> as regards to the coregistration: Is the MI-obtimaization function "good
> enough" to yield reasonable results, if you use a 12-parametric
> transformation
>
> instead of the 6-parametric transformation?
>
> > Actually, SPM's coregistration only does a 6 parameter transform by
> > default, although it can be coaxed into using 12 - but not via the user
> > interface.
This is a question that will differ from dataset to dataset. The validity
depends entirely upon the assumptions that can be made of your data. If you
use relatively new and well calibrated scanners, then the voxel sizes should
be set pretty accurately in the header. If this is the case, then a
6-parameter rigid body model will be better. Perhaps the question should be
rephrased:
"Is my data good enough to use a six-parameter rigid registration model
instead of a 12-parameter affine model?"
Best regards,
-John
|