slow drifting on EEG

[kyungho]

Dear all,

Hello, I am posting my issue for any help

Now I am conducting P300 BCI speller paradigm, and some parameters are set as below:

bandwidth: [0.1 70] Hz,
Notch: 60 Hz
Fs: 512Hz
Channel: 32 (MK2 box, pin type electrodes) + 5 EXG(ECG, EOG, EMG)
Reference: CMS electrode to left ear mastoid, DRL electrode to right ear mastoid (flat type electrodes)
Environment: laboratory
Data collection: BCI2000 software

Problems:
Some slow drifts frequently occur during recording EEG, and it makes P300 classifier failed to generate feature weight.
With the same parameters, someone did not show slow drift, and someone showed slow drift over recording. There was no severe body movement during experiment (due to ear reference, all channels contain ECG component though).

Since I do not have rigorous knowledge for inspecting EEG yet, I attached picture containing drift for your help.

re.png (108.92 KiB) Viewed 28078 times

[relatively good]

bad-b.png (136.6 KiB) Viewed 28078 times

bad-g.png (132.02 KiB) Viewed 28078 times

[bad]


My concern is whether using ear reference is good choice for P3 ERPs. We have other CMS/DRL electrode (pin type) around parietal area.
What do you think of ear mastoid reference or head reference?

Thank you.

Regards,
Kyungho Won.

[Coen]

Always reference the data before further processing, see https://www.biosemi.com/faq/cms&drl.htm

Raw data (with respect to CMS) is NOT valid data, do not regard or use CMS or DRL as reference. With proper referencing, location of CMS and DRL does not affect the data (but may affect interference level), also see:

viewtopic.php?f=1&t=33
viewtopic.php?f=1&t=40
viewtopic.php?f=1&t=177

Best regards, Coen (BioSemi)

[kyungho]

Dear Coen,

Sorry for my ignorance. I think I made a big mistake.

Can I ask you very very basic things?

According to my understanding,

The signal appearing without referencing is voltage-difference between electrodes and CMS, and it's not supposed to be used

In other words, if I want to use left/right ear mastoid reference, do I have to attach two extra channels (maybe EXG?) to left and right ear mastoid,

locate CMS/DRL to any other place (e.g., head cap) , and finally subtract the average of left/right ear mastoid from EEG?

Am I understanding right?

Thank you.

Regards,

Kyungho Won

[Coen]

The above is fully correct.

Best regards, Coen (BioSemi)

[DavidLindhagen]

Thanks for the posts and replies on this--I have recently run into a similar issue. I understand that you must add channels to the mastoids for successful offline re-referencing, but is it necessary to also change CMS/DRL locations? For example, could you keep CMS/DRL on each mastoid AND add new channels for mastoid references?

Thanks!
DL

[Coen]

CMS, DRL and the reference electrode can all be on the mastoids. However, CMS in the cap (approx. in the center of the other electrodes) is theoretically a bit better (Common Mode more equally distributed among the electrodes). DRL can be anywhere on the body.

Best regards, Coen (BioSemi)