UCSF

Magnetic resonance imaging (MRI) is a powerful non-invasive medical imaging modality, but imaging speed remains a fundamental limitation to many potential applications.

Using non-Cartesian trajectories to efficiently traverse k-space has been a promising research direction for rapid MRI. The ability to traverse k-space rapidly is also beneficial for applications where the signal is short-lived.

On the other hand, non-Cartesian trajectories are not widely adopted clinically because they are often sensitive to many error sources, such as gradient delays and eddy currents. Hence, it is crucial to develop non-Cartesian imaging methods that provide both rapid k-space traversal and robustness to potential system errors. This dissertation considers two particular applications that require rapid and robust imaging methods: hyperpolarized C-13 MR spectroscopic imaging (MRSI) where the polarized signal is short-lived, and pulmonary imaging where the transverse magnetization decays rapidly.

For hyperpolarized C-13 MRSI, this dissertation proposes using a concentric rings trajectory (CRT) as the image acquisition method. CRT provides the essential scan time efficiency and robustness to system imperfections: (1) acquisition time is halved compared with Cartesian counterparts; (2) CRT is inherently robust to first-order eddy currents and gradient system delays; (3) CRT results in low noise amplification for parallel imaging. Preclinical studies demonstrate the efficacy of using CRT in hyperpolarized C-13 MRSI.

For pulmonary imaging, this dissertation proposes using a 3D radial based ultrashort echo time (UTE) imaging sequence, combined with a novel reconstruction framework to provide motion robust high resolution 3D images. In particular, this dissertation develops: (1) a robust and quantitative dynamic 3D self-navigator; (2) image reconstruction techniques to compensate for respiratory motion with soft-gating and motion-resolved strategies. Both techniques are further incorporated with parallel imaging and compressed sensing.

Despite its motion robustness and fast acquisition, the proposed pulmonary imaging method is susceptible to gradient delays. Motivated by this, a general solution for correcting gradient system delays is proposed by exploiting data redundancy in multi-channel data. The proposed method requires no additional calibration scans, and estimates both gradient delays and auto-calibration data simultaneously. This work is general to many non-Cartesian trajectories as validated by both simulation and in vivo scans.