Dear Klara,
> I'm using nonlinear registration combined with full search (viz FSL
> gui options for first level analysis) and keep getting a really bad
> results for two of my subjects (please see
> http://picasaweb.google.com/australan/FSL#). These very same
> subjects can be easily and correctly registered linearly, but the
> nonlinear registration keeps failing.
>
> In one of those examples it's guite understandable - brain is alligned
> with the skull wich can't happen with a BET image.
> I'm at a loss why the fMRI data are flipped in the other example even
> if I'm using full search and there was no problem with linear
> registration.
I am not sure what has happened here. It appears to me like it is the
functional->structural registration that has failed, and for that you
should always use 6dof linear registration. If you have indeed used
linear registration, did you provide it with a betted (skull stripped)
structural to use for the functional->structural registration?
If not, I think that is probably the reason things went wrong. It is
also conceivable that you would be better off without the full search
provided that your images are oriented in the same way prior to the
registration.
Good luck Jesper
|