1. is there a way to pass an affine matrix (as stored in the
sn.mat-file) as a starting estimate to the segmentation routine in spm5?
No - but you can reorient your images via the Display button.
2. how can I derive the overall scaling as occurring during the affine
transform from the affine matrix ?
The determinant of an affine transform gives the relative volumes. The
cubed root of this may be one measure that can be used.
Background is, I know there is a way to disable affine registration for
segmentation in spm5 but that requires the images to be written out for
this step, introducing interpolation. If one could pass an an already
determined affine matrix to the algorithm, this would bypass this step.
Naturally, in order to then do modulation by hand one would also need
the overall affine scaling.
In earlier versions of SPM, I found that having the option to supply
starting estimates confused a lot of people, so I chose to simply assume
that images were oriented approximately correctly. The answer is
basically to reorient the images via the Display button.
Best regards,
-John
|