# EEG region of interest

I have EEG data collected from 32 channel electrode. I measured EEG across an entire condition. Then segmented each condition in one-second segments, then fast Fourier transformation. So I am looking at the general activity of the brain during task performance, and not event-related.

I however want to examine differences in frequency bands across the parietal regions across 3 different conditions. I have computed the average frequency power bands i.e., alpha, theta, gamma, etc for all electrodes. I am wondering how to compute the average power for my region of interest, for example, can I add up the frequency power spectrum for the electrodes in the region I am interested in? eg, the alpha spectrum for left parietal region P3 + P7 ) or find the average (eg, the alpha spectrum (P3 + P7)/2 ). Are there better ways to do that?

Also does the correlation between electrodes represent coherence across regions?

Hi reubebe. Yes, you can definitely average power across different electrodes, and the formula you suggest ( [p3+p7]/2 ) is the correct way to go.

As for correlation between electrodes: the raw correlation between electrodes reflects a combination of coherence (which you want) and volume conduction (which is a confound for coherence), so it’s not the best approach. In your case, I recommend using a phase-lag-based coherence measure. There are many of them, which I discuss in my book, but the most common ones are imaginary coherence and wPLI.

1 Like

Thank you for your response, very helpful. A follow up question, in a 32-electrode EEG, if I want lets say I am interested in the left central- parietal region (CP1, CP5 … , is it necessary to include the C3 in computing the average?) and if I want to look the left frontal central (FC1, FC5 does C3 appears again?). If the answer to both is yes what happens to the Cz and FCz respectively in both examples do they get included ? .

Also with the phase-lag-based coherence measure, is it feasible even though my EEG measure is not event-related, I have taken almost all your courses and most examples seem event-related, apart from that do you have any reference to guide me with the phase-lag-based coherence?