# Current of source

Robert Oostenveld r.oostenveld at FCDONDERS.RU.NL
Thu Sep 29 09:31:41 CEST 2005

```Hi Mikhail

On 19-sep-2005, at 17:38, Mikhail Zvyagintsev wrote:
> as I found dipolefitting.m calculates just dipole moment. Do you
> have any
> routines to transform dipole moment to the electrical current? I
> looked
> through the FieldTrip, but I did not find it.

Current in the brain is not like current through a wire, therefore
you cannot express it in Ampere alone. You can either express it as
current density (i.e. current per unit of volume) if you compute a
distributed source reconstruction or as dipole moment for a sparse
source reconstruction. A current dipole consists of a source at the
dendrite and a sink at the soma (or the other way around, depending
on whether it is an exitatory or inhibitory synaps). The current
source and sink together form the dipole. The distance between source
and sink determine the impact that the current generator has on the
secondary currents in the surrounding tissue. E.g. if the distance
were zero, there would be no impact at all. Therefore the current is
normalised for the distance between the two, i.e. source strenght is
expressed in Ampere times meter. If you make the distance twice as
small, and the current flowing from the source to the sink twice as
large, the impact on the secondary currents in the surrounding tissue
remains the same, just as the dipole moment.

> Accordingly to my previous question (I mean how to deal with
> dipolefitting.m
> and dataset from example) - it works now, but I would still wait
> for your
> recomendation how to use it properly in this special case.
> Another question is - how would you recommend to estimate activity of
> auditoria dipoles? Does nonlinear algorithm make any difference if
> I would
> do it separately - one after another one or it is always better to
> estimate
> them both in the same time? Some time they are not in exactly
> simmetrical
> positions... And if I estimate them one after another one, how
> should I
> project out the contribution of the first source from the measured
> signal?

In EEG (AEPs), you are required to estimate them simultaneously,
since they both project to a vertex extremum. In MEG, you can try to
estimate them separate, by taking subselections of sensors over both
hemispheres. Even then, about 10% of the field of the left auditory
dipole is still visible over the right hemisphere (if I recall
correctly). That means that in MEG you better also estimate them
simultaneously. You can start the estimation with a symemtry
constraint, and then use the positions of the symmetric dipole pair
as the initial guess in an unconstrained optimalisation.

If you have a single dipole model for your data, you can subtract the
model field distribution from the observed, and then continue with
the residual field. However, that is not optimal, since the field of
the first dipole is not orthogonal to the second dipole, and hence
you are subtracting part of the activity of the second dipole as well.

best regards,
Robert

=======================================================
Robert Oostenveld, PhD
F.C. Donders Centre for Cognitive Neuroimaging