Hi
Technically it's all possible, though clearly this is not optimal. If you need to retain all subjects then the the easiest is to use FEAT to pre-process all data separately and then use applywarp to get them all into a standard space. You can then use the melodic command line to run concat-ICA and follow this with dual regression. You might want to add a confound variable indexing the different groups of protocols in order to reduce the chance of false-positives.
hth
Christian
On 11 Oct 2011, at 17:39, Torsten Ruest wrote:
> Yeah I thought about something similar, like coregistering the 45 slices to a 47 slices scan using 6 DOF, then take it from there, ie register it to the corresponding highres, concat the transformations (TM to the 47 with the one to the highres) and then update the registration using update_feat but I am a bit worried whether this is legitimate - I hope the small difference in acquisition would be neglectible.
>
> Thanks,
>
> Torsten
|