Fsl_mrsi QuantificationError crashes entire run when Cr+PCr fits to exactly zero in a single voxel

MRS Software FSL-MRS
1

Hi all,

Apologies I’m fairly new to MRS.

I’m running fsl_mrsi for water-referenced quantification on 3T semi-LASER MRSI data (TE=40ms, 8x8 CSI, white matter), and a single problem voxel is crashing the whole run.

FSL-MRS 2.4.12, Osprey-converted basis set, mask thresholded on tissue fraction + fCSF, spline,stiff baseline.

Error: Exception (Metabolite reference [‘Cr’, ‘PCr’] has not been fit (conc=0). Please choose another or refine fit first.) occured in index (5, 4, 0).
fsl_mrs.utils.results.FitRes.QuantificationError: Metabolite reference [‘Cr’, ‘PCr’] has not been fit (conc=0). Please choose another or refine fit first.

I’ve tried tightening the mask (0.5 → 0.7 → 0.8 tissue fraction, plus fCSF <0.25). A different voxel seems to fail each time. I’ve also tried running without --h2o using relative concentrations only which seems to work fine so I’m guessing this is specific to the water scaling?

Is there a workaround for FSL to skip/flag voxel instead of crashing? Or is something like a pre-screening pass a standard workaround?

Would love to hear how others handle this for multi-participant MRSI pipelines.
Thanks!

2

Hi @andre

This has been a frustration for a while that we are finally fixing. You can get a development version (will be FSL-MRS V3 soon) from https://fsl.fmrib.ox.ac.uk/fsldownloads/fslconda/development/ and follow the instructions (option 2) modified slightly

conda install -c conda-forge \
              -c https://fsl.fmrib.ox.ac.uk/fsldownloads/fslconda/development/ -c https://fsl.fmrib.ox.ac.uk/fsldownloads/fslconda/public/ \
              fsl_mrs

Check you have the right after by using conda list and looking for version 2.4.17.2026072...

1 Like
3

Thanks so much @wclarke , I’ll give this a go.