John,
We have the same problem in coregistration. I tried MI registration with
partial brain to whole brain T2 (they got the same thickness and in-plane
resolution) but the MI doesn't seems to work (seems no registration).
T1 and T2 have pretty much same scan angle. Then we purposely tilted
partial T1 by 0.05 radian and 0.1 radian but after MI, the angle didn't
change at all. I wonder whether you could reproduce it in your site.
Just FYI, T1 is the object and T2 is the target.
Regards,
Kelvin
On Thu, 8 Mar 2001, John Ashburner wrote:
> | When we tried to coregister a T2 structural image to a
> | T2* functional image, the resulting coregistration seemed
> | reasonable in the xy plane. However, the coregistration in
> | the z axis was off by approximately 1 slice (slice
> | thickness=8 mm) in a majority of the subjects' data.
> |
> | Does anyone know why this might be happening ?
>
> There are any number of things that could go wrong, particularly
> because of the large dropouts and distortions that are present in
> fMRI data. The default coregistration method makes a number of
> assumptions, many of which are violated by many datasets.
>
> As an alternative, you may like to try the mutual information
> registration approach (set by the <Defaults> button). This
> makes much fewer assumptions, so may prove to be better for
> your data.
>
> The distortions present in fMRI data mean that a really good
> match can never be achieved using a rigid body transformation.
> Ideally, some kind of nonlinear warping should be used, but
> this is not currently available within SPM.
>
> Best regards,
> -John
>
|