Stripe Artefacts after Preprocessing

[xuyangwen]

What happened?

The preprocessed images show significant stripe artefacts in all our participants in both normalized and native spaces, including those with minimal head motion. These artifacts exist on version 24.1.1, but seem to disappear on 22.0.1 (at least not in an obvious way).

What command did you use?

/opt/conda/envs/fmriprep/bin/fmriprep /data/p_02817/Social/Neuroimaging/RawData /data/p_02817/Social/Neuroimaging/Preprocessing participant --participant-label sub-01 --output-spaces MNI152NLin2009cAsym:res-2 T1w:res-native fsaverage:den-10k:den-41k --fs-license-file /software/freesurfer/licensekeys --cifti-output 91k --work-dir /data/p_02817/Social/Neuroimaging/Preprocessing/Work --write-graph --stop-on-first-crash

What version of fMRIPrep are you running?

24.1.1

How are you running fMRIPrep?

Other

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

No

Please copy and paste any relevant log output.

Additional information / screenshots

[effigies]

Do these appear in 25.1.3?

[xuyangwen]

I haven't tried version 25.1.3 yet. The server of our institute has just been updated to 24.1.1. Do you think upgrading to a newer version might resolve this issue? If so, I'll request the IT department to consider updating to the latest version.

[effigies]

I'm not certain that this issue is resolved in the latest version, but we can't work on this until it's confirmed to still be an issue.

If you have access to a Linux machine with Docker or Apptainer, you should be able to test this. You do not need an exceptionally powerful computer to run fmriprep.

[xuyangwen]

Hi Chris, I just tried using version 25.1.3 today, and I have the same issue, if not worse. Do you know what might be causing this? Let me know if I can provide you with more information to solve it.

[effigies]

No, I haven't seen anything like that. What do your reports look like?

[xuyangwen]

As far as I can see, everything looks normal on the visual report, and there is no error reported. We also double-checked with the technician at the institute, who confirmed that the raw data also looks fine. Would it be a problem related to the slab orientation? It is not typically AC-PC, but aligned with the axis of the hippocampus. But why is it not a problem on the older version of the fmriprep 22.0.1?

[effigies]

Are you able to share a subject? I can try running it and tracking down where it happens.

[xuyangwen]

Sure. Thank you! I can share with you on Google Drive ([email protected])?

[effigies]

That will work.

[xuyangwen]

I already sent you the data. Let me know if you have not received it. Thank you!

[effigies]

It looks like an issue with Jacobian intensity correction when applying fieldmaps, which appears to be doing more harm than good in your data. If you add --ignore fmap-jacobian to your command, you get more reasonable looking intensities within the brain, at the cost of visible smearing outside the brain.

[xuyangwen]

Thanks for your response. I quickly run the preprocessing using 25.1.3 with the --ignore fmap-jacobian command. Although the stripe artefacts are gone, the texture in the whole brain appears twisted and has the ring artefacts. Do you know the reason?

[effigies]

What image are you looking at?

[xuyangwen]

The functional images (...space-MNI152NLin6Asym_res-2_desc-preproc_bold.nii.gz). They all have the same problems across all the tasks and runs.

[effigies]

I'm not seeing what you're describing in sub-03_task-mvpa_run-1_space-MNI152NLin2009cAsym_res-2_desc-preproc_bold.nii.gz, which is the only one I processed. I've looked through the first 20 volumes.

[xuyangwen]

Sorry, the subject I preprocessed using version 25.1.3 with the --ignore fmap-jacobian command was actually sub-10, not sub-03. I am currently running sub-03. Could you please email me the image file sub-03_task-mvpa_run-1_space-MNI152NLin2009cAsym_res-2_desc-preproc_bold.nii.gz at [email protected]? This will allow me to make a comparison once the processing is complete. Thank you!

[xuyangwen]

I preprocessed my images of sub-03. It still has the same problem (twisted texture). Although it is better than sub-10, it is certainly different from yours. I sent you one of the images via email. Could you please take a look at the images attached? Do you know what the possible reason is?

[xuyangwen]

I appreciate your help in looking into this. I have two further questions:

1. Regardless of whether using the old version 22.0.1 or the newer version with --ignore fmap-jacobian, there are smear artefacts, especially in the medial temporal lobe. Is there a way to solve this for the data collected?
2. Since my research focuses on the medial temporal lobe, I might consider changing the scanning parameters to improve the signal. Do you know the reasons for these artifacts, or which parameters I should modify to ensure that fmriprep does not produce images with artifacts?

[xuyangwen]

I had the opportunity to communicate with the technician at our institute yesterday. Is it possible that the stripe artifacts using the newer version of fmriprep are because my phase direction is not exactly along AC-PC but along the long axis of the hippocampus?

[effigies]

Sorry for letting this one fall through the cracks. If you're realigning your dataset to ACPC before passing to fMRIPrep, that could be a source of the behavior you see.

If you're not realigning, then it should not be a problem.

[xuyangwen]

Thank you for your response. I did not realign the data to AC-PC.