Fmriprep outputs for fsl analysis

[cchang]

Hi all,

I’m interested in using the fmriprep outputs for a fsl feat analysis.

It seems like there are a few existing threads (1 and 2), but was hoping to clarify a few things.

1. fmriprep does not give us the skull stripped preproc bold, therefore we need to apply a brain mask for all downstream analysis
2. Is it recommended to do physiological regression on skull stripped preproc bold? If so, why doesn’t fmriprep include a skull stripped preproc bold outputs?
3. I also do physiological regression with fsl_glm prior to running fsl first level feat. I’m using the output from the physiological regression analysis as an input for fsl first level feat. Does anyone have any recommendations on combining in the first level vs. doing it separately with fslglm?
4. One more question about registration! My registration after running fmriprep full pipeline looks a bit off. The image in the back is the bold data registered to standard space, and the one in the front is standard space.

Any feedback and thoughts would be appreciated! thanks all

[Steven]

Hi @cchang,

@cchang

fmriprep does not give us the skull stripped preproc bold, therefore we need to apply a brain mask for all downstream analysis

Yes

@cchang

• Is it recommended to do physiological regression on skull stripped preproc bold? If so, why doesn’t fmriprep include a skull stripped preproc bold outputs?

Denoising is fit on voxels individually, so whether or not it is skull stripped is not important for that.

@cchang

• I also do physiological regression with fsl_glm prior to running fsl first level feat. I’m using the output from the physiological regression analysis as an input for fsl first level feat. Does anyone have any recommendations on combining in the first level vs. doing it separately with fslglm?

You should do all denoising and first-level modeling in the same step.

@cchang

1. One more question about registration! My registration after running fmriprep full pipeline looks a bit off. The image in the back is the bold data registered to standard space, and the one in the front is standard space.

There is unfortunately not enough information in this post for me to help with this.

Best,
Steven

[cchang]

As always - thanks @Steven !

Would it be more fitting to open a new “software support” thread for the last question? Please let me know what information I can provide. I wasn’t sure if it was a recon issue or somewhere else along the pipeline since a different dataset that I’m preprocessing seemed to have a really good alignment in the MNI152 space.

[Steven]

No problem! Sure you can open new thread.