search

nipreps / smriprep

Structural MRI PREProcessing (sMRIPrep) workflows for NIPreps (NeuroImaging PREProcessing tools)
https://nipreps.github.io/smriprep
Apache License 2.0
131 stars 39 forks source link

Poor T1 MNI normalization #55

Open oliver-xie opened 5 years ago

oliver-xie commented 5 years ago

Dear fmriprep experts, we recently encountered an issue with our T1 skull stripping using fMRIprep. Part of dura mater was included in the brain mask leading to very poor EPI-T1 registration. The command used is as follows antsBrainExtraction.sh -a /data/sub-22146/anat/sub-22146_T1w.nii.gz -m /work/fmriprep_wf/single_subject_22146_wf/anat_preproc_wf/skullstrip_ants_wf/t1_skull_strip/T_template0_BrainCerebellumProbabilityMask.nii.gz -e /niworkflows_data/ants_oasis_template_ras/T_template0.nii.gz -d 3 -f /niworkflows_data/ants_oasis_template_ras/T_template0_BrainCerebellumRegistrationMask.nii.gz -s nii.gz -k 1 -o highres001_ -q 1 -u 1 image Using NKI template did not resolve the issue of skull stripping, below is the skull stripped image antsBrainExtraction.sh -a /data/sub-22146/anat/sub-22146_T1w.nii.gz -m /work/fmriprep_wf/single_subject_22146_wf/anat_preproc_wf/skullstrip_ants_wf/t1_skull_strip/tpl-NKI_res-01_class-brainmask_probtissue.nii.gz -e /templateflow/tpl-NKI/tpl-NKI_res-01_T1w.nii.gz -d 3 -f /templateflow/tpl-NKI/tpl-NKI_res-01_label-BrainCerebellumExtraction_roi.nii.gz -s nii.gz -k 1 -o highres001_ -q 1 -u 1 image I wonder if there is a solution to this. Thank you! Oliver

effigies commented 5 years ago

Thanks for this report. A couple things:

1) I think you should definitely report this upstream to ANTs. If they can help figure out how to better support cases like this, we can either incorporate their updated parameters or, if a code change is necessary, we can update to a new version of ANTs when it's available. 2) I notice the FAST tissue segmentation is much better than the skull stripping. Did you also run FreeSurfer? I believe @oesteban added some brainmask refinement using FreeSurfer tissue segmentation for cases like this.

oesteban commented 5 years ago

This could definitely be caused by the mask refinement

oliver-xie commented 5 years ago

@effigies Thanks for the advice. I will report it to ANTs. I did not run FreeSurfer and I am running fmriprep again with FreeSurfer. Will let you know if that improves the outcome.

oesteban commented 5 years ago

If you did not run FreeSurfer then please scratch this https://github.com/poldracklab/smriprep/issues/55#issuecomment-465786192

oesteban commented 5 years ago

@oliver-xie could you share this T1w privately?

oliver-xie commented 5 years ago

@oesteban I have shared the T1w images through google drive to your Stanford email address. I also checked with ANTs developers and was told to use another template.

Thanks for your help!

oesteban commented 5 years ago

Thanks, yes, I received the link. Which template did the ANTs devs recommend you?

oliver-xie commented 5 years ago

They did not recommend a specific template. Just told me to try another template.