LCMV source extraction for continuous resting data

External Email - Use Caution

Dear all,
I finally have some time to look into MNE as an alternative to spm/fieldtrip and Vladimir Litvaks amazing DAiSS toolbox. The way I used lcmv beamformer with DAiSS was to create surface projected raw data using the online montage function. https://github.com/spm/DAiSS
This allowed to create a resting dataset with e.g. 5000 virtual channels (as LFP/iEEG), from which I can analyse the spontaneous dynamics of oscillatory events in source space.

I have managed to get my resting data epoched and could project the epochs using mne.beamformer.make_lcmv and apply_lcmv_epochs.

  1. How would I do it for resting data?
  2. Is there a way to create a new raw data object with the new channels as chantype iEEG. This would make further processing much easier than with the source object.
  3. Finally, are the vertex coordinates stored somewhere in the source object?
  4. Is there a template solution for MNI2009b instead of fsaverage?

Thanks for your help
Julian

Dr. med. Wolf-Julian Neumann
Associate Researcher - Interventional Cognitive and Neuromodulation
Movement Disorder and Neuromodulation Unit
Department of Neurology
Charit? - Universit?tsmedizin Berlin
T: +49 30 450 660 359
M: +49 152 539 262 90

https://orcid.org/0000-0002-6758-9708
https://osf.io/uyg8w/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20191212/c8ce643f/attachment.html

External Email - Use Caution

How would I do it for resting data?

You can use apply_lcmv_raw
<https://mne.tools/dev/generated/mne.beamformer.apply_lcmv_raw.html&gt;\.

<https://mne.tools/stable/generated/mne.beamformer.apply_lcmv_raw.html&gt;

Is there a way to create a new raw data object with the new channels as
chantype iEEG. This would make further processing much easier than with the
source object.

You could hack the raw object to do this by creating a new RawArray
<https://mne.tools/dev/generated/mne.io.RawArray.html&gt; from the stc.data
you obtained from apply_lcmv_raw.

Finally, are the vertex coordinates stored somewhere in the source object?

Not in the object itself but in the source space used to create the
forward/LCMV filters. It depends on what type of source space you are using
but in general stc.vertices stores the vertex numbers used for the source
space, so src[0]['rr'][stc.vertices[0]] for example would give you the left
hemisphere source locations if you are using surface source spaces.

Is there a template solution for MNI2009b instead of fsaverage?

Not to my knowledge, but I think that fsaverage is already in MNI305 space
<https://surfer.nmr.mgh.harvard.edu/fswiki/CoordinateSystems&gt; so perhaps
that already gets you close to what you need.

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20191212/7c417798/attachment.html

External Email - Use Caution

hi,

I have managed to get my resting data epoched and could project the epochs using mne.beamformer.make_lcmv and apply_lcmv_epochs.

How would I do it for resting data?

I would split my raw in fixed length epochs and do as you did.

you can also use
https://mne.tools/stable/generated/mne.beamformer.apply_lcmv_raw.html
but then you keep the bad data segments

Is there a way to create a new raw data object with the new channels as chantype iEEG. This would make further processing much easier than with the source object.

what further processing do you have in mind?

Finally, are the vertex coordinates stored somewhere in the source object?

https://mne.tools/dev/auto_tutorials/source-modeling/plot_object_source_estimate.html

ALex

External Email - Use Caution

Great, thanks for the answer.

< what further processing do you have in mind?

I am looking at burst evoked connectivity.

Finally, are the vertex coordinates stored somewhere in the source object?
Page Redirection

So if I understand correct, the vertex coordinates are not stored in the source object. I like to use my own plotting and find the freesurfer images really ugly compared to what you can do with opaque colored source meshes (e.g. have a look at a recent poster: POSTIGO_2019.pdf)

Finally, has anyone created a MNI 2009b template? That would be a really great addition. Some research (deep brain stimulation) is entirely performed in volume space where the MNI2009b is simply the best template because of the high voxel resolution required for subcortical nuclei. Therefore, e.g. for LFP-MEG connectivity it is important to work in the same space between DBS electrode location and cortical projection.

Best wishes
Julian