Fieldmap correction gone wrong

[e0046902]

Summary of what happened:

I am running fmriprep on my data with fieldmap (2 magnitude images + 1 phase diff image). The output image looks weird, and very distorted. Any idea on possible reason?

Command used (and if a helper script was used, a link to the helper script or the command generated):

export SINGULARITYENV_TEMPLATEFLOW_HOME=/home/fmriprep/.cache/templateflow
/usr/bin/singularity run \
    --contain \
    --cleanenv \
    -B /tmp -B /data -B /data1 -B /data2 -B /data3 -B /n -B /cm/shared -B /data/fmriprep-workdir \
    /cm/shared/singularity/images/fmriprep-23.0-latest.simg \
    /data/achowdhury/Regret/derivatives \
    /data/achowdhury/Regret/fmriprepoutput \
    participant \
    --fs-license-file /cm/shared/freesurfer-license.txt \
    --participant_label 02 \
    --output-spaces MNI152NLin2009cAsym:res-2 anat func fsaverage \
    --n_cpus 4 \
    --mem-mb 8192 \
    --notrack \
    --dummy-scans 8 \
    --use-aroma \
    --skip-bids-validation \
    --fs-no-reconall \
    -vv \
    -w  /data/fmriprep-workdir/achowdhury

Version:

fmriprep version number. Default: 23.0-latest

Environment (Docker, Singularity, custom installation):

Singularity on a computing cluster

Data formatted according to a validatable standard? Please provide the output of the validator:

ExperimentalWarning: The fs.promises API is experimental
	1: [WARN] Task scans should have a corresponding events.tsv file. If this is a resting state scan you can ignore this warning or rename the task to include the word "rest". (code: 25 - EVENTS_TSV_MISSING)
		./sub-01/func/sub-01_task-regret_run-01_bold.nii.gz
		./sub-02/func/sub-02_task-regret_run-01_bold.nii.gz
		./sub-03/func/sub-03_task-regret_run-01_bold.nii.gz
		./sub-04/func/sub-04_task-regret_run-01_bold.nii.gz
		./sub-05/func/sub-05_task-regret_run-01_bold.nii.gz
		./sub-06/func/sub-06_task-regret_run-01_bold.nii.gz
		./sub-07/func/sub-07_task-regret_run-01_bold.nii.gz
		./sub-08/func/sub-08_task-regret_run-01_bold.nii.gz
		./sub-09/func/sub-09_task-regret_run-01_bold.nii.gz
		./sub-10/func/sub-10_task-regret_run-01_bold.nii.gz
		... and 28 more files having this issue (Use --verbose to see them all).

	Please visit https://neurostars.org/search?q=EVENTS_TSV_MISSING for existing conversations about this issue.

	2: [WARN] Not all subjects/sessions/runs have the same scanning parameters. (code: 39 - INCONSISTENT_PARAMETERS)
		./sub-01/func/sub-01_task-regret_run-01_bold.nii.gz
		./sub-02/func/sub-02_task-regret_run-01_bold.nii.gz
		./sub-03/func/sub-03_task-regret_run-01_bold.nii.gz
		./sub-04/func/sub-04_task-regret_run-01_bold.nii.gz
		./sub-06/func/sub-06_task-regret_run-01_bold.nii.gz
		./sub-07/func/sub-07_task-regret_run-01_bold.nii.gz
		./sub-08/func/sub-08_task-regret_run-01_bold.nii.gz
		./sub-09/func/sub-09_task-regret_run-01_bold.nii.gz
		./sub-12/func/sub-12_task-regret_run-01_bold.nii.gz
		./sub-13/func/sub-13_task-regret_run-01_bold.nii.gz
		... and 23 more files having this issue (Use --verbose to see them all).

	Please visit https://neurostars.org/search?q=INCONSISTENT_PARAMETERS for existing conversations about this issue.

	3: [WARN] The recommended file /README is missing. See Section 03 (Modality agnostic files) of the BIDS specification. (code: 101 - README_FILE_MISSING)

	Please visit https://neurostars.org/search?q=README_FILE_MISSING for existing conversations about this issue.

	4: [WARN] The Authors field of dataset_description.json should contain an array of fields - with one author per field. This was triggered based on the presence of only one author field. Please ignore if all contributors are already properly listed. (code: 102 - TOO_FEW_AUTHORS)

Relevant log outputs (up to 20 lines):

Screenshots / relevant information:

I do not have a pic of the erroneous outputs at this moment as I gave deleted them but I can try to upload them later. In any case,
I am attaching the BOLD file and fieldmap:

sub-02

Thanks in advance.

