Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Science


Medical Biophysics


Menon, Ravi S.


High-resolution fMRI using gradient-echo blood-oxygen-level-dependent (BOLD) contrast is beneficial for the non-invasive study of neural microcircuits. However, the signal spatial specificity of the BOLD contrast severely limits the ability to localize regions of neural activity at the mesoscopic scale in the cortex due to signal contamination from large veins. Phase regression is a venous bias correction technique that uses the correlation between magnitude and phase data in large veins to estimate and supress their contribution to the BOLD signal. This thesis further investigates the performance of phase regression by examining the laminar BOLD signal in human ocular dominance columns. Phase regression removes the venous bias from pial veins and large intracortical veins, while not removing the venous bias from venous vessel sizes within the cortex running parallel to the cortical surface. This thesis demonstrates improved laminar BOLD signal specificity that will be beneficial in future high-resolution laminar fMRI studies.

Summary for Lay Audience

Functional MRI is a popular non-invasive imaging modality that relies on changes in the concentration of blood oxygenation to map changes in neural activity associated with brain function. Neural activity is energy intensive and requires oxygen, leading to changes in local blood oxygenation in areas of neural activity. A major problem with this technique is that blood oxygenation changes are most prominent in large veins because blood drains away from many small vessels in activated regions and pools in fewer large veins. Large veins are more distant from the activated regions, meaning that they are not an accurate measure of blood oxygenation changes caused at the actual site of neural activity.

One technique that attempts to remove the venous bias from the signal is called phase regression and it relies on phase data. MRI acquisitions result in complex-valued data, which is commonly represented as magnitude and phase images, with the phase data typically being discarded. In phase regression, the phase data is used to estimate and suppress the venous bias from the signal. This helps ensure the measured signal is more spatially specific to the site of neural activity.

A laminar and columnar analysis was performed in ocular dominance columns in the human primary visual cortex to investigate the performance of phase regression for high-resolution functional MRI. Ocular dominance columns are vertical columns across the cortex with alternating sensitivity to right and left eyes. They also contain varying amounts of signal exchange between columns at different cortical depths (laminae). This makes ocular dominance columns well suited for performing a laminar and columnar analysis assessment as they are a cortical structure with relatively well-known mesoscopic functions.

It was shown that laminar signal profiles across cortical depths were improved by phase regression. However, phase regression did not help clearly define ocular dominance columns. This is convincing evidence that phase regression is only effective for the largest veins, and not smaller venous vessels within the cortex running parallel to the cortical surface. Overall, this thesis demonstrates that phase regression will be a useful tool for high-resolution functional MRI studies performing laminar analyses.