[Eeglablist] re-refeference? /correcting for the polar average reference effect

Sun Jul 5 11:01:41 PDT 2009

Zara,
    It's not clear to me that "an inverse relationship between the  
(vertex) reference and all (64) electrodes, getting larger as they get  
further away from the original common average reference." necessarily  
indicates a PARE problem, although I agree with Tom that it might be a  
contributing factor.  Since all electrical activity produces dipolar  
fields in which one direction is relatively positive and the opposite  
direction is relatively negative, measurements from the two sides of  
the dipolar activity will always be negatively correlated.  For  
example, if the data contains vertically oriented EEG activity then  
one end of the dipoles will be recorded at the vertex and the other  
end of the dipole will be recorded at the bottom sites and they will  
be negatively correlated. It is the tendency of mastoid referencing to  
make vertical voltage fields appear to be unipolar that is  
biophysically unrealistic and it is the average reference which is, in  
this sort of case, depicting a more realistic pattern of data.  That  
said, the electrodes near the vertex reference should be at least  
modestly positively correlated with the vertex.  The average reference  
essentially divides the electrodes in half (more or less, as the  
division is based on the total voltage recorded rather than by  
physical electrodes), allocating half to be the one side of the dipole  
and the other half the other side of the dipole.  The PARE effect  
results in the dividing line to be mislocated (e.g., in the case of  
the vertically oriented dipole, closer to the top than it should be).   
It should never be the case that ALL the other electrodes (in a high- 
density montage) are inversely correlated with the vertex, even with a  
PARE problem.  It sounds to me like either the rereferencing procedure  
was problematic or the manner in which the electrode correlations was  
calculated is problematic.  Also, one point of curiosity - I thought  
that the EGI system was the only one that used Cz as the original  
reference site.  Is this no longer the case?  Or are you actually  
talking about a ground electrode rather than a reference electrode?

Just my two bits.

Cheers!

Joe

On Jul 2, 2009, at 11:13 AM, Thomas Ferree wrote:

> Dear Zara,
>
> The following paper states exactly how to do this, and demonstrates
> numerically the advantages of this way of estimating average  
> reference.
> Furthermore, in my own studies I have seen that it improves the
> accuracy of topographic maps, which was your original question.
>
> There is not a new channel per se, so your question on that point need
> not be addressed.
>
> I have Matlab code for doing this that I might be willing to share.  
> It is
> set up to work with EEGLAB data structures, but I have not gotten  
> around
> to making a plug-in for it.  Please contact me in person about that.
>
> Best regards,
> Tom.
>
> Spherical splines and average referencing in scalp  
> electroencephalography.
>
> Ferree TC.
> Brain Topogr. 2006 Winter;19(1-2):43-52.
>
> 2009/7/2 Zara Bergström <zara.bergstroem at med.ovgu.de>
> Hi Andrew and list members,
>
> what you are describing sounds like a bias in the average reference
> transform known as the polar average reference effect (PARE, Junghöfer
> et al.,1999, see abstract below). Junghöfer et al recommend a  
> particular
> method for correcting this bias, described in their paper. I also  
> have a
> very large PARE in 64 channel data, and would like to use their method
> for correcting it, but I'm a little bit confused about how to  
> implement
> that correction in practice.
> So, I was hoping that someone on this list might be able to please
> explain it?
>
> In their paper, Junghöfer et al. describe how one can correct for the
> PARE by using the scalp electrodes to estimate the potential on the
> surface not covered by electrodes using a spherical spline
> interpolation, and then computing the average reference for the entire
> surface using the results from the interpolation.
>
> How exactly can I implement this in practice? I'm guessing that it  
> would
> involve creating a new "channel" representing the non-sampled area,  
> and
> then including that channel in the average reference (so, PARE- 
> corrected
> average reference = average of (all scalp channels + interpolated new
> channel)). But how can I compute an interpolated channel without
> inputting the coordinates of the channel to be computed? And how do I
> update the EEGLAB data structure to take into account this new channel
> when re-referencing? Will the re-referencing work without channel
> coordinates for the new channel?
>
> Or even better, does anyone know of an already existing EEGLAB  
> function
> or plugin (or a MATLAB script) that implements a PARE correction when
> re-referencing to an average reference?
>
> Any comments will be much appreciated!
>
> Thank you very much in advance,
>
> Zara
>
>
> Abstract from Junghöfer et al., The polar average reference effect: a
> bias in estimating the head surface integral in EEG recording. Clin
> Neurophysiol. 1999 110(6):1149-55
>
> "A reference-independent measure of potential is helpful for studying
> the multichannel EEG. The potentials integrated over the surface of
> the body is a constant, i.e. inactive across time, regardless of the
> activity and distribution of brain electric sources. Therefore, the  
> average
> reference, the mean of all recording channels at each time point,  
> may be
> used to approximate an inactive reference. However, this
> approximation is valid only with accurate spatial sampling of the  
> scalp
> fields. Accurate sampling requires a sufficient electrode density
> and full coverage of the head's surface. If electrodes are  
> concentrated
> in one region of the surface, such as just on the scalp, then the  
> average is
> biased toward that region. Differences from the average will then be
> smaller in the center of the region, e.g. the vertex, than at the  
> periphery.
> In this paper, we illustrate how this polar average reference effect
> (PARE) may be created by both the inadequate density and the uneven
> distribution of EEG electrodes. The greater the coverage of the  
> surface
> of the volume conductor, the more the average reference approaches
> the ideal inactive reference."
>
>
>
> [Eeglablist] re-refeference?
> Andrew Hill andrewhill at ucla.edu
> Tue Jun 16 19:01:59 PDT 2009
>
>    * Previous message: [Eeglablist] Research Assistantship/Fellowship
> position available at NIH/WRAIR in Bethesda MD
>    * Next message: [Eeglablist] 5th BCI2000 Workshop and Workshop on
> Electrocorticography, Oct. 1-3, 2009
>    * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
>
> hello folks,
>
> i've got a question on data referencing on some .CNT files.
>
> it seems that using reref to avgref state still shows an inverse
> relationship between the (vertex) reference and all (64) electrodes,
> getting larger as they get further away from the original common
> average reference.
>
> this data was from a 64-channel recording on a Synamps2 (32-bit).
> creating epochs/averages from this data either with linked ears re-ref
> or average reference, i still "see" the influence of the reference
> near Cz.
>
> is there any way to correct for this when performing a compute average
> reference in EEGLab?
>
> thanks,
> andrew
>
>
>
>
> --
> Dr. Zara Bergström
>
> Postdoctoral research fellow
> Memory and consciousness research group
> Department of Neurology
> Faculty of Medicine
> Otto von Guericke University Magdeburg
> Leipziger Str. 44
> 39120 Magdeburg
> Germany
>
> Tel. +49 (0)391 6117544
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu
>
>
>
> -- 
> Thomas Ferree, PhD
> Department of Radiology
> UT Southwestern Medical Center
> Email: tom.ferree at gmail.com
> Voice: (214) 648-9767
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu

--------------------------------------------------------------------------------

Joseph Dien
Assistant Professor of Psychology
Department of Psychology
419 Fraser Hall (by the coke machine)
1415 Jayhawk Blvd
University of Kansas
Lawrence, KS 66045-7556
E-mail: jdien07 at mac.com
Office: 785-864-9822 (note: no voicemail)
Fax: 785-864-5696
http://people.ku.edu/~jdien/Dien.html