Incomplete Gain matrix


I am trying to compute the gain matrix with the following specifications:

  • 63 EEG ch.
  • ~43000 cortical sources
  • The surfaces I am using are those generated by default (1922 nodes for head, inner- and outer skulls, and a cortical surface of 43000 nodes that I've uploaded and was obtained from the same MRI image)
    All the surfaces match perfectly and there is no overlapping. The electrodes are well positioned.

My problem when I "create head model" is that it gives me a Gain matrix with correct shape (63x43000) but only 3249 columns filled and the rest are zeroes. So it only solves the forward problem for 3249 nodes and I have no info (zeroes) for the other nodes.

Brainstorm does not report any error.

Let me know if I can provide you with further info or data.

Thanks for your help.

Are you using the OpenMEEG BEM model?
Have you tried with another forward model (3-shell sphere) and see if you get the same issue?

How have you generated and imported this cortex surface with 43000 sources?

Thanks François,

  • Yes I am using OpenMEEG BEM model.
  • With 3-shell sphere I don’t have any problem.
  • The 43000 nodes surface is a mesh generated using the freesurfer segmentation and the iso2mesh matlab toolbox. To import it, I generate the appropiate struct (surface.Vertices, surface.Faces, surface.Comment) and I use convert_cs to map nodes to scs space so that it fits the default 1922-node surfaces generated with OpenMEEG (outer, inner skull and scalp).

Why do you use this convoluted solution instead of using the default cortex surface available after importing the FreeSurfer output with the menu “Import anatomy folder” ?

Have you tried using the default cortex surface created by Brainstorm (15000V)?
and/or downsampling the high-resolution cortex surface to 43000 vertices using menu “Less vertices” and the option “reducepatch”?

Have you tried with less vertices on the various BEM layers?

(1) I need to use this specific surface to compare its reconstruction to some models that I have already on the same surface. Changing the surface wouldn’t be my favourite choice.

(2) I have just tried using the default 15000 surface generated by brainstorm and it doesn’t work either. The same error occurs (this time 1694 nodes have nonzero value and 13309 are zero)

(3) Using less vertices for the other BEM surfaces results in the same error.

A new version of OpenMEEG was just released, it might fix this issue.

First update Brainstorm manually (menu Update > Update Brainstorm).
The next time you use OpenMEEG to compute a forward model, the newest version (2.4.1) should be downloaded and installed automatically.