+ Check original region mappings.
+ Create one fake region per hemisphere in the connectivity matrix. These regions will represent the corpus callosum and the medial wall. They won't be connected to any other region (it's effectively padding the connectome with zeros)
+ For the surfaces: the vertices assigned to the 'fake' regions shouldn't be connected, therefore local connectivity be 0. To avoid sharp transitions at the boundaries, when computing the geodesic distance matrix, the distance for these vertices could be set to a large values (larger than a sensible cuttoff distance) so the local connectivity kernel will smoothly fall to zero and it will be effectively zero for the fake regions.
TODO: need to develop more the idea and how to make it work in TVB
+ decimate surfaces
+ create BEM surfaces
+ compute projection matrices
+ make sure to handle subcortical structures, that is, when the network size is nodes=vertices+subcortical regions.