Saemann Philipp wrote:
> the voxels/points in native space should first be available as image,
> [...]
> Then you could try the normalise=>write function
This sounds fine, but it should also be possible to directly answer
the question "where does target voxel (x,y,z) come from in the source
image"
E.g. with rigid/affine registration the answer to this is:
inv(Vsrc.mat)*M*Vtrg.mat*[x y z 1]') where M maps from target
world-space to source world-space (as e.g. the output from spm_coreg
does, before it is included in the source voxel-world mapping)
With HDW, under SPM5, the answer is similar, but requires loading the
5D NIfTI, something like:
N = nifti('y_src.nii'); % (named after source, but in target-space)
src_world = squeeze(N.dat(x,y,z,1,:));
src_vox = Vsrc.mat\[src_world(:);1]
With sn.mat files (e.g. from "normalise" or SPM5's unified seg) I'm
afraid I don't know the answer. Hopefully John or someone else can help...
Best,
Ged.
|