MR spectroscopic imaging (MRSI) is a powerful in vivo molecular imaging modality for simultaneously mapping many metabolites contrast-free. Multi-dimensional MRSI further extends this capability by producing spatially resolved, multiple metabolite profiles at each voxel via introducing additional encoding dimensions, e.g., TE (also referred to as J-resolved) encodings. This affords improved molecular specificity and quantification of molecule-specific biophysical parameters, e.g., T2 relaxation times, beyond the traditional concentration measure. However, multi-dimensional MRSI is highly challenging due to (1) the inherent low SNR of MRSI, (2) the higher dimensionality of the imaging problem, and (3) the lack of accurate methods to quantify the parameters of interest.
We proposed a novel multi-TE MRSI approach in this work to achieve fast, high-resolution, multi-parametric molecular imaging of the brain. The proposed method integrated (1) an augmented molecule-component-specific subspace model for multi-TE 1H-MRSI signals, (2) an estimation-theoretic experiment design for optimized molecular quantification, (3) a physics-driven subspace learning strategy for spatiospectral reconstruction and molecular quantification, and (4) an accelerated multi-TE MRSI acquisition for generating high-resolution data in clinically relevant time.
Our proposed experiment optimization significantly improved the estimation of metabolites, neurotransmitters and their T2’s over conventional imaging parameter choices (e.g., reducing variances of neurotransmitter concentration estimates by ∼40% and metabolite T2 by ∼60%). Simultaneous brain metabolite and neurotransmitter mapping was achieved at a nominal resolution of 3.4×3.4×6.4 mm3 in less than 15 minutes. High-resolution, 3D metabolite T2 mapping was made possible for the first time. Our method revealed interesting metabolic abnormalities in a post-traumatic epilepsy patient, demonstrating translational potential.
With the new capabilities to concurrently visualize and quantify multiparametric metabolite and neurotransmitter changes at millimeter-level resolutions, we expect our technology to provide more molecular-level insights into brain disorders such as neurological and neurodegenerative diseases, and to potentially improve disease diagnosis and treatment.