Acquisition and reconstruction methods for high-resolution, high-fidelity 3D multi-slab diffusion MRI

Thesis

Diffusion magnetic resonance imaging (dMRI) is a powerful tool for non-invasive mapping of brain microstructure and connectivity by measuring the random walk of water molecules. Conventional two-dimensional (2D) single-shot echo planar imaging (EPI) is widely used for dMRI due to its rapid acquisition and robustness against motion, but suffers from limitations such as long repetition times (TR) that reduce signal-to-noise ratio (SNR) efficiency and difficulty in achieving thin slice thickness, particularly for high-resolution imaging. To address these issues, three-dimensional (3D) multi-slab acquisition methods have been introduced, offering better SNR efficiency through shorter TR and the potential to achieve high, isotropic resolution dMRI for in-vivo human brain. However, key challenges remain for 3D multi-slab dMRI, including susceptibility to B₀ inhomogeneity induced geometric distortions that compromise image anatomical fidelity, inefficiencies in current navigator-based methods for correcting motion-induced phase variations, and the need for careful considerations of blurring and noise to achieve true high spatial resolution.

This thesis develops novel acquisition and reconstruction methods to overcome these hurdles. First, sampling strategies and a joint reconstruction for blip-up/down data are designed to correct geometric distortions and slab boundary aliasing for 3D multi-slab imaging without increasing the scan time. Second, a self-navigated 3D multi-slab dMRI framework is developed, eliminating the need for navigator acquisitions through optimised self-navigation sampling and structured low-rank reconstruction for motion phase correction, which effectively shortens scan time and improves SNR efficiency. Lastly, we leverage in-plane segmented 3D multi-slab EPI and a denoiser-regularised reconstruction to achieve 0.5-0.6 mm isotropic resolutions for in-vivo dMRI with superior SNR, minimal blurring and distortions. These advances are expected to improve the image fidelity and scan efficiency of high-resolution 3D multi-slab dMRI, with the hope to ultimately benefit neuroscience research of the human brain.

Authors: Li, Z

• DOI: 10.5287/ora-xrzapypby
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: MRI reconstruction; brain connectivity; pulse sequence; high resolution; diffusion MRI
• Subjects: Biomedical engineering; Diffusion magnetic resonance imaging