Dear Helmut and Marko,
I guess I can boil down the issue to the fact that I get the same mean
functional image if I manually rotate some of the volumes as when I
don't manually rotate some of the volumes - and this doesn't make sense
to me!
To clarify. I set my origin in the structural and functionals to match
my template, then manually reorient functional volumes that show
excessive rotation. When I then run realinment (reslice mean only) and
co reg structural to the mean functional, the mean functional is rotated
and does not line up with the structural. I get the same result as if I
had not manually reoriented the funtional vols that were rotated.
A
On 03/07/2014 19:01, Helmut Nebl wrote:
> What do you mean with "they are not aligned anymore" and "It seems as though realignment ignores the changes that I make". When looking at them with Checkreg? Or based on the plot of the realignment parameters?
>
> As Michael has already stated, there's world space and voxel space. The header part includes some information on how to convert voxel to world space. This is (hopefully ;) constant across time for raw data. However, the subject moves in space. During "Realign: Estimate" or "Realign: Estiamte & Reslice" (with only the mean volume resliced) this voxel to world matrix is updated for each of the files. The same holds if you reorient the volumes manually. If you then forward manually reoriented files into automatic realignment the different voxel to world matrices should be taken into account, but they should be ignored for the realignment plot / the rp file. The idea is to have the correct voxel to voxel transformation between different volumes, which you can add to the model and thus try to control for remaining inaccuracies / interpolation errors.
>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|