> Since I did not really find good information about this topic anywhere I
> would like to ask you, if
> there are any kind of draw-backs regarding of not using isotropic voxels
> (T1) as input images in
> SPM2 for optimized VBM.
High resolution data should be better, as there are less partial volume
effects. I can't think of any reason why the voxels need to be exactly
isotropic though.
> The script you provided writes isotropic 1 mm voxels as output image
> format, whereas the
> templates have 2 mm iso-tropic voxels. As far as I understand optimized VBM
> correctly, it
> segments twice once in native space and once in MNI space. If the image in
> native space is non-
> isotropic does this in anyway affect the re-application of the deformation
> field later on on an
> image, which is in isotropic voxel format?
I'm not aware of any such issues.
> So basically I am asking if you would advise to re-sample images to
> isotropic voxel size before
> running your optimized VBM script or do we not have to bother with it?
I would strongly suggest not doing any resampling beforehand.
> I am asking because we compared the output of VBM of resampled and
> non-resampled images
> and we could detect slight shape changes between both ( a few gyri were
> deformed slightly
> differently, which might indicate differences in spatial normalization).
I would expect to see slight differences, but the non-resampled data should be
slightly better.
Best regards,
-John
|