In FLIRT the origin is the 0,0,0 voxel coordinate, which is also the
coordinate, and is the left-most, bottom-most coordinate: i.e. all
go from 0 to N-1. The saved transforms relate to mm coordinates (which
obtained by multiplying the voxel coordinate by the voxel dimension in
Hence the translations you are interested in apply to the 0,0,0
The transformations totally ignore the origin set in the header file.
Talairach transforms, to get to MNI/Talairach mm conventions rather than
FLIRT mm conventions, use this origin.
Hope that answers your question.
All the best,
On Wednesday, February 18, 2004, at 06:31 pm, Christopher Joseph
> Dear list,
> I need to understand the logic of the transformation matrix produced by
> FLIRT. Specifically, how do the translation parameters (last column)
> relate to the "world" coordinates of the volumes in question?
> Example: T1 -> MNI. The T1-image has an (arbitrary) origin of
> (0,0,0)vox, dimensions 170x250x180 and voxel size 1mm^3. The MNI-space
> is of course centered at (46,64,37)vox; 91x109x91; 2mm^3.
> The 12-parameter "solution" is good and given by:
> 1.01368 0.0243093 0.027123 -4.2556
> -0.0721973 0.975255 0.28695 -36.0692
> -0.0102316 -0.335421 1.04256 25.7862
> 0 0 0 1
> What is the translation (-4.2556, -36.0692, 25.7862) (mm) relative to?
> I've read postings in the archives
> but cannot get this! I am missing something...
> Christopher Bailey
> Research Assistant (MSc)
> Center for Functionally Integrative Neuroscience (CFIN)
> Aarhus University Hospital, Denmark
> [log in to unmask]
> phn: +45 89 49 43 78
> fax: +45 89 49 44 00