[Mne_analysis] fiff data

[Eric Larson]

        External Email - Use Caution        

>
> 1) How can I modify or generate the sensors file (size, distance, ....) to
> fit it to our sensors?
>

For this take a look at coil_def.dat
<https://github.com/mne-tools/mne-python/blob/master/mne/data/coil_def.dat>,
but also more critically the forward solution documentation of coil def
<https://mne.tools/dev/overview/implementation.html#coil-geometry-information>.
I'd recommend reading the entire forward section, it seems necessary for
your work.
<https://github.com/mne-tools/mne-python/blob/master/mne/data/coil_def.dat>

Also check out the custom `coil_def.dat` that comes with the OPM example.
This, plus `info['chs'][ii]['loc']`, which produces a coil translation +
rotation matrix using `mne.io.tag._loc_to_coil_trans`, is used to orient
the sensors. Typically for SQUID MEG sensors these locations are stored in
the MEG device coordinate frame, but it might make sense for your use case
to figure out what they are (and store them) in the Neuromag head
coordinate frame. I think that this might be done in the example (?), but
note that their definition of the head coordinate frame is incorrect
(instead of being defined by LPA/Nasion/RPA it's set to be equal to the MRI
coordinate frame -- which is fine for the forward calculation, it just
means that you shouldn't interpret any results in the "head" coordinate
frame as being in the standard location/orientation). I think the forward
code should already recognize that they are in head coordinates, but in
case it does not, it would be safest to set `info['dev_head_t']` to an
identity transform.


> 2) We want to do simulation considering simple cases: radial dipole,
> tangential dipole + intermediate orientations, more or less deep
> dipole,...etc
> , is it possible to do the simulation in MNE python?
>

Yes, once you have the `modified_coil_def.dat`, set the `coil_type` and
`loc` in `info['chs']` properly, you can use any forward computing
functions you want from MNE, including the simulation code, as long as you
are in the mne.use_coil_def context manager (as in the OPM example).

3) Do you know what method is used to compute the forward model of the opm
> dataset you have?
>

It's the same as in other parts of MNE, we use the linear collocation
method (see for example Mosher et al 1999
<https://ieeexplore.ieee.org/document/748978>). But perhaps this is not
your question -- if you are asking if we compute the forward in all three
orientations for each source location, this is what make_forward_solution
will do. If you want just the normal direction of cortex you then use
convert_forward_solution(..., force_fixed=True).

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200407/c2abf2cf/attachment.html