Source: MAI 81/12(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Medical imaging, Artificial intelligence
Keywords: Artificial intelligence, Cardiac imaging, Magnetic resonance imaging, Medical imaging
Publication Number: 27957747
ISBN: 9798645497620
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest