I'd suggest looking at the Coreg and Realign options separately to
identify where the problems occur. I suspect that it may be the
coregistration, rather than the realignment that is having the
problems.
I'm not entirely sure of the cause of the problem but would suggest
using CheckReg to see how the initial alignment of the scans is. Also
look out for artifacts in the images. In particular, check out the
amount of intensity non-uniformity in both the scans you coregister.
Then there are a bunch of other fairly basic things, such as making
sure that the gaps between slices are included in the voxel sizes and
being aware that rigid-body alignment will not correct the nonlinear
spatial distortions that usually occur in EPI images.
Best regards,
-John
On 19 October 2010 20:34, Pierre Gravel <[log in to unmask]> wrote:
> Dear SPMers
>
>
>
> I have a question about coregistration.
>
>
>
> I can successfully coregister mean EPI images (4 x 4 x 4 voxel size) with their associated MPR images (1 x 1 x 1 voxel size) when I use the estimation options:
>
> realignment (estimate & reslice): separation of 4 mm, smoothing of 5 mm
>
> Coregister (estimate): separation of [4 2] mm.
>
>
>
> On the other hand I get a good alignment only 50% of the time when coregister new mean EPI images (3 x 3 x 3 voxel size) and MPR images (0.94 x 0.94 x 1 voxel size) with the following estimation options:
>
> realignment (estimate & reslice): separation of 3 mm, smoothing of 5 mm
>
> Coregister (estimate): separation of [3 1.5] mm.
>
>
>
> I tried this with both SPM5 and SPM8 with the same results.
>
>
>
> Why is coregistration worse when I use smaller voxels and adjust the estimation options accordingly? Did I forget something totally obvious?
>
>
>
> Thanks in advance
>
>
>
> Pierre
>
|