Diffusion MRI: Aspects of Reproducibility and Novel Segmented 2D and 3D Approaches for Higher Resolution and Geometric Fidelity

Abstract: Diffusion-weighted imaging (DWI) is a very useful tool for non-invasive imaging and clinical investigation of the human body. It provides data of the water diffusion process. Applied to biological tissues in a clinical setting it provides information that can aid diagnosis and characterization of lesions, as well as longitudinal monitoring of disease, mapping of fiber architecture for surgical planning, and much more. While DWI is a valuable clinical tool, it suffers from several limitations such as low spatial resolution, image distortion and inaccurate quantification of the diffusion process. These limitations prevent the DWI method to reach its full clinical potential. The work presented in this thesis aims to investigate limitations associated with DWI and develop novel methods to reduce them. The first study compared a new multi-shot DWI method called MUSE to a conventional single-shot approach in a clinical setting for a pediatric population (n=14). Improved resolution, geometric fidelity and image quality was demonstrated using MUSE, both with and without a method termed reverse polarity gradient acquisition. The second study investigated the influence of acquisition parameters on the apparent diffusion coefficient (ADC) using the standard clinical PGSE sequence on a modern clinical scanner with high gradient performance. Diffusion time, echo time and gradient amplitude dependence of the ADC was explored in healthy brain (n=10) and in brain tumors (n=10). Only minor ADC changes were observed in healthy tissue at group level. For tumors, however, the observed ADC changes with diffusion-time and/or echo time exceeded the repeatability coefficient in several individual subjects. This suggests that such dependencies should be considered when using the ADC as a quantitative biomarker. Even though considerable effort has been devoted by others to improve the image resolution and geometric fidelity in DWI, there is still improvement needed to reach the overall quality that is considered the standard in conventional T1- and T2-weighted imaging. The third and fourth study aimed to develop and test a novel DWI approach for 3D-DWI in healthy brain. A 3D imaging approach allows for high isotropic spatial resolution at superior signal-to-noise ratio (SNR) compared to its 2D counterpart. In Study III, single thick-slab 3D-DWI was demonstrated using motion-compensated diffusion encoding gradients to minimize shot-to-shot motion induced phase errors at the source. In Study IV, this sequence was further improved by incorporating three orthogonal 1D phase navigators to monitor and correct for phase variations stemming from higher order motion terms.