fMRIPREP: Which regressors for best Global Signal Regression Comparison

Hi all,
I ran into Chris M. at OHBM and he pointed out that when using the aCompCor regression strategy, the cosine regressors should be used too (this is not documented, we should change that)

Can anyone verify that the comparison below makes sense:

We are moving from bash scripts that implemented Thomas Yeo’s pipeline at MGH to fmriprep, and want to be able to compare traditional global signal regression to CompCor and/or ICA-AROMA:

For the aCompCor regression:
aCompCor00, aCompCor01, aCompCor02, aCompCor03, aCompCor04, aCompCor05, X, Y, Z, RotX, RotY, RotZ, (and we should also be adding in: Cosine00 Cosine01 Cosine02 Cosine03)?

For the standard regression:
CSF, WhiteMatter, GlobalSignal, X, Y, Z, RotX, RotY, RotZ

What if anything, should be regressed out from that nii.gz file to be most similar to the above?

For consistency’s sake, we’ve removed the first four frame first, but could add them back in, then add some extra non-steady-state columns


There is already an issue open for this:

I would add Cosines (they perform high pass filtering) and motion parameters in all variants. Even if potentially redundant it should not hurt.

1 Like

BTW - it would be great to get your help improving the Cosine documentation. It should be an easy PR!

I’d be happy to, if you can point me towards some reference regarding where these come from so I can explain it better. Are these just a linear decomposition of the temporal band passing?


1 Like

Yup! See “GLM with Discrete Cosine Transform (DCT)” at

1 Like

Hi Chris,

I was considering using the cosine components to include in an FSL first-level analysis instead of using FSL’s built-in high-pass filtering tool. However, when setting up the GLM, you’re given a choice to high-pass all the regressors or not, with the recommendation that you should if thedata is high-pass filtered. I am unsure how all of this information fits together.

My feeling is that I should not use fmriprep’s cosine components and just use FSL’s high-pass filtering (with the same frequency setting as fmriprep), and then click yes to high-pass model regressors. Would this interfere with fmriprep’s confound EVs in any way if they were made assuming the data was already high-pass filtered?



This feels like a whole new topic - do you mind reposting?