High resolution magnetic resonance spectroscopic imaging of the brain from 3T to 7T

The objective of this thesis research is to develop fast high-resolution magnetic resonance spectroscopic imaging (MRSI) methods of the brain at both high field (3 Tesla) and ultrahigh field (7 Tesla) MR. MRSI has long been regarded as a promising tool for non-invasive imaging of brain metabolism, with its potential shown in a large range of applications including energy consumption analysis, brain functional investigation, brain lesions characterization like tumors and stroke, neurodegeneration assessment, etc. However, its practical utility has been largely limited by several long-standing technical obstacles, including low signal-to-noise-ratio (SNR), long scan time, and limited resolution. Given these limitations, it has been very challenging to perform three-dimensional high-resolution MRSI experiments in a clinically feasible time. In recent years, a subspace-based MRSI method named as SPICE (SPectroscopic Imaging by exploiting spatiospectral CorrElation) has been proposed for accelerated MRSI. A number of proof-of-concept works using the basic SPICE method for 1H-MRSI at 3T have demonstrated its potential and advantages in significantly advancing imaging speed, resolution, and SNR. However, imaging capability of the basic SPICE method still cannot satisfy the ever-growing clinical needs. The proposed research further developed the SPICE method with multiple aspects of improvements, which significantly enhanced its imaging capability and made it a more practically powerful and clinically useful MRSI tool. In data acquisition, a novel pulse sequence was developed at 3T with multiple unique acquisition features for high imaging efficiency and robustness. More specifically, the proposed sequence used an FID (free induction decay)-based acquisition with ultrashort TE (echo time) and short TR (repetition time) for maximized SNR efficiency, removed water and lipid suppression pulses for minimized energy deposition, and employed fast spatiotemporal trajectories and highly sparse sampling of data space for fast imaging speed. As a result, three-dimensional metabolite signals (field of view: 240×240×72 mm3) at 2.0×3.0×3.0 mm3 nominal spatial resolution and unsuppressed water signals at 2.0×1.0×1.0 mm could be successfully acquired in an 8-minute scan. This method was also implemented and further developed at 7T MR systems. Taking advantage of the SNR benefit of ultrahigh field, the pulse sequence at 7T was pushed to achieve whole brain coverage (field of view: 240×240×160 mm3), high-resolution (3.0×3.0×3.2 mm3 for metabolites and 2.0×2.0×3.2 mm3 for water) imaging within the same scan time (8 minutes) via using a rapid spatiotemporal readout on slice direction and employing a higher sparse sampling strategy. Given these data acquisitions, the key processing and reconstruction issues included (1) separation of signals from water, lipid, and metabolites, (2) reconstruction from the sparse and noisy measurements, and (3) correction of effects and artifacts brought by various system imperfections. The first two processing issues were addressed using a union-of-subspaces model with subspace learning strategies, and the last issue was resolved by utilizing information derived from the unsuppressed water signals and several navigator signals embedded in the data acquisition. To demonstrate the feasibility and performance of the proposed method, in vitro experiments and in vivo experiments were carried out on a standard spectroscopic phantom and healthy volunteers, respectively. Given those technical advances, the proposed method successfully showed good accuracy and reproducibility in phantom experiments and obtained high-quality, high-resolution brain metabolite maps from healthy subjects. Moreover, the presented method was also applied for clinical tumor imaging to demonstrate its values in clinical environments. The feasibility studies showed its impressive imaging capability in capturing metabolic alterations in small-size tumors, imaging intra-tumor heterogeneity, classifying tumors with different grades, and monitoring therapeutic responses. In addition, the improved SPICE method was used for 31P-MRSI at 7T for high resolution mapping of high-energy metabolites. Experimental results also showed significantly advanced performance of the proposed method in generating high-resolution, high-quality 31P metabolite maps. In this thesis research, the feasibility of fast high-resolution MRSI at both 3T and 7T was successfully demonstrated. The described method is expected to provide a very powerful imaging tool in practical environments for a wide range of scientific and clinical applications.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Guo, Rong

Contributors dc:contributor

• Liang, Zhi-Pei
• Sutton, Brad
• Oelze, Michael L
• Yu, Xin

• MRSI
• subspace modelling
• high field
• ultrahigh field
• sequence design
• reconstruction.

Statement dc:rights

• Copyright 2022 Rong Guo

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/115880

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F115880