COMING SOON! PQDT Open is getting a new home!

ProQuest Open Access Dissertations & Theses will remain freely available as part of a new and enhanced search experience at www.proquest.com.

Questions? Please refer to this FAQ.

Dissertation/Thesis Abstract

Inter-Slice Compression and Reconstruction of Glioma Magnetic Resonance Imaging (MRI) Data Using Encoder-Decoder Neural Networks
by Karr, Gavin Michael, M.S., Rose Hulman Institute of Technology, 2020, 66; 28028237
Abstract (Summary)

Magnetic Resonance Imaging (MRI) scans of patients with brain tumors are an important source of pre-surgical medical information. These three-dimensional image volumes can be represented as a stack of two-dimensional image slices.

The objective of this thesis is to compress the size of these image volumes by removing the odd-numbered slices and reconstruct the image volume using an encoder-decoder convolutional neural network. This neural network architecture is based on a modified form of the U-net segmentation network, which has been adjusted to allow for multiple image inputs and to support a network capable of generating new image slices. A novel method of performing slice interpolation is introduced in which the image features extracted by the neural network are averaged at each network layer to form the intermediary slice from the two input slices.

The MRI volume reconstruction performed by the encoder-decoder neural network is compared against linear interpolation of the image slices, and the metric used is the peak signal-to-noise-ratio. The reconstruction of the volume by the neural network slightly underperforms the linear interpolation baseline due to both methods being close to optimal in performance. Overall, the reconstruction quality of both methods is high since the initial slice distance causes little variation between adjacent slices. This thesis concludes that the neural network method of compression and reconstruction has potential in cases where inter-slice resolution is initially poor, such as at 4 millimeters and higher, while linear interpolation is sufficient at resolutions below 4 millimeters.

Indexing (document details)
Advisor: McInerney, Michael
Commitee: Bryan, Kurt, Boutell, Matthew
School: Rose Hulman Institute of Technology
Department: Electrical Engineering
School Location: United States -- Indiana
Source: MAI 82/2(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Artificial intelligence, Medical imaging, Electrical engineering
Keywords: Compression, Convolutional Neural Networks, Deep Learning, Medical Imaging
Publication Number: 28028237
ISBN: 9798662568853
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest