Longitudinal Relaxation Dynamics in White Matter: Experiments in NMR and MRI

MRI-based assessments of the human brain are critical for research, diagnosis and treatment of neurological disorders. Future clinical practice will demand accurate and consistent quantitative methodology alongside today’s qualitative image evaluations. Consequently, MRI research focuses on developing physical understanding of prevalent techniques and establishing new methods for efficient quantitative analysis. The brain’s complex structure complicates this goal. Myelin, a lipid-rich tissue aiding in signal transmission along axons in white matter, creates useful image contrast but also complicates the interpretation of measurements using canonical models. In this dissertation, we conduct experiments using solid-state NMR and in vivo MRI to examine assumptions in current MRI methods leading to potential quantitive errors. We propose improvements to these methods and an adaptation to an existing model. 

A straightforward yet effective view of white matter is to separate proton populations into two pools, aqueous and non-aqueous, between which protons, and therefore magnetization, can exchange. This transfer can significantly affect the measured longitudinal relaxation (T1) depending on the preparation of each pool. First, we study the impact of adiabatic inversion pulses applied to white matter through NMR experiments on ex vivo brain samples and then compare these results to analogous in vivo experiments. We demonstrate that, contrary to common assumption, the non-aqueous pool is not saturated by typical adiabatic inversion pulses, although the aqueous pool is fully inverted, which results in bi-exponential longitudinal recovery. We compare this relaxation to that following hard and selective pulses, which are understood to result in mono- and bi-exponential recovery, respectively. Next, we perform NMR experiments on ex vivo brain samples using hard and selective pulses to initiate magnetization transfer demonstrating similar T1 biasing effects during Look-Locker and Variable Flip Angle sequences. We evaluate sources of systematic error pertinent to MRI applications. Finally, we modify the canonical Bloch-McConnell equations describing two-pool relaxation to incorporate fractional-order derivatives. We examine a numerical solution and provide an approximate analytical solution, which we use to model inversion recovery in heterogenous model systems, ex vivo, and in vivo brain. An additional fit parameter is introduced which may be used as a new contrast source.