Phantom Preprocessing in FSL-MRS: frequency and/or phase drift issues?

Hello FSL-MRS team and @wclarke,

Our group at UC Irvine has been working to quantify a series of spectroscopy phantoms to benchmark methods for in vivo hippocampal SVS in human adults. As a sanity check, we have been using PRESS sequences acquired from phantoms (reading out DICOM) and passed through spec2nii. While the resulting averaged FID peaks look plausible relative to simulations, fitting to several basis sets fails badly.

We noticed that this seems to occur as a function of frequency and/or phase drift wherein peaks are shifted .2-.3 ppm downfield relative to expected values. The phantoms are at room temperature during scanning (20–22°C) and are not necessarily at physiologic pH, so perhaps this explains some amount of deviation from in vivo expected values.

I have run fsl_mrs_preproc on the .nii.gz outputs of spec2nii dicom , but it outputs the following:

This data contains no unaveraged transients or uncombined coils. Please carefully ascertain what pre-processing has already taken place, before running appropriate individual steps using fsl_mrs_proc. Note, no pre-processing may be necessary.

Can you please advise on what further steps should be taken in FSL-MRS processing to appropriately prepare these phantom data for subsequent fitting?

Thank you!

Hi Tom,

If all peaks are shifted, then you could use fsl_mrs_proc fshift --file you_data.nii.gz --shiftppm 0.2 --output . --filename your_data_shifted. You might need -0.2 rather than 0.2, or 0.3 as you say. Let me know if that fixes the issue.

Re fsl_mrs_preproc, this assumes that there are multiple transients or coils in the data which require processing. It sounds like that isn’t the case here. What is the output of mrs_tools info on your data?