Dear Armin,
FNIRT does not use the EPI directly.
It only registers the T1 to the atlas.
You should be able to independently check if this registration looks good.
Assuming that the above is fine then the issue is the registration of
the EPI to the T1 image. This is done with FLIRT (and field map
distortion correction if you have them) and you should check this independently.
If the above also looks good then the remaining problem *might* be due to
the combination of the two registrations if the EPI is slightly misregistered
and ends up going outside of the T1 brain at all, since the FNIRT warps can
sometimes overstretch things just outside of the brain.
Anyway, please check on the above and let us know where exactly the
problem lies.
All the best,
Mark
On 2 Jan 2012, at 19:45, Armin wrote:
> To whom it may concern ,
>
> I want to know, for doing nonlinear registration (FNIRT) in melodic for instance, should I have the hole brain in EPI images? Since nonlinear registration just perform between T1 (high resolution) and Atlas (standard space), I believe it is not necessary. But it seems FNIRT try to fit EPI images (which are just part pf the brain) to the hole brain of MNI atlas?!!!
>
>
> Bests,
>
> Armin,
>
|