Hi Pawel,
Sorry for missing this.
You seem to understand correctly.
In your first example (-mask_import), the mask is assumed to be in the final registration space. And with -regress_ROI, it will regress out the average time series over that ROI. That is a single regressor applied to all voxels.
For the second example, instead of using the single regressor, each voxel gets is own regressor: the Gaussian-weighted average of the white matter, weighted by distance to the given voxel. Since the example uses -mask_import, the mask is still assumed to be in the final voxel space (and on the final EPI grid).
The third example uses -anat_follower*, which implies the mask is aligned with the INPUT anatomical dataset (passed via -copy_anat). The mask will be eroded and transformed to the final space in the same way that the original anat is transformed. After that, the anaticor_fast method will be applied to generate a regressor for each voxel as in the previous example.
So yes, if in the third example the WM_anat_mask is already in the final space, that would not be appropriate. If that is the case, start with -mask_import.
- rick