Dear Clark,
This is exactly the correct behaviour that we want!
The mm coordinates associated with the avg152 brains are true standard
space coordinates (MNI coordinates) and so we want to be able to keep
track of them even when the image is rotated (as you've done with
avwswapdim). You should find that if you picked a particular anatomical
point (e.g. the most anterior part of the right lateral ventricle) then this
would have the same mm coordinates in both the pre- and post-swapped
images (i.e. avg152T1 and tmp) because it is the *same* standard coordinate
for the *same* anatomical point, even though the voxel coordinate is now
completely different.
Hope that makes sense.
Deleting the orientation information obviously erases all knowledge of
the standard coordinates and gives you back the "scaled voxel" coordinates
which fslview defaults to.
One final, slightly tricky, note is that when you swap the left-right
(and avwswapdim
will warn you when this happens) some slightly unpredictable results can
occur in the current version of FSL, and we *do not* recommend changing the
left-right orientation during any pipeline. We also do not recommend mixing
images with different left-right conventions in the same pipeline, and
given that
the avg152 images are "radiologically-ordered" that is what we recommend all
images are left as. If you do create neurologically-ordered images with
avwswapdim
at the moment, the above explanation about keeping standard coordinates
constant
does not actually work. However, this is not what you are doing, so
this is not
directly relevant to your particular question - I just thought I'd be
complete in
my answer.
All the best,
Mark
Clark Johnson wrote:
> Greetings FSL superstars:
>
> I recently used avwswapdim on a HiRes Structural and subsequently was
> able to produce the same effect using avg152T1.
>
> So, for example
>
> avwswapdim avg152T1.hdr z x y tmp
>
> Generates a tmp.nii.gz. When I opened tmp with fslview and iterated
> the x-voxel dimension, the y-mm control changed. I've tried this with
> a variety of other images and they don't behave this way ( even
> without adjusting the header - as below). With the limited data I
> have it might only happen when the voxels are cubic (e.g. 1x1x1 or
> 2x2x2) -- but that is really only a guess.
>
> Removing the orientation per
>
> avworient -deleteorient tmp
>
> returns the fslview(er) behavior to what we would expect.
> Is this expected or possible a small "bug-let"?
|