Poor Skullstrip Brain Clip

Summary of what happened:

Dear all,

We used fmriprep-20.2.7 to preprocess our data, but we are facing issues with poor skull-strip which very consistent across subjects (see the attached image, x = 18; y = 44). For context, our MRI data were acquired by GE scanner, and the T1w contrast look fine and there was no intensity in-homogeneity issue.

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

> usr/local/miniconda/bin/fmriprep /bids /output participant 
--skip_bids_validation --participant_label CAMH036 --nprocs 4 --omp-nthreads 2 
-vv --output-spaces T1w MNI152NLin2009cAsym fsaverage fsaverage5 
--bold2t1w-init register --bold2t1w-dof 6 --use-aroma 
--aroma-melodic-dimensionality -200 --use-syn-sdc 
--fs-license-file /license/license.txt --no-submm-recon -w /work


20.2.7 LTS

Environment (Docker, Singularity, custom installation):

Anaconda Python

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

data were organized following BIDS standard.

├── ses-posttx
│   ├── anat
│   │   ├── sub-CAMH036_ses-posttx_T1w.json
│   │   └── sub-CAMH036_ses-posttx_T1w.nii.gz
│   └── func
│       ├── sub-CAMH036_ses-posttx_task-rest_bold.json
│       └── sub-CAMH036_ses-posttx_task-rest_bold.nii.gz
└── ses-pretx
    ├── anat
    │   ├── sub-CAMH036_ses-pretx_T1w.json
    │   └── sub-CAMH036_ses-pretx_T1w.nii.gz
    └── func
        ├── sub-CAMH036_ses-pretx_task-rest_bold.json
        └── sub-CAMH036_ses-pretx_task-rest_bold.nii.gz

Relevant log outputs (up to 20 lines):

docker is /usr/bin/docker
[ INFO ] Descriptor from execution saved to cache for future publishing as descriptor_fmriprep-(BIDS-app)_2022-10-21T21:52:28.712050.json
[ INFO ] Input: {'bids_dir': '/bids', 'output_dir': '/output', 'work_dir': '/work', 'analysis_level': 'participant', 'fs_license_file': '/license/license.txt', 'nprocs': 4, 'omp_nthreads': 2, 'verbosity': '-vv', 'use_aroma': True, 'use_syn_sdc': True, 'skip_bids_validation': True, 'output_spaces': ['T1w', 'MNI152NLin2009cAsym', 'fsaverage', 'fsaverage5'], 'participant_label': ['CAMH036']}
[ INFO ] Running: chmod 755 /tmp/nxf.Oxw0wp6DkJ/temp-1061311154703-1666403548720.localExec.boshjob.sh
[ INFO ] Running: singularity exec --cleanenv -B /mnt/tigrlab/projects/ttan/ThirtyFourD/THREED_BIDS:/bids -B /mnt/tigrlab/projects/ttan/ThirtyFourD/in_progress:/output -B /mnt/tigrlab/quarantine/freesurfer/6.0.0/build:/license -B /tmp/nxf.Oxw0wp6DkJ/work:/work -B /tmp/nxf.Oxw0wp6DkJ:/tmp/nxf.Oxw0wp6DkJ -W /tmp/nxf.Oxw0wp6DkJ /archive/code/containers/FMRIPREP/nipreps_fmriprep_20.2.7-2022-01-24-5df135ac568c.simg /tmp/nxf.Oxw0wp6DkJ/temp-1061311154703-1666403548720.localExec.boshjob.sh
/usr/local/miniconda/lib/python3.7/site-packages/bids/config.py:40: FutureWarning: Setting 'extension_initial_dot' will be removed in pybids 0.16.
221022-01:55:46,988 cli INFO:
	 Telemetry system to collect crashes and errors is enabled - thanks for your feedback!. Use option ``--notrack`` to opt out.
You are using fMRIPrep-20.2.7, and a newer version of fMRIPrep is available: 22.0.2.
Please check out our documentation about how and when to upgrade:
221022-01:56:01,226 nipype.workflow WARNING:
	 Previous output generated by version 0+unknown found.
221022-01:56:04,188 nipype.workflow IMPORTANT:
    Running fMRIPREP version 20.2.7:
      * BIDS dataset path: /bids.

Screenshots / relevant information:

We have tried solutions to address this major skullstrip problem in our dataset

  1. Use --skull-strip-t1w option. The output did not improve.
  2. Use Roburst Brain Extraction (ROBEX). This also did not improve the skull-strip
  3. defaced T1w.
  4. We are currently trying to change parameters in the ANTS brain extraction to see if it improve the skull-strip.
  5. Would running with newer fmriprep version fix our current problem?

We would really appreciate your expertise and comments on how to improve this skull-strip issue.



Hi @ThomasTT ,

Here are a fe suggestion that may provide different (and hopefully) better results:

  1. try a newer version of FMRIPREP: fmriprep v22 which is leveraging freesurfer 7.2 which may give different results, along with other internal changes and configurations of fmriprep functions.

  2. Use other software for skullstripping:

  • FSL bet which different parameters for -f and -g. You may also improve your results by running ANTs `N4BiasFieldCorrection before skull stripping the brain.
  • Synthstrip
  • HD-BET

In that latter case you will have to use the --skull-strip-t1w skip option when running fmriprep.

1 Like