I wish to normalise my T1 (to the MNI template) and use the obtained
parameters for normalisation of the EPIs, as is fairly standard. In
order to 'aid' the normalisation, I want to set the orientation to the
AC manually in the Display function, but I have a few questions about
this, that keep getting me confused:
- Does it matter if I set the orientation on the T1 before or after the
coregistration of the T1 to my mean_EPI?
If you coregister an image to another, and then move one of them then
they will no longer be coregistered. However, if you coregister and
then move all the images in the same way, then they should remain
coregistered. This is why the Display button allows a transform to be
applied to multiple images, and not just the one that is displayed.
- If I draw a lesion mask in MRICro, to be used in cost
function masking, should I do this on the coregistered T1, and should I
do this on the T1 before or after having set the orientation manually in
SPM?
I don't know enough about MRICro to properly answer this one. I believe
it does not handle the image orientation information so well, so you may
need to copy the appropriate orientation information to your lesion
mask. This bit opf code may work...
P1 = spm_select(1,'image', 'Copy orientation FROM');
Mat = spm_get_space(P1);
P2 = spm_select(1,'image','Copy orientation TO');
spm_getspace(P2,Mat);
Conceptually, my problem is that I would assume that if you change
information on the T1 (e.g. its orientation), you cannot use the
parameters obtained from normalising that changed T1 to normalise EPIs
that were not changed in the same way.
That is correct.
Or does setting the orientation not change the position of my T1
relative to the EPIs?
It will change the position relative to the EPIs. The orientation
information is essentially a 4x4 matrix that maps from voxel indices to
some real world coordinate.
xyz = Mat(1:3,:)*[i,j,k,1]'
Mapping between the voxels of pairs of images can then be done by
multiplying the inverse of one matrix by the other matrix. This maps
from voxels in one image to the real world coordinate system, and then
from the real world coordinate system to the voxels of the other image.
Likewise, if I draw a lesion map on a 'raw' T1 and then later change
the orientation settings of that T1, or coregister it to other images, I
would assume that I cannot use that unaltered lesion map to overlay it
on the altered T1 during cost function masked normalisation ...
Right again, although I believe that MRICro would lose the relevant
information anyway.
The Check Reg button is very useful for figuring out what SPM considers
to be the relative orientations of images. If in doubt about whether
images are in register, then try the Check Reg button.
Best regards,
-John
