Evaluating real-world performance of retrospective frequency correction in large samples for single-voxel MR spectroscopy at 7T

Motivation: Retrospective frequency correction methods have been shown to improve spectral quality. However, only limited studies have investigated their impacts on metabolite quantification and their real-world performances in large samples on 3T systems. Goal(s): To evaluate four different frequency correction methods in large samples for single-voxel MRS at 7T. Approach: Creatine fitting (Cre), residual water (RW), spectral registration (SR), and cross-correlation (Xcorr) methods were applied to correct frequency/phase changes. Results: With frequency corrections, rCRLB showed significant reductions (p<0.05) in Ins (Cre, RW, Xcorr), NAAG (Cre, SR, Xcorr), Scyllo (SR), Tau (Cre, RW, SR, Xcorr), and -CrCH2(Cre, SR, Xcorr). Impact: This study demonstrates subjection motion-induced frequency changes affect certain metabolites with small concentrations rather than affecting major metabolites as in the presence of gradient heating-induced frequency drift. The resulting increased rCRLB of metabolites can be mitigated using retrospective frequency corrections.