Committee Members:
Xiaoping Hu, Ph.D., Advisor (Georgia Tech and Emory University)
Xiaoming Huo, Ph.D. (Georgia Tech)
Shella Keilholz, Ph.D. (Georgia Tech and Emory University)
Helen S. Mayberg, M.D. (Emory University)
John Oshinski, Ph.D. (Georgia Tech and Emory University)

Functional magnetic resonance imaging (fMRI) based on blood oxygenation level dependent (BOLD) contrast has become a widespread technique in brain research. The central challenge in fMRI is the detection of relatively small activity-induced signal changes (< 3%) in the presence of various other signal fluctuations. Physiological fluctuations due to respiration and cardiac pulsation are dominant sources of confounding variability in BOLD fMRI. This dissertation seeks to characterize and compensate for non-neural physiological fluctuations in fMRI.

First, the dissertation presents an improved and generalized technique for correcting T1 effect in cardiac-gated fMRI data incorporating flip angle estimated from fMRI dataset itself. Using an unscented Kalman filter, spatial maps of flip angle and T1 relaxation are estimated simultaneously from the cardiac-gated time series. Accounting for spatial variation in flip angle, the new method is able to remove the T1 effects robustly, in the presence of significant B1 inhomogeneity. The technique is demonstrated with simulations and experimental data.

Recent fMRI studies have revealed that significant correlation between fMRI change and physiological change such as cardiac rate and respiratory volume change have been observed across widespread gray matter regions, even after removing short-term physiological effects. This dissertation describes a generalized retrospective technique to precisely model and remove physiological fluctuations from fMRI signal: Physiological Impulse Response Function Estimation and Correction (PIRFECT). Application of the proposed technique is observed to increase the intra-session reproducibility of resting-state networks.