University of Illinois Urbana-Champaign

MRI Registration and Segmentation for Machine Learning Diagnosis of Parkinson's Disease

Zeng, David

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/55634

Description

Title
MRI Registration and Segmentation for Machine Learning Diagnosis of Parkinson's Disease
Author(s)
Zeng, David
Contributor(s)
Liang, Zhi-Pei
Issue Date
2014-05
Keyword(s)
  • image registration
  • image segmentation
  • Parkinson’s disease
  • MRI
Abstract
Background: We have been provided MRI data of 10 healthy patients and 10 patients with Parkinson’s disease. Each of the patients has images in T1, T2, susceptibility-weighted, and diffusion kurtosis contrast schemes. With the opportunity this large dataset presents, our project is to develop a machine learning-based diagnosis of Parkinson’s disease. Methods: The first part of the project is to register the images. This is accomplished using Statistical Parametric Mapping (SPM8). The second part is to segment the images. Segmentation is performed in two ways: SPM8 segments the image by tissue type and a ramp transform method segments the image by features. The third part is to develop a machine learning algorithm to diagnose Parkinson’s disease. In this thesis, the first two parts are addressed. Results: The image registration in SPM8 for healthy brains passes visual inspection. For brains with Parkinson’s disease, there are concerns about distortions in the diseased areas of brains. Segmentation is performed on registered images so artifacts in registration propagate to segmentation. Segmentation by SPM8 of healthy brains passes visual inspection and segmentation of diseased brains have effects resulting from registration. Results from segmentation by the ramp transform have raised several issues on the limitations of the method and further work is needed to adapt the method. Conclusions: The registration and segmentation parts of the project have had successful results for images of healthy brains. There are reservations on the effectiveness of the image processing on brains with Parkinson’s disease but it will require development of the machine learning algorithm to fully evaluate the effects of the image processing.
Type of Resource
text
Language
en
Permalink
http://hdl.handle.net/2142/55634

Owning Collections

Senior Theses - Electrical and Computer Engineering PRIMARY
The best of ECE undergraduate research