fMRIPrep CompCor and cosine regressors: Include together with high-pass filtering in SPM12?

Hi,

I preprocessed fMRI data using fmriprep which produces CompCor and cosine regressors for each participant. Subsequently, I performed a first-level analysis in SPM12.

I would like to include the first 5 CSF and WM CompCor components, respectively, to account for physiological noise.

In the fMRIprep documentation (Outputs of fMRIPrep — fmriprep version documentation), is says that, as CompCor are calculated before temporal high-pass filtering, all cosine regressors should be included whenever you include CompCor regressors in a first-level design.

However, it also says that you should NOT include cosine regressors when the subsequent analysis includes separate high-pass filtering. As I understand it, high-pass filtering is included by default in SPM12 first-level fMRI analyses.

So my questions: Should I drop the cosine regressors from my GLM and let SPM perform its standard high-pass filtering, or should I include the cosine regressors and switch off high-pass filtering?

Am I misunderstanding something here?

Appreciate any feedback!

(Post is similar to High-pass filtering with fmriprep and FSL - #4 by Mkassaie, but concerning SPM instead of FSL).

Best,
Milena

Hi @Milena,

Do this. That way you don’t risk using a different filter cut off that would leave residual effects of drift in your compcor components.

Best,
Steven

Hi @Steven, thanks a lot for the quick reply!