Advanced high resolution diffusion tensor magnetic resonance imaging

Abstract

As a non-ironizing medical imaging method, MRI offers the visualization of biological tissues. However, the long acquisition time and spatial resolution confine the application of clinical MRI. In the mid-1980s, diffusion-weighted (DW) imaging was proposed to probe tissue microstructures by the molecular motion of water. However, high isotropic resolution DW imaging is hindered by the long acquisition time and low SNR. This thesis aims to design the methods to improve the acquisition efficiency and reconstruction performance for both 2D and 3D multi-shot DW Echo Planar Imaging (EPI) data. This purpose is hampered by many problems in sequence design and reconstruction. In this thesis, these problems are separated into five designed projects with different emphases and solved step by step. Nyquist ghost is a unique artifact in EPI, while its correction is restricted by shot number in parallel imaging reconstruction. In the first project, an alternated acquisition pattern and a MUSE based correction method are proposed for self-navigated Nyquist ghost measurement, which is also implemented in multi-band excitation. Another artifact of DW-EPI data is geometric distortion. The second project correction is designed to correct the off-resonance effect and eddy current effect for DW-EPI data. As a consequence, the self-navigated Blip Up-Down Acquisition is extended to a nonuniform undersampling in phase encoding direction. Correspondingly, the joint combined correction method is proposed to solve this phase accumulation in the k-space domain problem. In the third project, a reconstruction project for the self-navigated 2D DW PROPELLER acquisition is designed. The reconstruction model is based on the 2D spatial encoding of k-space data with aliasing and distortion artifacts correction. Correspondingly, an alternated phase encoding acquisition is designed for the proposed reconstruction method to improve efficiency and performance. To efficiently implement this reconstruction model for PROPELLER-EPI data, a preconditioned nonlinear iterative SENSE method is designed for accelerating the convergence of PROPELLER-EPI reconstruction with less memory consumption. In the fourth project, a high-performance reconstruction method for 3D multislab DW-EPI data with slab boundary artifact correction is proposed. To achieve the self-referenced slab profile estimation, the sliding-slab technique is utilized. The 3D diffusion-weighted SE-EPI 2D navigator-based sequence is developed on a Philips commercial scanner. Besides, the frequency of the RF pulse is varied in different interleave loops to achieve the sliding-slab pattern. In the conventional 3D DW data reconstruction process, reconstruction time is relatively long. Therefore, this project proposes a two-step reconstruction method to improve the efficiency of reconstruction. To inherently avoid the RF pulse induced artifacts of 3D DW imaging, the 3D single-slab DW-EPI acquisition can be used. However, the conventional 3D MUSER method requires long acquisition and reconstruction time. As a consequence, in the fifth project, the 3D single-slab DW-EPI data is undersampled by the pseudo-random pattern to accelerate acquisition. Correspondingly, the compressed sensing method is used in the reconstruction process to suppress artifacts. The preconditioned iterative SENSE method is used to accelerate the reconstruction process. Altogether, these projects comprise the theme of diffusion-weighted imaging and may open new pathways toward fast and high-performance clinical MRI.