[Eeglablist] Average Reference after measuring 128 Ch using one of the 128 channels as reference (Cz)?

Ernesto Gonzalez Trejo gonzalez at snn-unit.de
Tue Dec 1 01:35:14 PST 2015

Dear Yao,

thank you very much for your suggestion. I will review it as an alternative.
However, my question is still unanswered (as i am not looking for alternatives
to AR but instead i'd like to understand it better), so i will repost it here:

Hi everyone,

i have a brief question regarding a correct average referencing in EEGLAB, that
until now i've been unable to solve. I've read the EEGLAB tutorial and also
searched for some answers in the EEGLAB list but until now i didn't find one
that helps me solve the problem.

I am doing EEG measurements with 128 Channels, using the channel in the position
Cz as reference(ch. 30 in my cap). During the acquisition, the amplifier
properties have Cz set as reference for all electrodes, and  the Cz electrode is
then measured then against the ground. Additionally, two more channels (63-64)
are on the earlobes and are usually discarded since i'm not using them as
reference or for data analysis.

Given the nature of the measurement I consider the average reference the way to
go in order to represent better the brain activity.

The acquisition is done through SIMULINK and saved into a MAT file, for
post-processing in MATLAB. (The output is a 128xL matrix, where L is the length
of the measurement in samples).

The problem i have is that i can't seem to indicate correctly to EEGLAB that the
reference used in my measurement is also part of the channels. When i import the
mat file, i cannot give explicitely the channel (as stated on the import
window's help:

"Ref. channel indices or mode" - [edit box] current reference. This edit box
cannot be edited. To change data reference, use menu
Tools > Re-reference calling function pop_reref(). The reference
can be a string, 'common' indicating an unknow common reference,
'averef' indicating average reference, or an array of integer
containing the indices of the reference channels."

Then, after I import the data without a reference, I select the channel
locations (using a self-made loc file for the cap i use). Here, each one of the
128 channels has a label (10-10 system names for electrodes 1-64 and just the
electrode number for 65-128). Once loaded, i click on "Set Reference" and write
1:128 as channel indices and Cz as "Reference". (Channel 30 has the label Cz and
was the reference in the measurement). After the locations are set, i remove
channels 63 and 64, since i don't want to use them for AR.

So far, so good. After this i can just go and compute the average reference
using the tools from EEGLAB. I wanted to validate the results so i kept doing
some tests, and i realized that if i clicked on tools-rereference and then to
"add current reference channel back to the data", the error :

"There are no Reference channel defined, add it using the channel location
editor"

appeared, although in the main EEGLAB window the Reference is stated to be Cz
and Channel locations as "Yes".

Question 1:

Could it be that i am giving the Reference wrong in the channel locations? After
reading the tutorials i only got that the reference is usually given as an extra
channel, which is appended to the locations. In this case, it is not an extra
channel, but channel 30 of 128. Is it wrong then, to input this channel as i did
in the channel locations?

Question 2:

Since i wanted to see how exactly the average referencing was working, I checked
also the reref.m function from EEGLAB and tried to run it standalone as follows:

output=reref(eeg,30,'exclude',[63 64],'keepref','on');,

where i am only interesed on the average-referenced EEG (output) and i give a
128xL matrix (my EEG measurement), I indicate my reference was the channel (row)
30, and i exclude the earlobes (63 64).

The output is however very different to the one i obtain with the method
explained above in EEGLAB.

In order to obtain the (mostly) same result as in EEGLAB i have to input the
following:

output=reref(eeg,30,'exclude',[30 63 64],'keepref','on');

So, what exactly am i doing wrong here? I am probably specifying the reference
in a wrong way. How should one manage such a reference system within EEGLAB or
in the reref function? As i understand the calculation of average referencing, i
should NOT use the values of the reference to calculate the mean of all
electrodes, which would be afterwards subtracted from each one. And if i wanted
to see the signal of Cz post-AR, i would have it by inverting the mean of all
electrodes. If so, then the right way to go here would be to exclude also the
channel 30 in the reref function. Is that right?

I am using currently EEGLAB v13.4.4b on MATLAB R2013a

EG

> Subject: Re:[Eeglablist] Average Reference after measuring 128 Ch using one of
> the 128 channels as reference (Cz)?
>
> Dear Ernesto,
>  The  best choice is to use  the zero reference (REST) which has been
> confirmed much better than average reference.