Sdcfows find_estimators unable to find PEPolar epi maps

Software Support
,
1

Summary of what happened:

I am trying to diagnose an error in sdcflows that’s preventing fMRIprep from detecting pepolar estimators for my data. Going off of a previous thread, I verified that all metadata was correctly specified. (details in last section)

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

sdcflows was invoked via the python interpreter from an fmriprep 21.0.4 singularity container.

Version:

sdcflows: 2.0.11
fmriprep: 21.0.4

Environment (Docker, Singularity, custom installation):

fmriprep singularity image based off of official docker image.

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

BIDS only outputted warnings about inconsistent scan parameters (piloting, so parameters are changing).

Relevant log outputs (up to 20 lines):

>>> find_estimators(layout, subject="Pat01")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/conda/lib/python3.8/site-packages/sdcflows/utils/wrangler.py", line 271, in find_estimators
    _e = fm.FieldmapEstimation([
  File "<attrs generated init sdcflows.fieldmaps.FieldmapEstimation>", line 7, in __init__
  File "/opt/conda/lib/python3.8/site-packages/sdcflows/fieldmaps.py", line 398, in __attrs_post_init__
    raise ValueError("Insufficient sources to estimate a fieldmap.")

Screenshots / relevant information:

Pat01 Files

| sub-Pat01/
| | ses-1/
| | | anat/
| | | dwi/
| | | fmap/
| | | | sub-Pat01_ses-1_acq-dmri_dir-AP_epi.json
| | | | sub-Pat01_ses-1_acq-dmri_dir-AP_epi.nii.gz
| | | | sub-Pat01_ses-1_acq-fmri_dir-AP_epi.json
| | | | sub-Pat01_ses-1_acq-fmri_dir-AP_epi.nii.gz
| | | | sub-Pat01_ses-1_acq-fmri_dir-PA_epi.json
| | | | sub-Pat01_ses-1_acq-fmri_dir-PA_epi.nii.gz
| | | func/
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-1_bold.json
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-1_bold.nii.gz
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-1_sbref.json
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-1_sbref.nii.gz
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-2_bold.json
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-2_bold.nii.gz
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-2_sbref.json
| | | | sub-Pat01_ses-1_task-facename_dir-PA_run-2_sbref.nii.gz
| | | | sub-Pat01_ses-1_task-rest_dir-PA_bold.json
| | | | sub-Pat01_ses-1_task-rest_dir-PA_bold.nii.gz
| | | | sub-Pat01_ses-1_task-rest_dir-PA_sbref.json
| | | | sub-Pat01_ses-1_task-rest_dir-PA_sbref.nii.gz

Metadata for fieldmaps

$ grep 'B0Field' sub-Pat01/ses-1/func/*json
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-dmri_dir-AP_epi.json:    "B0FieldIdentifier": "pepolar_dmri"
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-fmri_dir-AP_epi.json:    "B0FieldIdentifier": "pepolar_facename"
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-fmri_dir-PA_epi.json:    "B0FieldIdentifier": "pepolar_facename"

$ grep 'PhaseEncodingDirection\"' sub-Pat01/ses-1/fmap/*json
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-dmri_dir-AP_epi.json:    "PhaseEncodingDirection": "j-",
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-fmri_dir-AP_epi.json:    "PhaseEncodingDirection": "j-",
sub-Pat01/ses-1/fmap/sub-Pat01_ses-1_acq-fmri_dir-PA_epi.json:    "PhaseEncodingDirection": "j",

IntendedFor is set for all fieldmaps and points to either a single diffusion or all bold + sbref runs.

