search

edickie / ciftify

The tools of the Human Connectome Project (HCP) adapted for working with non-HCP datasets
https://edickie.github.io/ciftify/
MIT License
116 stars 157 forks source link

T1w anatomical normalization issue #171

Closed eburkevt closed 2 years ago

eburkevt commented 2 years ago

Ciftify completes without error on the subjects I've tried so far, but I'm getting what looks to be bad Ciftify-based MNI normalizations for the fmriprep preprocessed T1w images. I believe where MNI normalization occurs is in the "Registering T1wImage to MNI template using FSL FNIRT" step of ciftify_recon_all.

The output from a subject from our lab shows the issue (note the distortion along the top of the brain, particularly in the pre/post central regions)

image

Have anyone seen this before? Is there a way to prevent this or improve the normalization? Does the T1w normalization issue impact the analysis of the surface fMRI data produced by Ciftify?

Also, here is was I got for one the Ciftify example subjects (sub-50004 from the tutorial). This subject also shows some slight deformity in the pre/post central gyral regions, although not as extreme as in the above case.

image

Here are my settings ... System Info: OS: Linux Hostname: mica Release: 3.10.0-862.3.2.el7.x86_64 Version: #1 SMP Mon May 21 23:36:36 UTC 2018 Machine: x86_64 ciftify: Version: 2.3.3 wb_command: Path: /onrc/home/pipeline/data/pipelineb/home/Chyatt/onrc/data/Apps/workbench/bin_rh_linux64/wb_command Version: 1.5.0 Commit Date: 2021-02-16 13:46:47 -0600 Operating System: Linux freesurfer: Path: /opt/freesurfer-6.0/bin Build Stamp: freesurfer-Linux-centos6_x86_64-stable-pub-v6.0.0-2beb96c FSL: Path: /opt/fsl/fsl-6.0.3 Version: 6.0.3:b862cdd5 ---### End of Environment Settings ###---

Issue occurs with FSL v5.0.11 also.

edickie commented 2 years ago

Thanks for posting this. The good news is that the T1w normalization shouldn't be to detrimental to the surface based fMRI. The same (bad) warp is used for both the functional and surfaces - so even though they look weird - they should still be aligned to each other. In this way, the quality of the T1w to functional registration is more important.

While I say that - the deformity after in the MNI warp in your participant does look particularly bad. This type of deformity is known to happen with FSL's FNIRT (which is what ciftify is using). I beleive it has something to do with the brain masking? Might be worth checking FSL supports on neurostars to see if there are any fixes?

eburkevt commented 2 years ago

Thanks for posting this. The good news is that the T1w normalization shouldn't be to detrimental to the surface based fMRI. The same (bad) warp is used for both the functional and surfaces - so even though they look weird - they should still be aligned to each other. In this way, the quality of the T1w to functional registration is more important.

While I say that - the deformity after in the MNI warp in your participant does look particularly bad. This type of deformity is known to happen with FSL's FNIRT (which is what ciftify is using). I beleive it has something to do with the brain masking? Might be worth checking FSL supports on neurostars to see if there are any fixes?

Thank you for your reply. I will look into what I can do regarding FNIRT based normalization.