Abstract
BACKGROUND: Magnetic Resonance Spectroscopic Imaging (MRSI), also known as Chemical Shift Imaging (CSI), is a pivotal tool in both clinical and preclinical metabolic research. Traditional MRSI offers high sensitivity to weak metabolites and covers a wide spectral bandwidth. However, the large number of RF excitations required for fully sampled 3D-MRSI acquisitions renders it impractical for hyperpolarized (HP) MRI applications, especially given the rapid signal decay and non-renewable magnetization of HP agents such as [1-(13)C]pyruvate. PURPOSE: This study aims to develop and validate an accelerated MRSI method that can preserve broad spectral bandwidth and weak metabolite detectability without aliasing, overcoming limitations of fast MRSI techniques such as echo-planar spectroscopic imaging (EPSI), which typically cause narrower spectral bandwidth and can suffer from spectral aliasing. METHODS: We implemented a sparsely sampled 3D-MRSI pulse sequence on an MRI scanner, acquiring data with large reduction ratios. A 4D compressed sensing (CS) reconstruction algorithm was developed to recover high-resolution spectroscopic data from undersampled measurements. The algorithm jointly reconstructs the three spatial dimensions and the frequency dimension, leveraging sparsity priors and iterative conjugate gradient optimization. The in vivo experiments were performed on a GE 3 T clinical MRI scanner (GE MR750W) using hyperpolarized [1-(13)C]pyruvate in one rat, with two acquisitions (R = 8 and R = 16) performed sequentially. RESULTS: Our method achieved high-quality reconstructions even at acceleration factors of R = 16 and R = 32, corresponding to 6.25 and 3.125% sampling, respectively. The normalized root-mean-square error (nRMSE) and structural similarity index (SSIM) remained low (nRMSE < 4 × 10(-3), SSIM > 0.95) even at high undersampling rates. In vivo experiments using hyperpolarized [1-(13)C]pyruvate in rat kidneys demonstrated the ability to resolve lactate, alanine, pyruvate, and bicarbonate distributions with high spatial and spectral fidelity. CONCLUSION: The integration of sparse MRSI acquisition and 4D-CS reconstruction enables rapid, high-fidelity MRSI with HP (13)C-MRSI. This approach reduces acquisition time by up to 32-fold, facilitating dynamic metabolic studies and improving feasibility for routine preclinical and future clinical use.