I went through the excellent tutorial on EEG and epilepsy, and I want to try out localization with MEG and/or eCOG.
I am wondering if there exists an MEG or eCOG dataset labeled with e.g. interictal spikes or pre-ictal activity, and anatomical data to support a realistic source model. Ideally something sufficiently close to the tutorial examples so I can extrapolate (my BST expertise is still very limited).
The two other brainstorm tutorials on epilepsy don't attempt source localization (e.g. beamforming or MNE). I'm aware of MEG data for a FieldTrip tutorial that does, but the epileptic activity is not labeled so it's hard to judge the outcome.
I imagine that inverse models should be applicable to eCOG similarly as to EEG (but with better results) but I haven't seen any examples. For SEEG, as far as I understand, the BST tutorial localizes sources to the vicinity of electrodes.
@alain Currently, we don't have any complete (Ephys + MRIs) and labeled Ecog/MEG data to use for the tutorial. We will consider this for future integrations.
Yes, but it may require more attention when computing the forward model since in iEEG both the sources and the sensors are in the same area.
Since the sensors are close to the sources, so there is a high chance that the localized source will be located next to the sensors ~(the sensors close to the source are more sensitive than the others).
What I'd need ideally is (1) a sample of MEG, eCOG or SEEG data with labels indicating events or intervals that need localizing (e.g. interictal spikes or early ictal activity), (2) a "head model" with a matrix of location coordinates and a matrix of location-to-electrode (or sensor) gains, (3) a surgeon not too far off to give feedback on whether an outcome makes sense.
I appreciate that proximity between source and electrodes might invalidate an approximation, but there is always the option to exclude some electrodes. Non-uniform or non-isotropic conductivity might also be a problem?
(and the patient)