fmriprep: Incorrect/dark functional images after SDC with SyN

What happened?

I am testing version 21.0.2 on an fMRI dataset for which fieldmaps were not acquired. As such, I am using the SyN method to attempt susceptibility distortion correction. However, I am getting strange results, inconsistent across subjects. I provide two examples.

[sub-01] The SDC estimation includes a considerable translation to match the functional and the anatomical images, which results in this estimated fieldmap

sub-01_fmapid-auto00000_desc-anat_fieldmap

and corrected image:

sub-01_task-control_run-1_desc-sdc_bold

I see no problems in the func to T1w or T1w to MNI registrations.

[sub-04] The estimated fieldmap seems more reasonable

sub-04_fmapid-auto00000_desc-anat_fieldmap

but the corrected image is mostly dark

sub-04_task-control_run-1_desc-sdc_bold

I have tested fmriPrep on the same dataset without the SyN correction and all seems fine.

What command did you use?

docker run -ti --rm \
    -v /SCRATCH/users/alexandresayal/BIDS-VP-HYSTERESIS:/data:ro \
    -v /SCRATCH/users/alexandresayal/BIDS-VP-HYSTERESIS/derivatives:/out \
    -v /SCRATCH/users/alexandresayal/fmriprep-workdir-hyst:/work \
    -v /SCRATCH/software/freesurfer/license.txt:/license \
    nipreps/fmriprep:21.0.2 \
    /data /out/fmriprep \
    participant \
    -w /work \
    --fs-license-file /license \
    --use-syn-sdc \
    --nprocs 10 \
    --mem-mb 35000 \
    --stop-on-first-crash \
    --fs-no-reconall \
    --output-spaces MNI152NLin2009cAsym \
    --participant-label 04

What version of fMRIPrep are you running?

21.0.2

How are you running fMRIPrep?

Docker

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

Please copy and paste any relevant log output.

No errors to report in either case.

Additional information / screenshots

I share the .svg images also in this Gdrive folder https://drive.google.com/drive/folders/1RcC4568D6_04gqNGFGp-j4Ec6o8_vdPp?usp=sharing

About this issue

Original URL
State: closed
Created 2 years ago
Comments: 16 (9 by maintainers)

Most upvoted comments

Please retry with 22.0.2. Skipping SyN-SDC should have been fixed.

effigies on Nov 18, 2022

Good to hear - the --use-syn-sdc flag is still fairly experimental, so I can’t say I’m surprised about these discrepancies. Do you know if there is something different (scanning parameters, missing metadata, etc) about the failing participants?

If you have any data that you can share with us (T1w + BOLD) of one of the failing runs, it will help us tremendously to improve the robustness of the SyN implementation.

mgxd on Aug 1, 2022

Are there any subjects in the dataset where the SyN correction worked as expected?

mgxd on Jul 27, 2022