Dear John and other SPM'ers
I must confess, that although I really like spm_orth_views in general,
and think it is a fantastic tool, I have always found the
reorientation facility to be counterintuitive the few times I have
used it, but I never found out exactly why. Yesterday I found out and
I would like to share my discovery with you: I have been doing my
transformations starting from the top of the list i.e. first
translations then pitch then roll and so on. I now realise that
because spm_matrix is used to convert my parameters to the
transformation matrix I should really do this in opposite order (the
same order as spm_matrix applies the transformations in). Just like it
is written in the help text of "Display Images":
" The images can be re−oriented by entering appropriate
translations, rotations and zooms into the panel on the
left. The transformations can then be saved by hitting the
"Reorient images..." button. The transformations that were
applied to the image are saved to the header information of
the selected images. The transformations are considered to
be relative to any existing transformations that may be
stored. Note that the order that the transformations are
applied in is the same as in spm_matrix.m."
Given this behaviour I think it would be more intuitive if the buttons
on the panel to the left had resize on the top, and also it would be
helpful if this was the order which the transformations appeared in
the help text:" The images can be re−oriented by entering
appropriate zooms, rotations and translations into the panel on the
left."
But actually I think it would be better so much better if an alternate
version of spm_matrix was not used in this case. I understand that it
would be problematic to change the order of the operations for the
realignment and coregister functions and this is not what I am
suggesting. For manual adjustments however, I think A=S*Z*R3*R2*R1*T
would be a much better choice.
Let me illustrate this with an example: Say I wanted to manually
construct an affine transformation which transformed a structural
image in to MNI space (Of course, in the real world, I would use SPM's
excellent spatial normalisation but this is an illustrative example).
I would press CheckReg and select my own anatomical image and a target
which was already in in MNI space. Next I would right click on my
anatomical image and select "Reorient Image(s) -- Current image" With
the new formulation of the A matrix (A=S*Z*R3*R2*R1*T) the procedure
would now be the following:
* First I set the cross hair position to (0,0,0).
* Next I locate AC in my anatomical image in world coordinates say
(9,27,-27).
* Then I specify the translation : (-9,-27,27) and then (0,0,0) is now
in AC.
* Then I locate PC in the new coordinate system say (0,-26,-3).
* I then perform a pitch rotation of atan(3/26) radians and PC is now
in (0,-26.1725,0).
* Then I would determine the apropiate scaling factors and input these
e.g. 1.1(x) 1.15(y) and 1.35(z).
* The Images are now registered.
Now try to do the same thing with the current implementation of A
(A=T*R1*R2*R3*Z*S). I will skip the zooming step, because before the
images are in AC-PC it makes no sense to estimate the scaling in x,y
and z.
* First I would loacate AC (9,27,-27) and then PC (9,1,-30).
* I would calculate the needed rotation i.e. AC-PC=(0,26,3) gives of
course the same atan(3/26) pitch rotation as before.
* I would then either relocate AC, or calculate its new coordinates:
(x,cos(atan(3/26))*y+sin(atan(3/26))*z,-sin(atan(3/26))*y
+cos(atan(3/26))*z). If I am lucky I will get it correct: (9, 23.7272,
-29.9169) which I will enter as the translation.
I think it would be a significant improvement to the already great
display programs in SPM if this behaviour was changed. What do others
think?
Best
Torben
Torben Ellegaard Lund
Assistant Professor, PhD
The Danish National Research Foundation's Center for Functionally
Integrative Neuroscience (CFIN)
Aarhus University
Aarhus University Hospital
Building 30
Noerrebrogade
8000 Aarhus C
Denmark
Phone: +4589494380
Fax: +4589494400
http://www.cfin.au.dk
[log in to unmask]
|