Feasibility of a Parallel Imaging Approach Exploiting Correlations in k-t Space for the Acceleration of ¹H Magnetic...

Magnetic resonance spectroscopic imaging (MRSI) provides spatially resolved metabolic information, but its clinical use has been limited by prohibitively long acquisition times. Parallel imaging methods such as GRAPPA have been applied to accelerate MRSI, though conventional implementations often reconstruct each time point independently and fail to exploit spectral correlations across free induction decays. In this work, we present a feasibility study of autocalibration region extending through time (ARTT) GRAPPA, a reconstruction approach that extends the autocalibration region through time (k-t space) to improve coil weight estimation by incorporating both spatial and spectral correlations. Data were acquired at 7 Tesla using a two-dimensional FID-MRSI sequence with a 24-channel array. A fully sampled dataset was retrospectively undersampled to simulate varying acceleration patterns, and reconstructions with ARTT GRAPPA were compared with conventional GRAPPA. A homogeneous phantom dataset was also evaluated to assess performance in the absence of spatial-spectral variability. Finally, prospectively undersampled acquisitions were performed in three healthy volunteers with effective acceleration factors of 2.50 and 3.58. ARTT GRAPPA enabled accurate metabolite reconstructions at acceleration factors exceeding 3.5, reducing scan times from over an hour to less than 20 min while maintaining normalized errors below 10% for key metabolites. Compared with the conventional method, ARTT GRAPPA consistently achieved more than twice the acceleration at equivalent accuracy. Phantom experiments confirmed that the advantage of ARTT GRAPPA arises from exploiting spatial-spectral correlations rather than intrinsic algorithmic changes. Prospectively undersampled in vivo data further demonstrated feasibility, producing reliable metabolite maps at substantially reduced acquisition times. These findings establish ARTT GRAPPA as a proof-of-principle approach that leverages k-t correlations for practical acceleration of MRSI, suggesting a path toward improved metabolic imaging in clinical research applications.