Transformation matrices different export options

Discussions
1

Dear all,

As I cannot transform directly to MNI in Brainstorm as my MRIs were not processed with Freesurfer before, I exported the volumes and normalized them to MNI space using fieldtrip/spm.
I would like to know intracraneal sEEG positions coregistered correctly with my preoperative MRIs in MNI space. I though I could use from my fieldtrip/spm normalizetion directly mri_norm (normalized mri ft structure).cfg.final transformation matrix. However, there is a problem.

If I export the MRI into Matlab as a brainstorm structure, I have the following transformation matrix:
T =

0.0149    0.9917    0.1277 -119.9746

-0.9998 0.0129 0.0161 79.6231
0.0144 -0.1279 0.9917 -77.3171
0 0 0 1.0000

ft_warp_apply(T, voxel) or ft_warp_apply(inv(T), electrode position) prefectly reproduces the coordinates in SCS (mm) or MRI voxel indices.

However, if I use my transformation matrix from my normalization step to transform the electrode positions in SCS to MNI space, I get wrong results. I realized that the transformation matrix of the exported to file (nifti) MRI is different than the exported to matlab one:

T(exported to file) =
1.0000 0 0 -84.0000
0 0.4688 0 -124.9824
0 0 0.4688 -131.0448
0 0 0 1.0000

The normalization to MNI space was done with the exported to file MRI and I guess there is an additional transformation done during export to file.

How can I get from the T to T(exported to file) so that I can use the transformation matrix from my normalization step? Maybe a stupid question but I get always confused with the transformation steps.

Thanks for your help!

Best,

Stephan Moratti

2

Hi Stefan,

The transformation saved in the .nii file is not the SCS (because no one would use it). The volumes are saved with no rotation, just a translation to the origin of the “MNI coordinates”. However if you didn’t compute the MNI transformation, we don’t know where the origin of the SPM/MNI coordinates is.

The code that defines the sform transformation in the .nii is in brainstorm3/toolbox/io/out_mri_nii.m, lines 97-106:

% Default origin of the volume: AC, if not middle of the volume
if isfield(sMri, 'NCS') && isfield(sMri.NCS, 'Origin') && ~isempty(sMri.NCS.Origin) 
    Origin = sMri.NCS.Origin - [1 2 2];
elseif isfield(sMri, 'NCS') && isfield(sMri.NCS, 'R') && ~isempty(sMri.NCS.R) && isfield(sMri.NCS, 'T') && ~isempty(sMri.NCS.T) 
    Origin = cs_convert(sMri, 'mni', 'mri', [0 0 0]) .* 1000;
elseif isfield(sMri, 'NCS') && isfield(sMri.NCS, 'AC') && ~isempty(sMri.NCS.AC) 
    Origin = sMri.NCS.AC + [0, -3, 4];
else
    Origin = volDim / 2;
end

Does this help?
I’m not sure why I did it this way… Let me know if you have a better suggestion.

Cheers,
Francois

3

FYI: We improved significantly the tools available in Brainstorm for processing and visualizing SEEG and ECOG data, including new options for volume coregistration. They are now documented in a new tutorial:
http://neuroimage.usc.edu/brainstorm/Tutorials/Epileptogenicity