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

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

Dear Nina,

Well, it sounds as if you’re almost there! Indeed, as you indicate yourself, dipole orientation plays a role here. The ’times 3’ in the output to ft_compute_leadfield results from the fact that a forward model has been computed for 3 dipoles at each of the locations (i.e. a unit amplitude dipole pointing into the x/y/z/ direction respectively). If your generative model for your data is based on the idea that a single point dipole with fixed orientation per region describes the sources per regions well enough, then you’d need to decide on an optimal orientation for each dipole, express this as a unit norm vector, and use these dipole moments to reduce your 128 x 270 matrix to a 128 x 90 matrix. This could be either done per region (i.e. take each 128 x 3 ‘leadfield’, and multiply that with the each regions 3x1 dipole moment), or in a single matrix multiplication, where you make a large block diagonal matrix of all dipole moments (which would result in a [128x270] * [270x90] matrix multiplication, where the latter matrix is a (in matlab speak) blkdiag of all [3x1] moment vectors). Indeed, once you ‘lf’ is a 128 x 90 you can create your (average referenced) EEG timeseries just by means of a matrix multiplication (plus the optional addition of electrode noise).

I guess the challenge now is to come up with a meaningfully defined orientation for each of the regions. Given that the atlas is a volumetric one, it’s not straightforward to apply a heuristic that is based on the dominant orientation of the cortical surface in the parcel (which is the first thing I’d consider). Is there a possibility that you could use a surface based atlas, which would allow for such a heuristic?

Best wishes, and keep up the good work,
Jan-Mathijs

On 15 May 2023, at 14:01, Nina Omejc via fieldtrip <fieldtrip at science.ru.nl<mailto:fieldtrip at science.ru.nl>> wrote:

Dear Fieldtrip community,

I am a second year PhD student from Slovenia, looking for some help. As the title suggests, I am trying to project simulated source activity to the surface (EEG data simulation from sources).

Up to now, I have simulated 90 timeseries at the source level using Jansen-Rit neuronal mass models, based on the AAL volumetric atlas (activity is placed in each center of the 90 regions), so I have 90x3000 matrix (sources x time points). I have decided to project the source activity to 128 channel standard Biosemi 128 EEG setup. Now I have issues with defining forward model from the AAL sources to the EEG channels.


Currently, I have followed your online examples to create headmodel (BEM) and the forward model, with the important part of the code being:

headmodel = ft_read_headmodel('standard_bem.mat');
[headmodel, sens] = ft_prepare_vol_sens(headmodel, sens);
[lf] = ft_compute_leadfield(atlas_rois, sens, headmodel); where atlas_rois is a 90x3 double, with 3d positions of source dipoles (at AAL centers)


What I don't understand is the output of the function ft_compute_leadfield, which is a 128x270 double. Could you please explain what it represents? As far as I understand, the surface EEG should be obtained by the equation surface_time_series = forward_model * source_time_series, meaning that the dimensions should have been [number of electrodes) x (number of time points)] = [number of electrodes) x (number of sources)] * [(number of sources) x (number of time points], which comes down to: [128x3000] = [128x90] * [90x3000]....   Is there a reason 3*90=270? Do source (dipole) orientations play a role?

Additionally for clarification, to obtain the final EEG projection, would it be then correct to directly apply the equation surface_time_series = forward_model * source_time_series, with forward model being the variable "lf"?

I have tried various toolboxes for this simulated sources to EEG projection (TVB, MNE, SPM, Fieldtrip) and had issues everywhere, but came the furthest with Fieldtrip, so I would really appreciate your help. General recommendations on this topic would also be greatly appreciated.

Thank you very much!

Kind regards,

Nina Omejc


_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://urldefense.com/v3/__https://doi.org/10.1371/journal.pcbi.1002202__;!!HJOPV4FYYWzcc1jazlU!-pwnzebYdwU0MsbC5LpPVLirsl_jIXoSfVuy5cESHsSs-IyQ-8aBFQLd5LUgYaj2YpIlh09uai4LCOug7yd2P1Al0Gp3qW1QVQipXQ$

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20230515/e2bc8c1e/attachment.htm>