As I begin working with simultaneous MEG-EEG data, I’m trying to better understand how MNE handles BEM models & conductivities.
From what I understand, MNE always requires a BEM model to compute a forward solution (mne.make_forward_solution). However, if only MEG channels are present, is the BEM model data actually used ?
If yes, does it means that MNE consider different magnetic permeability values inside vs. outside the outer skull surface ?
If no, Would it be possible to skip BEM surface computations entirely for MEG-only recordings? Also, it is correct to assume that the conductivity parameters passed to make_bem_model are only relevant for EEG ?
The BEM model data is actually used, also by MEG and yes that means it considers different permeability values inside vs. outside the outer skull surface if you setup the BEM model like that.
To ignore BEM surface computations, you can use a spherical model: mne.make_sphere_model — MNE 1.10.1 documentation That by default gives you a 3-layer spherical BEM model with different permeability values for the brain, CSF, skull, and scalp. You can also set head=None to obtain a model without any layers, so ignoring permeability completely.
Maybe a slight addition, but at the frequencies of MEG, the relative magnetic permeability values of the brain tissues can all be seen as 1. The different electrical conductivities do affect the MEG signal a bit, since they affect the secondary currents. Their influence is of course less than in EEG
Following your answers, I still have a couple more questions regarding conductivities and permeabilities. Since MNE uses different permeability values, how are these defined? Are they constant values stored somewhere in the code ? Is there a reason why one can adjust the conductivities but not the permeabilities ?
For MEG-only recordings, the permeability values are used in the magnetic field propagation and the conductivities (specified when creating the BEM model ) are used for the secondary currents. Is that interpretation correct ?
So the magnetic permeability is usually not used for these kinds of models. They are a measure of magnetization when a magnetic field is applied. Since the human tissues are not really magnetic, they do not really play a role.
The conductivity values do play a vital role, though, and they can be implemented in different ways. In the BEM models, these values are specified per tissue type (brain, skull, scalp). In FEM models, they can be assigned to each voxel and even made anisotropic (direction dependent).
In almost all research, the conductivity values are assumed to be a constant value across the population (which is definitely incorrect, but estimating individual conductivities is very difficult).
For MEG, the primary magnetic field created by the dipoles in the brain is not really dependent on any parameter. The vectorial position from the dipole to the sensors in that case defines how the dipole translates to the MEG. However, since the neuronal currents also flow and thus create a magnetic field as well (since a current always carries a magnetic field), the conductivities influencing the currents do play a role as well.