Hi Janina,

1) That is what is done in example 11. Actually, it regresses
not only local WM but average WM, as you noted.

2) It is both. See -regress_anaticor_fast.

3) The ventricles are good to use because they should not have
any draining veins. Example 10 uses 3dSeg for segmentation,
which is not as good because the CSF is not restricted to the
ventricles. Getting the first 3 principle components seems to
work well. Steve Gotts used that method here.

Note that example 11 does not use bandpassing.

Along those lines, example 11 might be good for high motion
subjects. One might rather censor more and avoid the bandpass
step (which might otherwise drive most subjects toward zero
remaining degrees of freedom, or worse).

Do you have physiological signals?

- rick