Reduction of voxel bleeding in highly accelerated parallel 1H MRSI by direct control of the spatial response function

Purpose To substantially improve spatial localization in magnetic resonance spectroscopic imaging (MRSI) accelerated by parallel imaging. This is important in order to make MRSI more reliable as a tool for clinical applications. Methods The sensitivity encoding acceleration technique with spatial overdiscretization is applied for the reconstruction of parallel MRSI. In addition, the spatial response function is optimized by minimizing its deviation from a previously chosen target function. This modified minimum-norm sensitivity encoding-MRSI reconstruction approach is applied in this article for in vivo pulse-acquire MRSI of human brain at 7T with simulated acceleration factors of 2, 4, and 9 as well as actual 4-fold accelerated MRSI. Results The sidelobes of the spatial response function are significantly suppressed, which reduces far-reaching voxel bleeding. At the same time, the major enlargement of the effective voxel size, which would be introduced by conventional k-space apodization methods, is largely avoided. Regularization allows for a practical trade-off between noise minimization, effective voxel size, and unaliasing. Although not aiming at increasing the nominal spatial resolution, a better spatial specificity is achieved. Conclusion Simultaneous suppression of short- and far-reaching voxel bleeding in MRSI is analyzed and reconstruction of highly accelerated parallel in vivo MRSI is demonstrated.