Creating connected WM mesh for EEG localization

[grova]

Hello everybody,

I m working in the context of EEG source localization using distributed source approaches.
I m using brainvisa and notably the anatomical pipeline to extract the white matter mesh of a T1 MRI to generate my source space.

Would it be possible to generate a WM mesh of both hemisphere connected together through the corpus callosum and the brainstem. It would be very usefull when using some realistic spatial smoothness contraints in the EEG inverse problem.

Thanks in advance for your help

Christophe

[Jean-Francois Mangin]

Well,
do you need that for practical reasons? in which case you can just merge the two meshes (There should be a AimsMeshCat somewhere). If you really want a continuous mesh, I am a bit surprised. The cortex, indeed, is interrupted at the level of corpus callosum, so there is no real reason to build a continuous mesh?

If you really want it:
There is in anatomy/triangulation a process giving the whole brain mesh. It should not be difficult to modify this process in order to get the white matter spherical mesh. Have a look at the process giving the half white matter mesh as model.

[riviere]

Just a precision: the command to merge meshes is AimsZCat
Denis

[grova]

Hello ...thanks for your answer, actually I was already using AimsZcat to fuse in one mesh both hemispheres.

Concerning the connection through the corpus callosum, it is clear that there are no EEG sources there, but connections thanks to WM fibers, so I would suggest that it may be correct to think that neurons are connected through the corpus callosum.

Moreover, concerning the brainstem, where we may find EEG sources, the cut between Left and Right hemispheres is artificial.

If possible, I would really need a fully connected WM mesh, notably when including some prior about spatial smoothness in EEG source localization. Such method is using a discrete Laplacian computed on the surface. We observed very strange behaviour when both hemisphere are disconnected (ex, the method put one whole hemisphere completeley positive and the other completeley negative) ...

By making just an articial link, manually for fiew vertices of the left and right brainstem,s those instabilities caused by the Laplacian disappeared ! That s why, I dont want to put any EEG source within the corpus callosum, but I was wondering if it was possible to get nice and realistic connection between sources of both hemisphere through the brainstem and the corpus callosum !

What do you think ???

[riviere]

Hi Christophe,

The pipeline is designed to process each hemisphere separately. If you want to process both together, you are going outside of the standard procedure. However there should be some things that you can do, but you'll have to hack a bit:
For instance, in the pipeline, after the hemishperes split (which is basically the step you want to remove), you could try to provide the brain mask as "brain_voronoi" parameter to make it process the whole brain instead of a single hemisphere (but you'll certainly have to change the brain mask value: the mask is value 255, and brain_voronoi is 1 and 2 for R/L hemispheres, you can replace 255 with 1 or 2 using AimsReplaceLevel or AimsLinearComb), and see what the following steps do.
Otherwise you will have to get to lower-level things, either in BrainVisa with parts like "Ana Compute Hemi Grey White Classification", or directly use some commandlines (you'll have to read the code of BV processes to get the list and order of commands calls.
These are just ideas that I have not tried myself...

Denis