Dear FSL folks,
I'm trying to understand some of the guts of FSL's registration tools,
and I was wondering if the following is correct:
Let's say that I register original.img to standard.img using flirt
> flirt -in original -out registered -ref standard ....etc
and it produces a matrix original2standard.mat and a
original2standard.xfm
Now, let's say that I choose a point xyz from the original.img and read
the coordinates.
I should be able to read in the original2standard.mat into matlab and
do the following operation:
Amatrix = load ('original2standard')
xyz2 = Amatrix * [ x y z 0]'
The actual questions:
1) Should the xyz2 coordinates should be the coordinates for the same
structure after the transformation (ie- registered.img)?
2) is the frame of reference for that rotation the same as the one in
the reference image, the original image (ie- origin, voxel sizes...),
or is it relative to the first voxel in the file?
3) I should also be able to invert the transformation matrix and go
back and forth between the spaces, no?
4) Do all these things hold true in the case of 12DOF transformations?
I'm thinking they should ...
5) what is the .xfm file for?
I'm sorry if this is a newbie question...I've been trying to test these
questions on a synthetic image made up of a few dots but I'm getting
weird results. Maybe I'm using the functions wrong ...
thanks a lot
-Luis
--
( http://www.eecs.umich.edu/~hernan )
