Dear John and experts,
This is an unsophisticated observation, but as a preliminary and routine
check on my understanding of John Ashburner's spm_Deformations code, I
spatially normalized the canonical single subject T1 to itself (i.e., I
chose a copied version of this image as the image to be normalized, and I
chose the original image as the target) using SPM99. My idea was simply to
ulitmately see a Jacobian determinant image uniformly of ones after using
spm_Deformations on the output sn3d.mat file. But to my surprise, the
spatial normalization routine did seem to do something (albeit not gross,
but also not trivial) to the image, causing the corresponding Jacobian
determinant image to deviate seemingly subtantially from one (its range was
about .8 - 1.2).
I think my question does not concern spm_Deformations per se, but the
spatial normalization routine of SPM99. That is, is it to be expected that
using identical images (in identical coordinate frames) as source and
target in this algorithm would lead to a non-identity transformation? If
so, is this because the algorithm must take a non-zero step in parameter
space, even if this worsens the value of the cost function? I apologize if
questions are no longer being fielded for SPM99 algorithms.
Many thanks,
Eric
|