Clustering EEG channels to few groups of electrodes

Hi Mike,

I am working on a resting-state EEG dataset of 54 healthy aging subjects. I did a power spectral analysis and now I want to explore if there is any association between EEG power in different frequency bands and behavioral cognitive scores in these subjects. I’d like to cluster electrodes into multiple groups like middle frontal, lateral frontal, temporal, and so on. could you please let me know where I can learn how to do that? and what electrodes are better to be averaged as, for example, middle frontal and so on? I read some paper that did this but they did not explain how they decided about what electrodes should be averaged as a specific group.

I have 54 electrodes that you can see in the attached picture.

I appreciate your response.


Hi Sam. I don’t know if there is a standard way to group electrodes like that. I guess if I were doing this, I’d just draw regions on the topoplot, thus basically doing it by eye based on what looks reasonable. Or you can find a paper where they for regions and then use their clustering.

One way I’ve thought about doing this is to convert electrode locations to MNI coordinates. These coordinates would be outside the brain, of course, but the nearest brain regions defined in, say, the Jülich atlas could be used to determine the brain regions closest to that electrode. There will be overlap, so some weighting factor might be necessary. For example, CP1 might be 70% sensory cortex and 25% parietal cortex.
Other issues are differences in brain anatomy and cap placement on the head. Digitized electrode locations would help with the latter, MRI scans with the former.