In the central nervous system, axon and myelin membranes are physical barriers to water diffusion. As a consequence, diffusion weighted magnetic resonance imaging (DWI) is uniquely well suited to study the changes in tissue microstructure due to neurological disease or injury. However the conclusive assignment of changes in measured diffusion properties to axon and/or myelin damage has not been straightforward. This has been hampered, in part, by the fact that axon and myelin damage are often histopathologically linked, and that diffusion tensor imaging (DTI) implicitly assumes free diffusion (i.e., as described by a Gaussian distribution).

The primary focus of this dissertation is the development of q-space DWI techniques to study water diffusion in the human and rat spinal cord, both model systems for the study of restricted diffusion. Q-space analysis computes the probability density function for molecular diffusion and provides a more comprehensive and biophysically rigorous means to quantify diffusion than conventional DTI. The specific aims of this research are to improve the understanding of diffusion biophysics in the central nervous system, and thereby develop DWI techniques that are more sensitive and specific to axon and myelin damage.

This is accomplished by first investigating the effect of signal to noise on the accuracy and reproducibility of in vivo DTI in the human brain. Low signal to noise is shown to introduce significant bias into diffusivity and anisotropy measurements, and guidance is provided for the design and interpretation of DTI experiments performed at multiple imaging sites. Secondly, the feasibility of in vivo q-space DWI of the human spinal cord is demonstrated in healthy volunteers and patients with multiple sclerosis. Q-space contrasts are shown to be more sensitive to the abnormal water diffusion in damaged white matter than conventional diffusivity measurements. Thirdly, ex vivo q-space DWI was performed in the rat spinal cord after dorsal root axotomy to investigate the effects of axon and myelin damage on water diffusion perpendicular and parallel to the long axis of the spinal cord. Comparisons with histology show that an increase in perpendicular diffusion is not specific to myelin damage; however a decrease in parallel diffusion may be a specific marker for axonal damage. Furthermore, measurements of the diffusion kurtosis excess (KE) (i.e., the deviation from Gaussian diffusion) show that KE is sensitive to white matter damage. Specifically, the decrease in perpendicular KE implies that white matter damage decreases the barriers to perpendicular diffusion, whereas the increase in parallel KE suggests that axonal injury results in more restricted diffusion in the parallel direction. The effect of the maximum q-value on the accurate quantification of the KE of free and restricted diffusion is also elucidated, which has important ramifications for the adoption of the KE metric to study non-Gaussian diffusion in biological tissues. Finally, a Monte Carlo simulation study of diffusion in impermeable compartments suggests that bulges of the axonal membrane may account for the decrease of parallel diffusivity and increase in KE noted in DWI studies of animal models of axonal damage.