Hello Colin,
you should definitely coregister the structural volumes to your EPI series in case you want to apply the DARTEL parameters to your functional volumes as well.
Concerning missing data, you should use the same normalization procedure for the whole group. In your case, the only available option seems to be normalization of the functional mean to SPM's EPI template. You would have to make sure that there is not too much stretching in areas in which your data have some artefacts which are absent in the EPI template. Another option would be to rescan these three subjects of course... Leaving this aside, the number of subjects would still be < 20. It is generally recommended to have a reasonable large group of subjects to create your own high-resolution templates. Although there is no definite answer, < 20 should be too few.
Best,
Gabor
|