Page 1 of 1

Pairs of channels showing the same EEG activity

Posted: Mon Jan 08, 2018 7:16 pm
by nicolas
We are collecting EEG data with a Biosemi ActiveTwo box (either using 128 or 64 channels cap). So far we have collected data from 19 subjects with 128 channels, and 12 subjects with 64 channels. In all the data collected, when trying to run an ICA, we receive a message saying that "The proposed dimension for ICA is 124 (out of 128) channels" (when using the 64 channels cap the message is "... 60 (out of 64)". We have narrowed the electrodes to the following pairs:
A25 - A29
A26 - A30
A27 - A31
A28 - A32

The problem when running the ICA seems to disappear when an average reference is set before running the ICA (both with 128 and 64 channels).
Additionally, we try the bucket test (https://www.biosemi.com/faq/check_electrodes.htm) with the 32+CMS/DRL channel A-set and apparently there is no problem with the channel set.
Our first thought was a gel bridge, but we think is unlikely that the same gel bridge with more than 20 subjects is bridging the same pairs of electrodes.
At this point, we don't know what else to try to solve the problem. Any idea what is the problem and how to fix it?

Re: Pairs of channels showing the same EEG activity

Posted: Tue Jan 09, 2018 1:39 pm
by Coen
The listed electrodes are close to the CMS. In the unreferenced signals, there will be very little EEG activity, just electrode noise and 50/60 Hz interference (because of the reduced CMRR in unreferenced form). The noise and interference probably "fools" the ICA.

It is essential to always reference the data before doing any analysis. The raw unreferenced data are not valid signals, see viewtopic.php?f=1&t=33&p=86 and viewtopic.php?f=1&t=40&p=109 and viewtopic.php?f=1&t=177

Best regards, Coen (BioSemi)