[Steven]

Hi @e0046902,

I have relabeled your post as Software Support and added the corresponding post template, please fill in the other requested information by editing your so we an best help. Also, in the Screenshots section, please post the SDC images from the HTML report.

Best,
Steven

[e0046902]

hi @Steven @Chris_Rorden, thank you. I have populated the related fields in the form, except the picture of the output from preprocessing. I deleted that previously but should be able to upload it later. Otherwise, I have provided all required information.

Thank you very much for your help!

[jsein]

Maybe this thread can enlighten you?

Making Sure that the SDC has well-Applied

@jsein Thank you so much again for your time and the detailed guidance. I have applied SDC using fMRIPrep LTS v20.2.7 as suggested, the output looks pretty reasonable. [SDC_LTSv20.2.7] I checked the fieldmap less SDC by including --use-syn-sdc flag with both v23.0.1 and v20.2.7_LTS. It did work fine to some extent for v.20.2.7 but not good for v23.0.1. The outputs are as follows: synsdc with LTS v20.2.7 [synsdc_LTS20.2.7] synsdc with v23.0.1 [synsdc_v23.0.1] I started skullstripping t…

[e0046902]

Hi @jsein @Steven,

I tried skull-stripping only the phasediff image before feeding it into fmriprep, and the output is still very distorted:

[jsein]

Indeed, I looked at your dataset and tried fmriprep v23 vs fmriprep v20.2.6 on your raw data and the difference is dramatic:

Référence image, not corrected for SDC:

Référence SDC-corrected: fmriprep v20.2.6:

Référence SDC-corrected: fmriprep v23:

Note 1: The difference of contrast in image for correction version of v20.2.6 vs v23 is that in fmriprep v20.2.6, I used the SBref image as reference, whereas for fmriprep v23, I ignored the SBref.

Note 2: In fact in fmriprep v23, I did not proved a masked 'phasediffimage but the phasediff`image was well masked when I looked at it in the working directory. The issue seems to be coming in the creation of the field coefficient, with B-Spline functions. Here is the look at the fieldmap applied for fmriprep v23:

[e0046902]

@jsein Thanks.

So would it be right to say fmriprep v23 is not suitable for use with magnitude+phasediff fieldmap correction?

[wwgraves]

I’m having what looks like the same issue. Used fMRIPrep v23 and got “corrected” BOLD images that look quite distorted in posterior areas around the occipital lobe:

sub-NTR0019_task-words_run-01_desc-sdc_bold.svg | Powered by Box

Any new info on this issue?

[jsein]

Please try fmriprep v23.0.2 where a fix may have solved this issue:

Image orientation and PE direction for fMRIPrep

I had assumed they were reversed because the phase encoding direction given in the json file for the fmap is opposite (i.e. i vs i-), For phase-difference fieldmaps, that does seem likely to produce a sign-inverted estimate of the field, which would then lead to incorrectly expanding/contracting when applying correction. We’re using fMRIPrep 20.2.7 as advised here after we were having issues with SDC in v22 If you wouldn’t mind trying out 23.0.2 for the fieldmaps that are working well, I …

[wwgraves]

Thanks @jsein, and apologies for not being more clear. I did use the latest version:
About
fMRIPrep version: 23.0.2
fMRIPrep command: /opt/conda/bin/fmriprep /home/cmccabe/data/NTR/data/NTR0019_bids /home/cmccabe/data/NTR/data/work participant --participant-label NTR0019 --fs-license-file /home/wgraves/software/freesurfer/license.txt
Date preprocessed: 2023-05-04 09:30:50 -0400

Happy to try other stuff though!

[jsein]

Thank you @wwgraves for sharing this example of SDC going wrong with fmriprep v 23.0.2! Do you confirm that the SDC correction method used in your case by fmriprep v23.0.2 is the “phase difference” case?

May you try with fmriprep v20.2.7 and confirm that the SDC correction is still much better with this earlier version?

[wwgraves]

Yes, as far as I know it’s a “phase difference” case. I used the field map sequence that came with our Siemens Prisma. Using dcm2bids output two magnitude images, and one phase image, where the phase image had two echo times. That’s why I assume it’s a phase difference map. Please let me know if there’s a better way to verify.
I’m running fMRIPrep v20.2.7 now. I’ll post here once I’ve gotten a chance to take a look at the results!

[jsein]

Indeed, what you report sounds exactly like the the phase difference method for SDC.

[wwgraves]

Yes @jsein, it turns out changing only the fMRIPrep version to 20.2.7 led to WAY better distortion correction:

sub-NTR0019_task-words_run-1_desc-sdc_bold.svg | Powered by Box

In light of this, I’ll be interested to hear what developers think about next steps.