Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Medical Biophysics

Supervisor

Dr. Ravi Menon

Abstract

Performing magnetic resonance imaging at high magnetic field strength promises many improvements over low fields that are of direct benefit in functional neuroimaging. This includes the possibility of improved signal-to-noise levels, and increased BOLD functional contrast and spatial specificity. However, human MRI at 7T and above suffers from unique engineering challenges that limit the achievable gains. In this thesis, three technological developments are introduced, all of which address separate issues associated with functional magnetic resonance neuroimaging at very high magnetic field strengths.

First, the image homogeneity problem is addressed by investigating methods of RF shimming — modifying the excitation portion of the MRI experiment for use with multi-channel RF coils. It is demonstrated that in 2D MRI experiments, shimming on a slice-by slice basis allows utilization of an extra degree of freedom available from the slice dimension, resulting in significant gains in image homogeneity and reduced RF power requirements.

After acceptable images are available, we move to address complications of high field imaging that manifest in the fMRI time series. In the second paper, the increased physiological noise present in BOLD time series at high field is addressed with a unique data-driven noise regressor scheme based upon information in the phase component of the MRI signal. It is demonstrated that this method identifies and removes a significant portion of physiological signals, and performs as good or better than other popular data driven methods that use only the magnitude signal information.

Lastly, the BOLD phase signal is again leveraged to address the confounding role of veins in resting state BOLD fMRI experiments. The phase regressor technique (previously developed by Dr. Menon) is modified and applied to resting state fMRI to remove macro vascular contributions in the datasets, leading to changes in spatial extent and connectivity of common resting state networks on single subjects and at the group level.


Share

COinS