Subcortical source simulation

💬 Support & Discussions
1
  • MNE version: 1.3.1
  • operating system: Windows 10

Hi MNE community,

I am trying to use mne source simulator to simulate hippocampal source data, and I am following this tutorial Generate simulated source data — MNE 1.4.2 documentation. The problem is that I can use only cortical surface labels from parcellation to add data to the simulation (source_simulator.add_data), while I should use aseg for the hippocampus. Do you know any way around it?

Thank you.

2

@larsoner this was discussed today in office hours, and both @agramfort and I were stumped. Do you have any ideas?

3

You can use a “discrete” source space where you explicitly specify the dipole positions and orientations with pos=dict(rr=..., nn=...) in mne.setup_volume_source_space

4

Sorry I should have unpacked our discussion a bit more. We got as far as “use a discrete source space” but then the issue is: if your hippocampus label is volumetric, you can get the positions easily enough but how do you choose the orientations? Some ideas we kicked around:

  • find the centroid and assign several orientations to a dipole there, and see what patterns show up as orientation varies
  • convert the volumetric label to an encapsulating surface (somehow?) and use the surface to define dipoles normal to it (not very physiologically plausible but might be OK depending on what you’re trying to learn from the simulation)
5

You can do this with marching_cubes, smooth as much as you want, and get the normals, etc.

For code to get part of the way there using FreeSurfer see also ENH: API for subcortical surface source spaces · Issue #6784 · mne-tools/mne-python · GitHub

2 Likes