Dear Brainstorm developers,
I was wondering if there are any future plans to apply ICA decomposition on the source localized data (using the voxel time series data)?
It would seem like a great application in the context of identifying brain networks for connectivity studies (instead of using the seed based approach).
Best,
Lars
No, we do not have such plans at this moment.
Since the minimum norm source time series are only linear recombinations of the EEG or MEG signals, there are no additional independent components that could be "created" by projecting the recordings to source space. This massive linear mixing should not change what you can identify as independent components, it would just increase tremendously the duration of the ICA computation.
ICA at the sensor level can be an interesting path to modeling the dynamics between multiple "sources". If you consider that one IC corresponds to one brain source, you can select only this one component and localize it in the brain with a minimum norm/beamforming/dipole fitting approach. You can save also save the IC time series and do some connectivity analysis at the IC level.
I think it's more interesting to consider either IC analysis OR source analysis, not both.
Dear Francois,
I understand. Thank you for the explanation.
Best,
Lars