Robert Terwilliger wrote:
> I create an MNI152 brain by isotropically resampling avg152T1_brain (from
> ${FSLDIR}/etc/standard), using flirt, from 2x2x2 to 1x1x1 mm^3 voxels. The
> resulting image is 182x216x182 mm^3 voxels.
>
> Is there a specific coordinate in this MNI152 brain that is considered the
> standard "center coordinate"?
The centre is the origin of the world coordinate system, i.e. the
voxel which maps to (0,0,0) under the voxel-world mapping. You can see
this in the screenshot here:
http://www.cs.ucl.ac.uk/staff/gridgway/tmp/ss.jpg
On the left is the original avg152T1 and on the right is a 1mm
isotropic version (created with ApplyXFM, selecting the identity
matrix and entering twice the number of voxels, and 1mm voxel sizes).
As you can see, the world origin corresponds to the same anatomical
location in both, though the voxel coordinates necessarily differ.
The MNI space origin is apparently about 4mm above the anterior
commisure (from http://imaging.mrc-cbu.cam.ac.uk/imaging/MniTalairach
again).
If you have registered an image to the MNI/avg152 template with FLIRT,
then the mm coordinates of the transformed image reported by FLIRT
should correspond with MNI space.
> The context of my question is with regards to TBSS. I have run tbss on a
> series of FA images, and I currently have a mean_FA_skeleton
> affine-registered to the aforementioned MNI152 brain.
>
> Suppose I am interested in a voxel in the skeleton at (104,125,104) as
> viewed with fslview. Although the affine registration of the skeleton is but
> a close approximation of the MNI152 morphology, I would like to know what
> are the nearby anatomical structures at or nearby (104,125,104).
>
> Since my coordinates are in "voxel space", they are not the correct ones I
> would use as input to find the corresponding structures in an MNI152 atlas.
>
> Is there a straightforward way I could translate this particular coordinate
> into the "atlas space"?
If you record the world/mm coordinates reported by flsview for this
skeleton, then I think they should be in MNI space, and hence directly
usable with an MNI atlas.
> I don't know if I have merely restated my previous post
I'm a little worried that I have done this too! If I am missing your
point, hopefully someone else on the list can help...
Best,
Ged.
P.S. If for some reason you have FLIRT matrices, but not the
transformed images, then you can convert from voxel to world space by
first converting voxel coordinates to scaled mm coordinates (just
multiply by the unsigned voxel dimensions) and then applying the
stored transformation matrix (i.e. pre-multiply the vector
[scaled_x scaled_y scaled_z 1] by the 4x4 matrix.
|