[Eeglablist] dipfit uses average reference? (fwd)

[Jonathan R Folstein]

Robert already answered me backchannel, here is his reply.

(Robert, I don't know you are on this listserv or not, so I'm including
you on this email.  If you would prefer to give a different answer or not
have your answer posted, go ahead and let the moderators know.  If this is
all redundant, I apologize for the spam...)


Jonathan Folstein, Doctoral Candidate
Cognitive Psychology
University of Arizona
email: folstein at u.arizona.edu
lab phone (520)-621-3265

---------- Forwarded message ----------
Date: Wed, 29 Nov 2006 21:30:22 +0100
From: Robert Oostenveld <r.oostenveld at fcdonders.ru.nl>
To: Jonathan R Folstein <folstein at U.Arizona.EDU>
Subject: Re: [Eeglablist] dipfit uses average reference?

Hi Jonathan

On 29 Nov 2006, at 21:11, Jonathan R Folstein wrote:
> This thread is extremely useful, thank you.  Would a non-cephalic
> reference at the base of the neck be a reasonable substitute for
> average
> reference?

Regarding dipole modelling: any non-cephalic electrode would be
difficult to model in a volume conduction model: the lower part of
the head is difficult to image with MRI (susceptibility artifacts,
also a problem for fMRI). Even if the geometry would be known (e.g.
from a mixture of a CT and MRI scans), it would be a very complex
geometry with also very difficult volume conduction properties. So I
would not trust the volume conduction model for a very low electrode,
hence I also would not use the measurement at a low electrode for
inverse source modelling. Using a poorly modelled electrode as the
reference makes the situation even worse (as explained in my previous
mail), but I would not consider the potential of that electrode at
all (also not referenced to a "good" electrode).

Note that those low eletrodes are still usefull for ICA
decomposition, since in ICA there is no assumption about a volume
conduction model. The low electrodes in neck and face are actually
helpfull in disentangling the components corresponding with muscle
and eye artifacts. This is actually the reason that in DIPFIT the
channel selection for dipole fitting can be a subset from the
channels used in ICA decomposition.

About "average" reference: that assumes that all electrodes are
known. The potential at Cz is different when average referenced using
a 19 channel 10-20 electrode cap, compared to when average referenced
to a 256 channel cap as Scott and colleagues are using at the SCCN.
The easiest way to think about the influence of the reference
electrode on your potentials is by assuming a dipole in the brain
stem (i.e. approx in the middle of the head). If the brainstem dipole
is pointing towards the vertex, the potential at the neck would be
lowest and at the vertex the highest. Using a base-of-the-neck
reference electrode, the potential at the lower part of the head
would be zero (per definition), at T7/T8 it would be 50% of the
maximum, and at the vertex it would be 100% of the maximum. So it
would be zero or positive everywhere: hence the average of the
potential at all channels is NOT zero. So to answer your question: a
base-of-the-neck reference electrode is different from an average
reference (for this specific dipole).

best regards,
Robert