Filters in functional preprocessing should be orthogonal about fmriflows HOT 3 CLOSED

Additionally, I'm not sure how to do low-pass filters in a GLM. I think high-pass can be achieved with cosine filters (DCT), but I don't know how to specify the cosine filters for low-pass.

from fmriflows.

I think I'm getting there... So, CompCor, FD and DVARS are just estimations, while motion correction and temporal filters actually change the data.

Orthogonality between motion correction and temporal filters

Therefore, the motion parameters used for motion correction should first be temporally filtered. That should take care of their orthogonality. This is especially important in the context of high temporal resolution fMRI, (see Scheel et al., (2014)).
Also, if users don't want to apply a high-pass filter on their data, but only specify it in the GLM, then the motion correction will be as usual.

Open Questions

1. What's the correct order concerning the Friston24 components? First, compute Friston components, then apply temporal filters or reversed?
2. Should FD be computed on temporally filtered motion parameters or not?
3. CompCor, DVARS and the extraction of average signal (GM, WM, CSF) is computed on realigned files, should I remove motion artifacts afterward?

from fmriflows.

PR #20 takes care of many questions in this section. At the current standpoint, I believe that fmriflows implements the ideal combination of filtering and motion correction.

from fmriflows.