Fmriprep not finding SBRef


I am using bidscoiner to transform my data to BIDS format, which works pretty well. it finds and renames all of my scans so that my directory (took as example ses02 because it-s most complete) looks like this

│ ├── ses-02
│ │ ├── anat
│ │ │ ├── sub-003_ses-mri02_acq-MPRAGE_run-1_T1w.json
│ │ │ ├── sub-003_ses-mri02_acq-MPRAGE_run-1_T1w.nii.gz
│ │ │ ├── sub-003_ses-mri02_acq-SPC_run-1_T2w.json
│ │ │ └── sub-003_ses-mri02_acq-SPC_run-1_T2w.nii.gz
│ │ ├── fmap
│ │ │ ├── sub-003_ses-mri02_acq-MB_dir-COL_run-1_epi.json
│ │ │ └── sub-003_ses-mri02_acq-MB_dir-COL_run-1_epi.gz
│ │ ├── func
│ │ │ ├── sub-003_ses-mri02_task-1back_run-1_echo-1_bold.json
│ │ │ ├── sub-003_ses-mri02_task-1back_run-1_echo-1_bold.nii.gz
│ │ │ ├── sub-003_ses-mri02_task-1back_run-1_echo-1_sbref.json
│ │ │ └── sub-003_ses-mri02_task-1back_run-1_echo-1_sbref.nii.gz

Unfortunately if I look in my output file from fmriprep

191015-08:56:30,989 nipype.workflow IMPORTANT:

Running fMRIPREP version 1.4.1:
  * BIDS dataset path: /project/xxxx/analysis/BIDS.
  * Participant list: ['003'].
  * Run identifier: 20191015-085619_1d98ff63-0f80-4427-8246-047980302c41.

191015-08:56:45,604 nipype.workflow IMPORTANT:
Creating bold processing workflow for “/project/xxx/analysis/BIDS/sub-003/ses-mri02/func/sub-003_ses-mri02_task-1back_run-1_echo-1_bold.nii.gz” (0.62 GB / 724 TRs). Memory resampled/largemem=2.47/6.94 GB.
191015-08:56:45,690 nipype.workflow IMPORTANT:
No single-band-reference found for sub-003_ses-mri02_task-1back_run-1_echo-1_bold.nii.gz
191015-08:56:47,514 nipype.workflow IMPORTANT:
Slice-timing correction will be included.
191015-08:56:47,730 nipype.workflow IMPORTANT:
SDC: fieldmap estimation of type “epi” intended for /project/xxx/analysis/BIDS/sub-003/ses-mri02/func/sub-003_ses-mri02_task-1back_run-1_echo-1_bold.nii.gz found.
191015-08:56:48,615 nipype.workflow IMPORTANT:
Creating BOLD surface-sampling workflow.

How does fmriprep search for single-band reference images? How can I fix it?
Thanks a lot!

Hi @Flaarn - looking at your directory structure, it looks like there is a difference between the subject’s session label (ses-02) and the information in the bold/sbref (ses-mri02). Could you try unifying the labels and rerunning with a clean working directory?

Hi @mgxd, Thanks for your reply! I’m sorry that went wrong while I copied it here: it is indeed called ses-mri02 also as a folder! If it helps I found out that other people at Donders Institute have the same issue (and I think most of us are using first bidscoiner and fmriprep installed as a singularity contianer)

This looks like a bug. I made a directory structure similar to yours and simulated the sbref querying used by fmriprep 1.4.1 with pybids 0.7.1, and layout didn’t find any matching sbref scans. If you can, try updating to a more recent version of fmriprep. Since the pybids requirement has increased, this problem should be fixed.