Evere fit distortion in FSL-MRS despite good preprocessed spectrum

MRS Software FSL-MRS
1

Hello FSL community,

I am encountering an issue with data fitting for the PRESS sequence acquired on 3T Siemens scanner. It generates a severely distorted fit for some of my datasets (not consisitent), even though the preprocessed input spectrum looks clean and well-phased. The data is recorded from vmPFC and is comparitvely noisy.

The fitting code used for the data was:
fsl_mrs
–data “$workdir/fsl_mrs_preproc_low/metab.nii.gz”
–basis “$BASIS_SET”
–metab_groups MM09 MM12 MM14 MM17 MM21
–TE 30
–output “$workdir/fsl_mrs_quant_lowmm_default”
–tissue_frac “$sesdir/segment/segmentation.json”
–h2o “$workdir/fsl_mrs_preproc_low/wref.nii.gz”
–report

done

Image
Image1280×720 94.9 KB

I tried different baseline setting also but that did not work (and baseline might not be the only issue). In fact when I tried using the spline baseline (mutiple stiffness) and it was unable to fit the data and gave the error “Metabolite reference [Cr and Pcr] has not been fit”. I have tried fitting the same data with Opsrey (both Osprey default fit and LCModel fit) and it works well. The basis set used for quantification in is the same in both the softwares.

I will really appreciate insights and help into this issue.

2

Hi @Harleen ,

I think this arises from some old data driven estimation of the phase/frequency handedness of the data . Spec2nii now handles this, and this “feature”, which goes wrong if there is a string residual water signal will be removed in the next full version. For now try adding --no_conj_fid to the call.

3

Thank you Clarke for the quick response. The data fits with the addition of --no_conj_fid to the call.
Looking forward to the next version.