Dear Akos,
of course, it would be best to avoid these task-correlated artifacts, especially if they are not only present in the skull, but also on the boundary surface between grey matter and CSF. Does your design include a motor task (e.g. response) or eye movements that might induce head movement as well? If you are still acquiring data it might help to review subject instructions to emphasize the importance of not moving the head. Also, you should ensure good positioning and fixation of the subjects in the scanner.
Unless absolutely necessary, I would not recommend to run lots of "repairs" before statistics. In many cases, inclusion of realignment parameters as regressors in first level statistics should explain movement related signal change away. You might also try to use Realign&Unwarp instead of Realign, even if you do not have a fieldmap or voxel displacement map to correct for overall distortion.
If you are primarily interested in group results, it should be sufficient to mask your group statistics. You could use a mask created from grey and white matter segments (wc*) of your subjects, or a mask based on SPMs tissue probability maps.
Hope this helps,
Volkmar
|