Error running QSIPrep, V2.3.1

[harpine]

Dear Ido,

Thanks a lot for patching the version solving the duplicate error for dicom to nifti conversion. I had to delete & reinstall the docker images, but then it worked !

I am now facing a new error during the preprocessing;

From what I have seen, the tissue analyzer step works perfectly
( /ti-toolbox/tit/pre/tissue_analyzer.py:411: UserWarning: This figure includes Axes that are not compatible with tight_layout, so results might be incorrect.
plt.tight_layout()
Skin: Volume=109.55cm³, Thickness=1.45±0.59mm
Tissue analysis: ✓ Complete (79.9s)
QSIPrep DWI preprocessing: Started )

Then, I have an error running QSIPREP:

File "/ti-toolbox/tit/pre/qsi/qsiprep.py", line 99, in run_qsiprep
raise PreprocessError(f"QSI Docker preflight failed: {preflight_error}")
tit.pre.utils.PreprocessError: QSI Docker preflight failed: Host project directory does not exist: C:/Users/user/Documents/TI-toolbox2
Process returned non-zero exit code (1) while running: simnibs_python -m tit.pre /tmp/pre_5tetrvaz.json. Check the console output above for the backend error. If this was caused by missing inputs, verify the selected project, m2m folder, simulation TI/mTI outputs, atlas/ROI files, and Docker/X11 setup. Last output: File "/ti-toolbox/tit/pre/qsi/qsiprep.py", line 99, in run_qsiprep | raise PreprocessError(f"QSI Docker preflight failed: {preflight_error}") | tit.pre.utils.PreprocessError: QSI Docker preflight failed:
Pre-processing failed or was stopped. Check the console above and logs under derivatives/ti-toolbox/logs/sub-{subject}/. Common fixes: verify sourcedata DICOM layout, FreeSurfer license, CHARM/SimNIBS inputs, and Docker/QSI access.

However, the project folder does exist.

I first thought this would be because the DTI images are not automatically converted to nifti files as for T1w and T2w (please correct me if the TI-toolbox does it and I missed it ?), but after converting the files manually and placing them in the subject order, it still doesn't work.

Do you have any clue about what could cause this problem ?

Just for the record, I work on a computer with windows 10, TI-toolbox version is 2.3.1. I ran convert dicom to nifti, run freesurfer recon-all, create simnibs 2m2 folder and tissue analyzer successfully (at least I think) before running into the issue.

Don't hesitate to tell me if I should provide more information.
Best,
Aline

[idossha]

Hi @harpine ,

Thanks for the detailed report.

The preflight that failed was checking whether your host project path (C:/Users/user/Documents/TI-toolbox2) was visible from inside the container. On Windows that's never the case because of some internal Docker logic, so the preflight aborted even though your project was perfectly fine. I've pushed a fix and QSIPrep should get past the preflight now.

On your other question: it should be converting the DWI DICOMs into sub-{id}/dwi/ (with the .nii.gz .json sidecars next to it, BIDS-style).

Please continue to report if you encounter any problems, it makes the software better!

Best,
Ido

[harpine]

Dear Ido,

Thanks a lot, the error with the project folder not found has been solved!

About the DWI dicom, they have not been converted to nifti.

This is the path to my DWI folder:
C:\Users\user\Documents\TI-toolbox2\sourcedata\sub-wpo001\dwi\dicom

In sourcedata \ sub-wpo001, I have T1w, T1w, dwi (tried with DWI as well, doesn't work either).
Nothing shows up in sub-wpo001/dwi... Am I doing something wrong ?

Thanks a lot in advance for your help,
Best,
Aline

[harpine]

Dear Ido,

I re-ran the preprocessing and I have a similar error (output already exist), as before with another step;

Subject atlas segmentation: ✓ Complete (29.1s)
Traceback (most recent call last):
File "", line 198, in _run_module_as_main
File "", line 88, in _run_code
File "/ti-toolbox/tit/pre/main.py", line 57, in
main()
File "/ti-toolbox/tit/pre/main.py", line 35, in main
exit_code = run_pipeline(
^^^^^^^^^^^^^
File "/ti-toolbox/tit/pre/structural.py", line 501, in run_pipeline
return _run_pipeline_inner(
^^^^^^^^^^^^^^^^^^^^
File "/ti-toolbox/tit/pre/structural.py", line 701, in _run_pipeline_inner
d = _run_subject_pipeline(
^^^^^^^^^^^^^^^^^^^^^^
File "/ti-toolbox/tit/pre/structural.py", line 238, in _run_subject_pipeline
if _should_run_output_step(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/ti-toolbox/tit/pre/structural.py", line 98, in _should_run_output_step
raise PreprocessError(
tit.pre.utils.PreprocessError: FreeSurfer recon-all output already exists for subject wpo001: /mnt/TI-toolbox3/derivatives/freesurfer/sub-wpo001. Choose skip existing outputs, replace existing outputs, or disable this step.
Process returned non-zero exit code (1) while running: simnibs_python -m tit.pre /tmp/pre_7io4xy8e.json. Check the console output above for the backend error. If this was caused by missing inputs, verify the selected project, m2m folder, simulation TI/mTI outputs, atlas/ROI files, and Docker/X11 setup. Last output: File "/ti-toolbox/tit/pre/structural.py", line 98, in _should_run_output_step | raise PreprocessError( | tit.pre.utils.PreprocessError: FreeSurfer recon-all output already exists for subject wpo001: /mnt/TI-toolbox3/derivatives/freesurfer/sub-wpo001. Choose skip existing outputs, replace existing outputs, or disable this step.
Pre-processing failed or was stopped. Check the console above and logs under derivatives/ti-toolbox/logs/sub-{subject}/. Common fixes: verify sourcedata DICOM layout, FreeSurfer license, CHARM/SimNIBS inputs, and Docker/QSI access.

I re-try with an empty folder and it happened again... (I ticked all boxes in the preprocessing) Any idea how I could solve this ?

Thank you a lot for your help in advance !
Best,

[idossha]

Hi Aline,

Thanks again for the careful reports — both of these turned out to be real bugs on our side, and I've now fixed both.

1. "recon-all output already exists" on a fresh project. This was the pipeline tripping over its own scaffolding: it was pre-creating an empty derivatives/freesurfer/sub-{id} folder at startup, and then the "already exists" check saw that empty folder and aborted — even on a clean project. The pipeline no longer creates that empty folder, and the check now ignores empty directories, so a fresh run will go through.

2. DWI DICOMs not being converted. You were right — this was never actually implemented; the converter only handled T1w/T2w. It now converts the DWI series too, reading from sourcedata/sub-{id}/dwi/dicom/ and writing sub-{id}/dwi/sub-{id}_dwi.nii.gz together with the .bval/.bvec files QSIPrep needs (folder name matched case-insensitively, so either dwi or DWI works).

I apologize for the oversight. I guess this is the price to pay for rapid solo development :) Thanks again for helping make the toolbox better — please keep reporting!

p.s. Happy to schedule another Zoom meet if anything is still unclear or buggy.

Best,
Ido