$FSLDIR not found during segmentation task

[bellarude]

Dear #wtclarke

I am experiencing this fsl.utils.run.FSLNotPresent: $FSLDIR is not set - FSL cannot be found! error while testing a segmentation task on the provided example files: T1.nii used as reference for metab.nii . I attach here the print out of the error.

I tried to look it up the original FSL installation website and other similar reported issues for heads up but yet I struggle with the path definition within the fsl_mrs implementation.

The installation of my fsl_mrs followed the conda option and other commands (or at least the few I tried so far) do run.
Would you have any suggestion?

Many thanks

[wtclarke]

Hi @bellarude, did you follow the Windows specific instructions for setting up FSL-MRS with FSL? If you did, is WSL running?

[bellarude]

Dear @wtclarke

thanks for the prompt note. Yes I did install the full FSL package on WSL. Btw the FSL-MRS that comes with that is in a older version 1.1.8. Now I figured I can either run this FSL-MRS on Ubuntu WSL ( which I will call case 1.) or the other one I have installed via conda in my window machine (which is a FSL-MRS version 2.0.2and I will call case 2.).

case 1.: I can run segmentation with FSL. However, if I try to run the segmentation with FSL-MRS it gives me errors with directories, such as the one attached in the print out. If I try to set up an output directory with the -o command it still tells me it cannot create the directory even if I run the WSL as admin.

case 2.: despite having added the environment variable correctly upon having FSL correctly installed in \\wsl$\usr\local\fsl I still run into the error mentioned in the former post. I tried to run the fsl_mrs_verify and it fails on the segmentation task as well. FSL is running on WSL while try to run FSL-MRS on windows, as you indicated in the installation guidelines.

I think I am stepping into either a newbie error with this ubuntu setup or I am missing some perhaps naïve setup on my windows machine. Would you have any heads up on what to test out for debugging?

Many thanks!

[wtclarke]

I think in case 1 the error is referring to the fact that fsl_anat has already been (partially?) run. As such the command should be run with the -a /path/to/fsl_anat.anat option rather than -t.

For case 2 it's not immediately clear what is happening. Could you run set fsl in the same environment and post the output?

[bellarude]

Dear @wtclarke thank you again for your impressive promptness!

case 1. good heads up indeed. I provided .anat reference and it worked. case 2. I attach here the output from set fsl

[wtclarke]

Hi @bellarude, Please forgive the delay in replying, holiday and fellowship interviews cut across August. Did you manage to get the setup to work with the second case? If not I can suggest trying to do a quick debugging session virtually. I'm keen to get this all working on Windows but I don't have much access to lots of different setups.

BW

Will

[bellarude]

Dear @wtclarke, quick update on my side. I successfully managed to replicate the mrsi segmentation on the example data that are provided. I had to exploit the _fslmrs in my ubuntu environment for that but it works just fine.

On the other hand as soon as I try to extend it to my data I run into troubles. While I run the mrsi_segmentwith the .anat reference, the segmentation stops yielding the following error:

I must admit that I am not familiar with fsl at all. Therefore, I have been looking into this mismatch between sform and qform and seems to be related with fsl itself rather than _fslmrs. However, I could not find any heads up on how to prevent this mismatch to happen.

The .anat reference is generated from a T1w image that is segmented using the plain fsl_anat -i <input.nii.gz>. The .nifti version of the mrsi scan is produced via fls-mrs command spec2nii dicom.

Any suggestion? Many thanks and BW. Rudy

[wtclarke]

I think I've seen this before, will get back to you on it. Off the top of my head try fslorient -copysform2qform T1_fast_pve_0.nii.gz on that file in the first line of the error. Also on pve_1 and pve_2 as well

[bellarude]

Thanks Will. However, so doing it seems that the mismatch of qform and sform error is postponed at the tmp.nii.gz that (I guess) mrsi_segment is generating. I am willing to share an example of my dataset where I wish to run the software, to help solving the issue.

Thanks a lot for the support. Rudy

[wtclarke]

Sorry I don't understand. What error do you now get?

Yes, please to the dataset.

[wtclarke]

P.s. what version of FSL-MRS are you using? fsl_mrs --version

[bellarude]

Hi Will, impressed by the promptness of your answer. Thank you.

• fsl version is 1.1.8 (is the one that comes with fsl installation on my ubuntu system)
• the error I get is the same as yesterday: sform and qform not identical in image .... tmp.nii.gz and tmp.nii.gzis the nifti file that is getting generated on the destination folder where I wish to have the segmentation output.

I will send you an email with an example of the dataset. Perhaps knowing that everything would work for you could help me debugging it further.

BW Rudy

[wtclarke]

Hi @bellarude ,

Could you try updating to the newest version of FSL-MRS. Unfortunately the one which comes packages with FSL is trailing behind at the moment (the problem of being a newer tool in the library).

To get the newest version can you follow the following instructions? https://open.win.ox.ac.uk/pages/fsl/fsl_mrs/install.html#option-1-using-conda