Hi,
I've been looking at the forward computations generated by MNE. I've
been generating my fif files from data I have in matlab. I have an
average reference montage, and when I look at the forward model there
are large variations in the average across channels. Since it is an
average reference montage I expected the forward model to have zero
mean. I'm not sure if I don't understand the forward model output from
MNE, or if I'm not importing my data correctly.
Thanks,
Justin Ales
I think that the MNE forward matrix is just the distribution of
potentials on your electrodes produced by a current dipole at a given
location times all such locations. There is no reason for the sum of
potentials over the electrodes to be always 0. For example, if the
electrodes cover mostly the area of the scalp to where one pole of the
dipole points, then the averaged potential will be quite different
from zero. You could add a constant potential to turn the average to
zero, but this would only work for one dipole out of many.
I'm confused. I don't understand the hypothetical potential
distribution in your example. I thought the forward operator should
take into account the reference of the electrode measurement montage.
The sum of the potentials over the electrodes in the recorded data
would be 0 because of the explicit average referencing.
Justin
The forward solution itself is never computed using the average-
reference montage. Rather the average EEG reference is taken into
account when the inverse operator is computed. To get the average-
reference montage in matlab, you need to do it yourself.
- Matti
As Matti said, the average reference only matters when you compute the
inverse. The forward solution is computed for each hypothetical dipole
assuming, for example, that the dipole potential at infinity is 0
(which is the usual choice of the potential reference in
electrostatics). The forward solution has nothing to do with the
actual measurement, except for the positions of the electrodes on the
scalp.