Functional Magnetic Resonance Spectroscopic Imaging: Method Development and Applications

Magnetic resonance spectroscopy (MRS) is a non-invasive method to study metabolic processes in vivo. Functional MRS (fMRS) extends this capability to probe neurochemical dynamics during brain activation, complementing fMRI by assessing metabolic responses underlying neural activation. Glutamate (Glu), the major excitatory neurotransmitter, is of particular interest. However, conventional fMRS is restricted to single-voxel acquisitions, limiting its ability to investigate spatially resolved responses. Magnetic resonance spectroscopic imaging (MRSI) provides information about the spatial distribution of metabolites. Conventional phase-encoded MRSI suffers from long acquisition times, limiting its applications in functional studies. Recent advances in acceleration techniques, such as spatial spectral encoding and k-space undersampling, make functional MRSI (fMRSI) feasible with improved spatial coverage and temporal resolution. In this context, fast and reliable 2D/3D 1H-MRSI acquisition and reconstruction pipelines with rosette trajectories were first developed. An optimized water suppression method, FAST (Five-variable-Angle Gaussian pulses with Short Total duration), was proposed to improve insensitivity to B1+ inhomogeneity. Lipid signals were removed with an adaptive L2-regularization method. The 2D implementation achieved 4.48x4.48 mm2 in-plane resolution (10 mm thick) with 22 s per average, and the 3D implementation provided whole-slab coverage (224x224x90 mm3; 50x50x20 matrix) in 5:38 min. Retrospective compressed sensing showed potential to reduce scan time to ~2 min. Reproducibility analysis showed intra-session CVs below 6% for major metabolites, including Glu. Building on this framework, the feasibility of fMRSI was investigated with visual stimulation, targeting Glu response mapping. A block design with interleaved 5-min REST and STIM periods was employed. The 2D sequence, adapted to 30s temporal resolution, yielded 50 time points for the 25 min paradigm, enabling application of statistical analyses commonly used in fMRI. General linear model (GLM) and independent component analysis (ICA) consistently revealed Glu-activated clusters in the occipital cortex across participants, but not in NAA, tCr, or linewidth. A mean Glu increase of 5-8% was detected, corresponding to an estimated increase rate of ~0.26 mM/min and ~17% rise in CMRO2 during activation. These values are consistent with the reported 7-30% increase upon focal visual stimulation in previous fMRI and PET studies. In addition, as a technical pilot towards future energy-related studies, a 31P diffusion-weighted STEAM sequence was developed. With TE = 15 ms, the apparent diffusion coefficients (ADC) of phosphocreatine (PCr), ATP, Pi, and GPC were successfully quantified in the human calf muscle within 40 min. The ADCs of ATP, Pi, and GPC were reported for the first time in humans, while PCr values matched literature. This study demonstrates the feasibility of 31P diffusion MRS