Dear Satra and others,
Dont have the (complete)solution, but perhaps some thoughts that might be
helpful. I have also been thinking about this for a while, since i am
planning to re-map normalised activation coordinates (from another study by
someone else) to a subjects own brain (for stereotactic purposes.
The problem is that the normalization (which isnt an affine transformation
in the mathematical sense, by the way), is in principle not reversible, i.e.
the inverse transformation doesnt (always) exist. If i understand it right,
by the non-linear part of the normalization process 2 original locations
could potentially be mapped to the same point in MNI space (John, is that
true?), meaning that the inverse process is ambiguous. Th coregistration
(linear) part, which is an affine transfromation (allowing rotations,
translations, sheers and zooms) is reversible (the inverse matrix does
exist). For the latter it is sufficient to load the affine transformation
matrix (say M) from the mat-file, type inv(M) in Matlab and there you have
the reverse of an affine transformation.
During the SPM short course in London recently I asked John about this, and
he told me, if i remember well, that the inverse of the normalization does
in principle not exist, although in practice the non-linear coefficients do
not reach the point where the process gets irreversable. Since i didnt
examine the non-linear math of normalisation in detail yet, i cannot give
you more detail. I plan to get into this in the near future, I'll keep you
posted.
Regards,
Bas
--------------------------------------------
Dr. S.F.W. Neggers
dept. of Psychonomics,Helmholtz Institute
Utrecht University
Heidelberglaan 2
3584 CS, Utrecht, the Netherlands
Tel: (+31) 30 253 4582
Fax: (+31) 30 253 4511
E-mail: [log in to unmask]
Web: http://www.fss.uu.nl/psn/pionier
--------------------------------------------
-----Original Message-----
From: Satrajit Ghosh [mailto:[log in to unmask]]
Sent: vrijdag 23 augustus 2002 4:39
To: [log in to unmask]
Subject: Inverse Affine Transform
Dear Users,
I have a very simple transform question. I have an affine normalized T1
image from which I have extracted a brain surface. I would like to
transform this image to its original space. So essentially I have the
following:
- *_sn3d.mat file
- a set of vertices
Is there a one liner to do this?
Thanks,
Satra
|