I have been trying to use brainstorm to perform source localization on (epilepsy) intracranial data. For two patients implanted with ECOG and depth electrodes, I computed a BEM head model for 5000 cortical sources locations from their MRI data with scalp, skull and brain (either 1000 or 1400vertices each, default conductivities). I then computed the constrained MNE current density maps with brainstorm's default parameters. In one case the simulated recordings look fairly accurate, but for the second patient the simulated recordings seem to show periods of activity almost identical in all channels and that shouldn't be there, as you can see on the image in the link below.
I suspect that this activity 'spreads' from one of the channels (presumably the first) during the procedure of projection to sources & backprojection to the channels, possibly due to the fact that for a large portion of the brain (that which isn't covered by grids) the reconstructed sources are noisy, to say the least.
The electrode localisation looks like this:
Do you have any idea what might be causing this issue and if there is any way around? Is it ill-advised to perform source analysis when a large part of the cortex isn't covered with electrodes as is typically the case with ECOG data?
I wanted to compute the sources to be able to estimate activity more reliably in between regions covered by electrodes (and get rid of noise originating from other parts of the brain), but I find from simulations that the further from the electrodes the noisiest the localization (even when the sources are not outside of the region of interest covered by grids but simply in sulci). Therefore, the advantage of performing source analysis rather than simply projecting the channels' activity on the cortex seems quite low.
I would greatly appreciate it if you had any insights on this issue or general advices!
Thank you very much for your help.