[FieldTrip] Projection of simulated source activity to the surface EEG

[Schoffelen, J.M. (Jan Mathijs)]

Dear Nina,

> i) Afaik, none of the atlases provide normals calculated up front. I calculated the normals for each dipole (center of mass of an atlas region) as a normalized, averaged normal of all normals that appear in one atlas region. There were as many normals as were triangles in a region and one normal was calculated as a cross product between two edges of a triangle (of a surface) and oriented outwards. Is that approach ok or is there nevertheless some standard way of doing it? Papers usually don't go in such details.

Yes indeed, averaging the vertex-normals is based on the assumption that each and every dipole in a cortical patch has exactly the same activation (note I say vertex normals, rather than triangle normals -> one easy way of computing this is specified in fieldtrip/forward/private/surface_normals.m). One could of course come up with more fancy models of parcel-based activation, e.g. having some multivariate activation with a different weighting per vertex, but unless otherwise specified in publications, the safest guess would be that just plain averaging has taken place.

> ii) I am slightly confused which type of surface should I take for the calculation of normals. Most likely that should be the original folded surface, and not modified pial, white or inflated, as are also the options? When I was looking at the results, they clearly (and logically) differ.

Well, I think it would make sense to indeed use a surface with original folding (i.e. no inflation). Typically, people take the midthickness (the average betwen the white and pial surfaces if it does not exist yet).

> iii) could I still use volumetric atlas, where orientations of the cortical regions would be defined as commonly in the surface based atlas (perpendicular to the surface), even though it would be volumes instead of surfaces? In a sense that I take surface areas of the cortical volumes to calculate the normal, and then use that normal as orientation of the dipole, that is in a volume, not necessarily on a surface of a volume. Then, the subcortical regions could either be excluded in the projection (and only had influence through time series couplings in the simulation), or the orientation could be defined based on structural connectivity (e.g. sampling from probabilistic distribution determined based on structural connectivity)?  Thank you for any comment on that topic.

You lose me a bit here, but I would typically stick to the surface for cortical sources. If you want to include subcortical sources in your model (which don’t exist in a surface atlas), you could combine the cortical parcels with volumetrically defined subcortical ones. Yet, there it is tricky to define an orientation, in part because many of those patches don’t produce an open field which can be picked up outside the head (but see https://urldefense.com/v3/__https://www.sciencedirect.com/science/article/pii/S1959031809000414__;!!HJOPV4FYYWzcc1jazlU!4LVbJAqNJ55iTqWsvK5C3TBx7MJl2l9D4SoXKSGm7YSYsWcL5q6WzdEb6MPSf-T1x00PNgUpmwj6txxdvj2_VEuZh2HoDH0NU-f18w$  ).

Good luck and keep up the good work,
Jan-Mathijs
 

> 
> 
> I hope I'm not asking too much. Thank you again for your time!
> 
> Best,
> 
> Nina
> 
> 
> 
> _______________________________________________
> fieldtrip mailing list
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> https://urldefense.com/v3/__https://doi.org/10.1371/journal.pcbi.1002202__;!!HJOPV4FYYWzcc1jazlU!_AieN513ZaKvbpO3Muq_RIY4bdfKioc6zklTohTCh_2M41wvd7aupcDTtl7DKDm1DpMqBdhuAHXRPWlHXyZFIrw2mzKzTIRIqaE8HA$