Files in this item



application/pdfNGO-DISSERTATION-2018.pdf (17MB)
(no description provided)PDF


Title:Model-based reconstruction for correcting magnetic susceptibility-induced artifacts in magnetic resonance imaging
Author(s):Ngo, Giang-Chau
Director of Research:Sutton, Bradley P.
Doctoral Committee Chair(s):Sutton, Bradley P.
Doctoral Committee Member(s):Insana, Michael; Liang, Zhi-Pei; Sadaghiani, Sepideh Friberg
Department / Program:Bioengineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Magnetic Resonance Imaging
Magnetic susceptibility
Magnetic field inhomogeneity
Model-based imaging reconstruction
Functional MRI
R2* estimation
Quantitative susceptibility mapping
Abstract:While magnetic susceptibility is an essential contrast mechanism in magnetic resonance imaging (MRI), it also causes significant disruptions to the imaging process. Strong differences in magnetic susceptibility exist at the air and tissue interfaces in the brain, resulting in a magnetic field distribution across the brain which is called the background magnetic field. Although many methods have been developed to correct the image distortions and signal loss due to the background magnetic field, there has not been a thorough analysis of the impact of the spatial gradients of this background magnetic field on MR imaging. This dissertation demonstrates, models, and corrects the important effects of the background magnetic field and the corresponding susceptibility-induced magnetic field gradients (SI-MFGs) in MRI. The background SI-MFGs have particularly a strong impact on applications using gradient echo acquisitions, such as functional MRI (fMRI) and R2* mapping. In this work, SI-MFGs are shown to lead to a non-uniform fMRI sensitivity through the brain due to induced echo time shifts. A method based on estimated background magnetic field maps and SI-MFG maps is used to predict the impact of SI-MFGs on the fMRI sensitivity. Using the proposed approach, the potential bias introduced by the background SI-MFGs is evaluated for an aging study, by looking at the magnetic field distribution differences between younger and older subjects. Although aging introduces changes in the magnetic field distribution across the brain, I demonstrate that the magnitude of the changes between the two groups is small, and thus no measurable age-related bias in fMRI contrast due SI-MFGs is expected at 3 T. Background SI-MFGs also lead to bias in R2* mapping. R2* values are crucial for applications such as quantitative functional MRI and relative cerebral blood volume estimation. Therefore, for these applications, it is important to obtain R2* estimates which are not dependent on the background magnetic field distribution. However, background SI-MFGs can lead to an overestimation of R2* values if standard estimation approaches are used. A model capturing the SI-MFG effects on the R2* decay is developed and implemented in this work. A novel k-space-based method using this signal model and a joint estimation framework is developed to obtain R2* estimates free of SI-MFG bias. I also demonstrate that accurate modeling of the background magnetic field effects also enables faster acquisitions for quantitative susceptibility mapping. Fast imaging can be achieved by using long readout imaging but at the expense of increased artifacts induced by the background magnetic field. Correcting for the background field effects is crucial to provide accurate magnitude and phase images. In this work, a model-based reconstruction framework is combined with an efficient 3D spiral-in acquisition for susceptibility mapping. Susceptibility maps free of background magnetic field artifact are obtained at a resolution of 1 x 1 x 1 mm3 in 46 s, which is 17 times faster than a standard gradient echo spin-warp acquisition. This approach can further enable applications such as functional susceptibility mapping but also improve patient comfort by considerably decreasing the time spend in a scanner. Overall, this thesis develops models of the background magnetic field and its spatial gradients and uses those models to reconstruct accurate images of brain function and tissue susceptibility. I also demonstrate that this imaging approach enables fast and more accurate quantitative imaging.
Issue Date:2018-02-27
Rights Information:Copyright 2018 Giang-Chau Ngo
Date Available in IDEALS:2018-09-04
Date Deposited:2018-05

This item appears in the following Collection(s)

Item Statistics