creating BEM head models for M/EEG forward modeling MNE-style

[cjayb]

Documentation and best-practices for creating multi-layer BEMs for source localisation

• software (freesurfer, simnibs, brainstorm,...)
• running on cluster
• checking/visualisation

@MadsJensen @sarathykousik

Perhaps a line or two on the wiki about how SPM does this? @martinjdietz

[MadsJensen]

+1 I have some code for simnibs to cluster, and some bash wrappers for visualization in freeview

[sarathykousik]

Separate advice for EEG and MEG?
I have some submission scripts but Mads' is probably better.

[MadsJensen]

@sarathykousik we can compare and merge at the Hack. But as it is for the a multi-layer BEM there should be no difference between MEG & EEG.

[sarathykousik]

Code-wise yes. But pipeline-wise we should discuss it. But that's more a pipeline issue. 👍

[martinjdietz]

BEM forward modelling in SPM is shared with Fieldtrip and done using Robert Oostenveld's "forward" toolbox.

[sarangnemo]

FieldTrip supports several BEM variants: OpenMEEG (contributed by my group), dipoli, and bemcp. For MEG, a fast and for many 'good enough' alternative is their "Single Shell" model (originally devised by Guido Nolte), which uses a brain or scalp segmentation as a 'shell'. BTW it seems bemcp isn't used much (possibly crashy?), but my group is also using dipoli and we will soon compare OpenMEEG to dipoli...

[sarathykousik]

+1 for singleshell option in FT. All you need from FS is a cortical surface and co-reg from somewhere.

[MadsJensen]

If the aim is both EEG & MEG we should have instructions on a three layer BEM as this is needed for EEG and combined EEG/MEG.

@sarathykousik @sarangdalal Is it fairly straight forward to make a BEM in FT based on imported surfaces in FreeSurfer format? Simnibs (if given the right MRI data) creates good surfaces and from visual inspection they are a better fit than anything else I have seen so far.

[sarathykousik]

@MadsJensen: I haven't tried multi-layer BEM + FT on my data. Only tried out this tutorial. Once seg is done within FT (I think it uses some SPM functions), then you can use one of the variants Sarang mentioned.
Tutorial here. [http://www.fieldtriptoolbox.org/tutorial/headmodel_eeg_bem]

[sarangnemo]

@MadsJensen , as long as your meshes are 'BEM-compatible' -- i.e., the surfaces cannot intersect themselves, and the different layers cannot touch -- then they should be pretty straightforward to import and use with FT. Even if not, perhaps the same mesh-fixing routines that we use for SPM/Fieldtrip-generated meshes would also work.

[sarangnemo]

Important note: The "dipoli" BEM variant in FieldTrip is for EEG only