Dear Gabriele,
Like John has stated, the between-modality registration is way off (corpus callosum is below the blue line on the structural volume, well above in the functional one). The affected subjects might indeed have moved between those two to a rather large extent, in which case the coregistration might fail as it requires a relatively good starting point.
However, given your preprocessing pipeline it's quite likely that you introduce some bias yourself: Instead of just reorienting the structural volumes manually you should apply these parameters onto the functional volumes as well; alternatively, e.g. in case they were acquired on different days / subjects were repositioned in between, reorient the structural volumes based on the structures and reorient the functional series based on e.g. the mean EPI images. Otherwise it's quite likely that the files in native space are in good alignment, by reorienting one part the discrepancy is increased unnecessarily (and possibly results in a failure of the automatic coregistration).
Best
Helmut
|