Metadata for bold + sbref

$ grep 'B0Field' sub-Pat01/ses-1/func/*json
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-1_bold.json:    "B0FieldSource": "pepolar_facename"
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-1_sbref.json:    "B0FieldSource": "pepolar_facename"
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-2_bold.json:    "B0FieldSource": "pepolar_facename"
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-2_sbref.json:    "B0FieldSource": "pepolar_facename"
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-rest_dir-PA_bold.json:    "B0FieldSource": "pepolar_facename"
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-rest_dir-PA_sbref.json:    "B0FieldSource": "pepolar_facename"

$ grep 'PhaseEncodingDirection\"' sub-Pat01/ses-1/func/*json
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-1_bold.json:    "PhaseEncodingDirection": "j",
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-1_sbref.json:    "PhaseEncodingDirection": "j",
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-2_bold.json:    "PhaseEncodingDirection": "j",
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-facename_dir-PA_run-2_sbref.json:    "PhaseEncodingDirection": "j",
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-rest_dir-PA_bold.json:    "PhaseEncodingDirection": "j",
sub-Pat01/ses-1/func/sub-Pat01_ses-1_task-rest_dir-PA_sbref.json:    "PhaseEncodingDirection": "j",

N.B. IntendedFor points to bold + sbref; B0FieldSource is also set for both bold + sbref.

2

Nice report with a lot of details!

Everything looks mostly fine to me. One thing perhaps, since there is only one direction present for acq-dmri_dir-AP_epi in fmap/, perhaps the fieldmap estimator is complaining about that?
For testing could you remove (or filter out) sub-Pat01_ses-1_acq-dmri_dir-AP_epi.json and sub-Pat01_ses-1_acq-dmri_dir-AP_epi.nii.gz and try again to launch fmriprep (removing all temporary files from previous run before launching)?
It would be also helpful to see you fmriprep command.

Good luck!

3

Assuming it’s the presence of the DWI fieldmap that’s causing issues, this is a bug. We should be dropping incomplete fieldmaps (possibly with a warning), rather than raising an error.

4

I think I might have found the solution:

The “incomplete” fieldmap is meant to be paired with a dwi scan. I included B0FieldSource but I did not include B0FieldIdentifier in sub-Pat01_ses-1_dir-PA_dwi.json. Including B0FieldIdentifier seemed to do the trick.

Strange though. I included the dwi scan in the IntendedFor for sub-Pat01_ses-1_acq-dmri_dir-AP_epi.json. If both IntendedFor and B0Field* tags are included, does sdcflows use the other as fallback?

5

If B0FieldIdentifier is present in a fieldmap, then IntendedFor is ignored and no associations are attempted. If both are present and no BOLD file specifies B0FieldSource, then a fieldmap could be estimated, but not applied.

6

Still sounds like a bug. A warning sounds reasonable and might help you catch the issue. A crash means that a misconfigured DWI scan could break fMRI tools.

7

Interestingly enough, when I ran sdcflows on the corrected set, find_estimators was able to correctly guess the fieldmaps.

However, when I ran fmriprep 21.0.4, I was still getting ValueError: "Insufficient sources to estimate a fieldmap."

When I removed all B0Field* tags, fmriprep 21.0.4 was finally able to find the estimators. Might have been inconsistent tagging on my part.

8

21.0.4 will be using an older version of sdcflows. Bug fixes march ever on.

1 Like
9

Updated to 22.1.1, unfortunately the extra dMRI images are still throwing off find_estimators(). Had to split off a b0 and save as a separate epi image.

For personal reference, is this scheme not supported by sdcflows? That is, one diffusion sequence paired with one reverse-PE as a separate image?

10

I don’t think there’s any BOLD-specific logic that would keep SDCFlows from finding a diffusion+reverse fieldmap pair. However, we exclude dwi from indexing for efficiency in fmriprep. I suspect this is causing sdcflows to fail to find the matched fieldmaps in dwi.

11

@ajschadler If you are able to patch in an experimental copy of sdcflows, I have a PR that might fix this: FIX: Log incomplete fieldmaps, rather than erroring by effigies · Pull Request #341 · nipreps/sdcflows · GitHub

Using fmriprep-docker:

git clone https://github.com/effigies/sdcflows.git
git -C sdcflows checkout fix/log_incomplete
fmriprep-docker -i nipreps/fmriprep:unstable --patch sdcflows=sdcflows/sdcflows \
    bids_dir output_dir participant [...]
12

Looks like your fix worked! Ran swimmingly.

1 Like