Thank you very much for your quick response.
"It has a large negative value which is not infeasible but I'm still
surprised there is such a big difference between Brainstorm and SPM
and also the focality of the peak is so different"
The difference between the focality of the peak was present because we had incorrectly co-registered the electrodes for SPM, this has been fixed. We now get peaks that are close in the two gain matrices (within 1 or 2 vertices).
I just have two more questions :
1) what is the unit of the gain matrix created by spm ? I understood that it was microvolts but I'm not sure.
2) Why is there the need to wrap the subject's meshes (and other surfaces) in the MNI space for the forward model? This lead to the field D.other.inv{1, 1}.forward.mesh that have different coordinates and units that in the field D.other.inv{1, 1}.mesh.tess_mni.
I work only on the subject space so to avoid this wraping, I just replace the template surfaces and mri in the canonical folder of spm by those of the subjects in the subject space. This work well and avoid this step of wraping on the MNI space.
Thank you,
Gaelle Nicolas
----- Mail original -----
De: "Vladimir Litvak" <[log in to unmask]>
À: "Gaelle Nicolas" <[log in to unmask]>
Cc: "spm" <[log in to unmask]>
Envoyé: Vendredi 7 Janvier 2022 17:19:03
Objet: Re: [SPM] Question about leadfield for EEG source reconstruction
Dear Nicolas,
> 1) during the coregistration in SPM, the sensors are "scale until they fit to the head surface according to our visual judgement". Is there a way to not scale the electrodes ?
> I would like to use the transformation (rotation and translation) without the scale because in my case, the sensors position are real positions for each subject recorded during the experiment.
>
If you used both individually measured electrodes and an individual
MRI then no scaling is done exactly as you say. Otherwise, scaling is
still necessary,
> 2) why the vertex which contributes the most to the electrode (in the figure bottom right) is the one which has the minimum value ?
>
It has a large negative value which is not infeasible but I'm still
surprised there is such a big difference between Brainstorm and SPM
and also the focality of the peak is so different. Maybe try comparing
to the spherical model as well and plot the correlation between sensor
topographies for each vertex as a map as in this paper
https://www.sciencedirect.com/science/article/pii/S1053811914000172?via%3Dihub
. If everything works well the correlations should be above 0.9
everywhere even for quite different forward models.
> 3) if it is complicated to have the same gain matrix for brainstorm and SPM. Is it correct to replace the file called "SPMgainmatrix_..." by the gain matrix created by Brainstorm ?
You can't just replace one file with another but you could possibly
put Brainstorm lead fields in the same data structure that's in the
SPMgainmatrix_ file and replace what's there. As long as SPM then is
not triggered to recompute the lead fields, they should be used.
Would be good to get to the bottom of these differences. Please keep me posted.
Best,
>
> Thank you in advance for your response,
>
> Gaëlle Nicolas
> --
> Gaëlle Nicolas
> PhD Student
> GIPSA LAB/Equipe ViBS
> 11 rue des Mathématiques, 38402 Saint Martin D'Hères
> GIN/Equipe Neuroimagerie fonctionnelle et perfusion cérébrale
> Bâtiment E.J. Safran, Chemin Fortuné Ferrini, 38700 La Tronche
--
Gaëlle Nicolas
PhD Student
GIPSA LAB/Equipe ViBS
11 rue des Mathématiques, 38402 Saint Martin D'Hères
GIN/Equipe Neuroimagerie fonctionnelle et perfusion cérébrale
Bâtiment E.J. Safran, Chemin Fortuné Ferrini, 38700 La Tronche
|