Yi Zhang1,2, Refaat E. Gabr1,
Michael Schr1,3, Robert G. Weiss1,4, Paul A. Bottomley1,2
1Division of MR Research,
Johns Hopkins Univesity, Baltimore, MD, United States; 2ECE, Johns
Hopkins University, Baltimore, MD, United States; 3Philips
Healthcare, Cleveland, OH, United States; 4Division of Cardiology,
Johns Hopkins Univesity, Baltimore, MD, United States
Scan-time and signal-to-noise ratio (SNR) are central problems for MRS of low-concentration metabolites. Matching voxel sizes to anatomical compartments a priori, yields better SNR than adding signals from smaller chemical shift imaging (CSI) voxels post-acquisition to create an average spectrum from the same volume. While several solutions were proposed for CSI, the necessary a priori compartmentation and tailoring of gradient-encoding was never really implemented in humans. Here, we use linear algebraic modeling and a reduced phase-encoding set to achieve manyfold gains in SNR and/or scan-time reduction for 31P MRS in the human leg and heart vs. the conventional CSI experiment.