[Eeglablist] Dipole-Fitting - Behaviour of fitTwoDipoles

Nils Hachmeister nils.hachmeister at uni-bielefeld.de
Mon Jun 12 00:45:48 PDT 2017

Hello everyone,

I'm a neuro-researcher from Bielefeld University, Germany. My question 
is about the dipole fitting: As I understand, pop_multifit only ever 
fits one dipole per IC (never two) and then the fitTwoDipoles adds a 
second dipole to some ICs when some heuristic indicates a second dipole 
would be beneficial, right?

When scanning through the dipoles in my data sets after running these 
two function using the fine-fit EEGLAB GUI function, I realised that for 
the vast majority of components with two dipoles the RV could be 
decreased significantly by fitting position and moment (again). While 
for single dipoles the RV never changed. Investigating this issue 
further I realised that if I only fit dipoles' moment of a IC with two 
dipoles nothing is changes. Hence, after adding the second dipole 
fitTwoDipoles seems only to fit the dipoles' moment.

Furthermore I realised that the second dipole's position always seems to 
be the mirror position of the first dipole, i.e. the sign of the first 
coordinate of the original dipole's position is flipped. The first 
coordinate describes the position in front-back-direction (along the 
saggital axis, the dipole would, thus, be mirrored at the coronal 
plane). It would have been my expectation that the dipole is mirrored in 
left-right-direction (along the coronal axis, mirroring the dipole at 
the saggital plane and thereby following the somewhat symmetric layout 
of the brain). The symmetric dipoles check-box in the fine-fit-GUI seems 
to do just that. Why does fitTwoDipoles behave differently?

This might also explain, why in very rare cases I observed that fitting 
the (already added) second dipole increased the RV. As I understood it, 
fitting dipoles really is an optimisation process aiming to minimise the 
RV. Hence, hitting the "Fit dipole(s)' position and moment"-button 
should never increase the RV. But if the dipole position to start with 
is not considered a potential solution by the fine fit, this would 
explain why.

Thanks in advance

Best Regards

More information about the eeglablist mailing list