Purpose: Averaging multiple repetitions to improve signal-to-noise ratio is common practice in magnetic resonance spectroscopy (MRS). However, temporal variations in scanner B0 due to motion or gradient heating may cause spectra to become misaligned, broadening and distorting peaks and impacting on processing and quantification. We present a comparison using in vivo data of different methods for correcting these errors.
Methods: Three different correction methods were applied to 53 brain scans: residual water peak alignment, creatine fitting, and spectral registration. In 32 of 53 subjects, diffusion tensor imaging (DTI) was acquired prior to the MRS scan. We compared the resulting linewidths to find the most effective technique. In addition, the impact on metabolite concentration estimates was evaluated.
Results: MRS data acquired after DTI imaging exhibited a frequency drift four times higher than data without DTI, resulting in changes to metabolite concentrations, particularly glutamate/glutamine. All three correction methods produced significantly improved linewidths relative to uncorrected data, with spectral registration performing best by a small margin.
Conclusion: Frequency correction is an important step in processing MRS data, significantly impacting metabolite quantification, particularly after echo-planar imaging that often occurs with MRS scans in clinical studies. Spectral registration proved most effective at frequency correction.
Keywords: 1H; MRS; frequency correction; signal processing; spectral registration; spectroscopy.
Copyright © 2016 by the American Society of Neuroimaging.