Doctor of Philosophy
Menon, Ravi S
High resolution functional MRI allows for the investigation of neural activity within the cortical sheet. One consideration in high resolution fMRI is the choice of which sequence to use during imaging, as all methods come with sensitivity and specificity tradeoffs. The most used fMRI sequence is gradient-echo echo planar imaging (GE-EPI) which has the highest sensitivity but is not specific to microvasculature. GE-EPI results in a signal with pial vessel bias which increases complexity of performing studies targeted at structures within the cortex. This work seeks to explore the use of MRI phase signal as a macrovascular filter to correct this bias.
First, an in-house phase combination method was designed and tested on the 7T MRI system. This method, the fitted SVD method, uses a low-resolution singular value decomposition and fitting to a polynomial basis to provide computationally efficient, phase sensitive, coil combination that is insensitive to motion. Second, a direct comparison of GE-EPI, GE-EPI with phase regression (GE-EPI-PR), and spin echo EPI (SE-EPI) was performed in humans completing a visual task. The GE-EPI-PR activation showed higher spatial similarity with SE-EPI than GE-EPI across the cortical surface. GE-EPI-PR produced a similar laminar profile to SE-EPI while maintaining a higher contrast-to-noise ratio across layers, making it a useful method in low SNR studies such as high-resolution fMRI. The final study extended this work to a resting state macaque experiment. Macaques are a common model for laminar fMRI as they allow for simultaneous imaging and electrophysiology. We hypothesized that phase regression could improve spatial specificity of the resting state data. Further analysis showed the phase data contained both system and respiratory artifacts which prevented the technique performing as expected under two physiological cleaning strategies. Future work will have to examine on-scanner physiology correction to obtain a phase timeseries without artifacts to allow for the phase regression technique to be used in macaques.
This work demonstrates that phase regression reduces signal contributions from pial vessels and will improve specificity in human layer fMRI studies. This method can be completed easily with complex fMRI data which can be created using our fitted SVD method.
Summary for Lay Audience
Functional MRI investigates brain function using the changing concentration of blood oxygen in the brain. This process has several pitfalls, one of which is inaccurate signals in large vessels far from the activating region caused by the pooling together of changing blood oxygen signals from many small vessels. One possible solution to these spurious signals is using a secondary imaging contrast from the MRI machine, the phase, to estimate these pooled signals and remove them. This technique has previously shown success in resting state and task based human studies. This thesis extends upon this work by investigating this technique at high resolution.
The first chapter of this thesis describes a method for the combination of phase data from a multi-coil radio-frequency array. High resolution fMRI requires a multi-coil radio-frequency array to acquire a high signal-to-noise image. These arrays require additional steps to create a high-quality phase image. This method estimates and corrects offsets for these arrays using data routinely acquired throughout the imaging session.
The second chapter of this thesis goes on to investigate phase regression at high resolution in a visual task. This chapter compares images collected with and without phase regression as well as a control image technique sensitive to small vessels. The findings show that fMRI with phase regression resulted in less pooled signal in the observed activation while retaining a higher contrast-to-noise ratio than the control condition.
The third chapter of this thesis details the study of phase regression in macaques during resting state. Macaque fMRI also contains pooled signal and is a common model for high resolution imaging studies. Phase regression did not perform well due to the presence of system and breathing noise in the acquired images. Several recommendations pertaining to quality determination are discussed to improve this experiment and phase regression studies generally.
Overall, this thesis extends the use of phase regression to high-resolution human fMRI and designed a multi-coil combination method for this application. A pilot of this procedure in animals was completed but requires further correction for phase artifacts, like system and breathing noise.
Stanley, Olivia W., "Phase imaging for reducing macrovascular signal contributions in high-resolution fMRI" (2021). Electronic Thesis and Dissertation Repository. 8167.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.