Hi,
I used SPM with fairly good results in the past for this purpose. Problem is that EPI volumes usually have a slightly different shape from the anatomicals to begin with due to the distortions caused by tissue boundaries, which distortions also vary as your subject is moving. For more info on "susceptibility-by-movement interaction", and on corrective methods implemented in SPM5 see the SPM5 manual (currently at [
www.fil.ion.ucl.ac.uk]) page 26 (mid-page).
(Another issue with taking the part SPM/part AFNI route is that one'd have to convert the files back and forth between SPM and AFNI.)