[Eeglablist] Signal/noise loss, re-referencing Biosemi data

Thomas Ferree tom.ferree at radiology.ucsf.edu
Mon May 9 21:10:40 PDT 2005

By definition, voltage is always measured relative to a reference.
One can re-reference in software to any electrode, by subtracting
that electrode time series from all the others.  This is what any
software will (i.e., should) do when re-referencing.  But certainly
the Biosemi system is using some reference choice when acquiring.

I think you are confusing linked reference with average reference.

Collecting with a linked reference, e.g., linked mastoids, is a
mistake.  It installs a new, low-resistance current path from one
site to the other, that otherwise would not be there.  That forces
the potential at the two sites to be equal, when otherwise they
would not.  It fundamenatally alters the data you collect in a way
that can not be remedied by re-referencing off-line.  It violates
the no-current-flow-through-the-scalp boundary condition that is
implicit in every head model on the market, including EMSE.
However you can re-reference to the average of both mastoids
off-line if you want.

Now, average reference: That is an off-line signal processing
step.  As far as I know, no system has the capability to collect
data in an average reference mode.  Without getting into the
pro's and con's of average reference, let me say that EMSE has
the option to transform both data and head model foward solutions
to the average reference, for the purposes of visualization and
improved inverse solutions.  For the latter, although choice of
reference does not matter to a first approximation, the average
reference is a good idea after a rather subtle argument.

See Nunez (1981) for an authoritative discussion of reference
choices and average reference in EEG.  See also the discussion
of scalp surface Laplacian.

At 12:34 PM -0600 5/9/05, Jim Kroger wrote:
>Recently, the following was posted by Arno, and I'm looking for more 
>details. I've searched the archives thinking there was an earlier, 
>longer post on this, but can't find it. I have 2 questions about it 
>Finally, do not forget to re-reference your data (otherwise there is a
>40dB loss in signal/noise).
>1. Biosemi is collected and stored raw, as I understand it, and 
>referencing is done in software; when I use EMSE I select "common 
>average reference."  Is a similar operation what is meant by 
>re-referencing? I gather from vague recollection of the previous 
>post that Arno meant the data must be referenced with a single 
>electrode as the reference.....that is, the common average reference 
>scheme should NOT be used, or the S/N loss will occur. Is this 
>correct? If so, since we did not use a reference electrode for some 
>collected data, which electrode is recommended as the reference 
>electrode? We are looking at visual attention and working memory, so 
>occipital, frontal, and parietal are important.
>2. Then, should we be collecting data with a reference electrode? 
>Any reason to do with nose, earlobes, mastoids?  Linked?
>eeglablist mailing list eeglablist at sccn.ucsd.edu
>Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu


Thomas Ferree, Ph.D.
(415) 353-9474

More information about the eeglablist mailing list