[Eeglablist] When using DIPFIT, what are the units of the calculated dipole moments?

[Makoto Miyakoshi]

Hi Peter,

Finally, I see someone ask that question on this list. I am very happy to
find your post.

I recommend you read this section and check out my forward model
simulation. Note the unit for the 'Est dip moment': nAm. So the short
answer is nAm
https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Does_a_broad_dipole_layer_produce_a_depth_bias_when_fitted_with_a_single_dipole.3F_.28For_190.2C000_page_views.2C_02.2F22.2F2022_added.2C_07.2F18.2F2022_updated.29

The question is why those values are larger than usually expected (50-100
nAm) by three orders of magnitude.
You can find a link to my poster for Neuroscience 2019. My
explanation there must be pointing at least one of the explanations.

The fundamental mechanism I think is that ICA-derived scalp topos are not
the same as those of sensory evoked potentials after averaging 100 times
(which is the example for observing a typical dipole fitted to S1 with a
100 nAm dipole moment, according to an expert). I found that ICA-derived
scalp topos can represent a current source (in the sense of ICA i.e.,
temporally maximally independent sources, which may or may not be
identified as a biophysical current source) that has a very broad
distribution. In the case of current source in S1, the ground truth of the
current-generating source is known to be small, so the dipole's assumption
holds. But this assumption is severely violated for the case of cortical
current sources larger than > 20 cm^2 (this estimate is based on Nunez and
Srinivasan, 2006). In my simulation, you can see that a dipole layer with
20 cm^2 will have about x1500 larger dipole moments compared with the one
of a single dipole source. This is about four orders of magnitude, which is
close to our estimate (10000 / from 50 to 100), isn't it?

I have been working with one of the subscribers of this mailing list to
publish a study on this basic dipole problem. Do you want to join us? I
have published several studies on EEGLAB mailing list-derived problems.

Makoto

On Tue, Feb 4, 2025 at 11:40 AM Peter Elovsson via eeglablist <
eeglablist at sccn.ucsd.edu> wrote:

> Dear all,
>
> I am trying to use DIPFIT in EEGLAB at the moment. However, when I
> perform dipole source localisation, all of my dipoles have moment
> magnitudes in the order of 10000s, and there is no corresponding unit.
>
> I would have assumed the unit for moment is nAm, however this wouldn't
> make much sense considering how large my calculated moment magnitudes
> are.
>
> It could be that my co-registration isn't accurate enough which is
> causing such large and unreasonable moment magnitude (if the unit is
> nAm), but I am unable to reduce the mean distance between template
> electrodes and measured electrodes from the study I'm using to less
> than 5mm.
>
> Can anybody help clarify, is the co-registration the problem, or is it
> that the dipole moment magnitude calculated by DIPFIT is a strange
> unexpected unit and that's what's causing my problem?
>
> Thanks so much,
> Peter
> _______________________________________________
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu or visit
> https://sccn.ucsd.edu/mailman/listinfo/eeglablist.
>