Degree
Doctor of Philosophy (PhD)
Department
Physics and Astronomy
Document Type
Dissertation
Abstract
Diffusion Magnetic Resonance Imaging (dMRI) measures the diffusion of water within the body and can be used for inspecting white matter tracts within the brain. dMRI data can be used for estimation of white matter fiber orientations as well as for estimating measures of tract integrity. In order to utilize state-of-the-art analysis techniques, large quantities of dMRI data are necessary, and scanning times have become a limiting factor in the implementation of dMRI clinically. In this dissertation, we propose machine learning methods to reduce the amount of data that needs to be acquired at the scanner. In chapter 2, we have proposed a model that estimates dMRI data with high levels of diffusion weighting from volumes with a lower level of diffusion weighting. In chapter 3, we applied this model to data collected from the Bogalusa Heart Study and analyzed lifespan effects of cardiometabolic health on white matter health in midlife. In chapter 4, we proposed a second machine learning model that improves the angular resolution of dMRI data by mapping the spherical harmonic representation of dMRI data acquired with a small number of diffusion weighting directions to the spherical harmonic representation of the same volume with a greater number of diffusion weighting directions. The proposed models enable data acquired in a short period of time to be used with advanced analysis methods which would normally require significantly more scan time and provide more sensitive measures of white matter health.
Date
4-1-2024
Recommended Citation
Dugan, Reagan T., "AUGMENTATION AND ANALYSIS OF DIFFUSION MRI DATA VIA MACHINE LEARNING METHODS" (2024). LSU Doctoral Dissertations. 6393.
https://repository.lsu.edu/gradschool_dissertations/6393
Committee Chair
Carmichael, Owen