It depends on whether or not your data are distorted, whether your voxel sizes
are correctly set, as well as how rigid the head and brain is. The brain
does slop about in the skull a little bit, but if the tilt of the head is
similar in both scans, then it should slop into similar positions in both
scans. There is also quite a bit of possible movement in the tissues outside
the skull, which the algorithm will also try to fit.
My best guess is that if a rigid body fit is not good, then it is something to
do with image distortion. Although not as distorted as EPI images, the
anatomical scans are still likely to have some degree of distortion. A
phantom could be used to check that the voxel sizes are the same as those
reported in the image headers, as well as the level of distortion.
If the images are good and the head is actually rigid, then a rigid-body model
should give the best results. If we know that zooms should be set to 1 and
there is no shear, then setting them to these values will be more accurate
than estimating them from the images.
Best regards,
-John
On Wednesday 17 September 2008 20:47, Henrique Amaral wrote:
> Hi SPMers
> The papers of validation of co-register methods don't explain what tecnique
> would be better, if a rigid-body transformation or a non-rigid-body
> transformation.
> In my work i coregister MR and SPECT in SPM, but some people in
> departament are questioning that non-rigid transformation is more
> appropriate.What are your opinion about this?
> It's possible update the coregister routine to perform non-rigid-body
> transformation, or would be necessary create a specific function for this?
>
> Thanks
>
> Henrique Amaral
> Clinics Hospital of Ribeirão Preto of the University of São Paulo - Brazil
|