Incorrect surfaces from mne.bem.make_watershed_bem()

💬 Support & Discussions
1

Dear MNE experts,

I am new to MNE and I would have a question. I need to use the mne.bem.make_watershed_bem() command to create the EEG forward model (the output will be used in the script to create TVB input rows 271-316 and 426-459 https://github.com/BrainModes/tvb-pipeline-converter/blob/master/convert2TVB_format.py), however I got the error “the surface inner skull is not completely inside surface outer skull”.
I improved the surface extraction as much as possible avoiding manual interventions (only scripting), as described in https://mne.tools/stable/overview/faq.html#forward-and-inverse-solution
using Freesurfer7.3.2 .
So I reran -autorecon1 with the -expert option adding in the file xopts.txt:
mri_normalize -mprage -b 20 -n 5 -gentle \n mri_watershed -h 5 -atlas -useSRAS -surf sub-01_recon

This improved a lot the surfaces, however the inner_skull and outer_skull surfaces, which are created respectively shifting the brain surface and the outer skin surfaces, are still crossing (I suspect this is due to the fact that I am using pediatric data and the shift of the surfaces are too big maybe?).

So I created a new inner_skull and outer_skull surfaces giving a shift of +1mm and -2mm respectively using Freesurfer commands:

  • mris_expand sub-01_recon_brain_surface 1 sub-01_recon_inner_skull_surface # expand the surface of brain created by “mri_watershed” command of +1mm radially and overwrite the existing one
  • mris_expand sub-01_recon_outer_skin_surface -2 sub-01_recon_outer_skull_surface # expand the surface of skin created by “mri_watershed” command of -2mm radially and overwrite the existing one
    They are now much better, see attached screenshots (before shift correction and after it).

Surfaces_without_shift_correction
Surfaces_without_shift_correction1730×767 192 KB

Surfaces_with_shift_correction
Surfaces_with_shift_correction1730×767 194 KB

However if I use now the MNE command:
mne.bem.make_watershed_bem(subject= recon_all_name, subjects_dir = recon_all_dir, overwrite=True, atlas=True, gcaatlas=True, preflood=5) this recreates the mri_watershed command of Freesurfer creating the inner_skull and outer_skull surfaces as they were before creating those with a reduced shift (+1mm inner_skull and -2mm outer skull, see above) and hence with the crossing.

I see that the MNE command mne.bem.make_watershed_bem() creates in the “bem/watershed” folder the four files for surfaces and brainmask ws.mgz (which are already created by Freesurfer), plus in the folder /bem it creates other four surfaces with .surf extension (that I suspect are the same of those in the folder watershed/bem only renamed with extension .surf ) and a -head.fif file (which is brand new from the mne.bem.make_watershed_bem command), see attached screenshot.
Files

So these are my questions:

  • how can I create the -head.fif file and the .surf files using the surfaces created with Freesurfer (those with a corrected value of the shift)?
  • is the approach I used correct in your opinion?

I greatly appreciate your help!

Mic

  • MNE Python version: 1.4.2
  • operating system: macOS 12.6.8
2

Just another question: the brain_surface (and hence inner_skull_surface) still contains three pieces of skull: how can I remove them manually?

3

See Fixing BEM and head surfaces — MNE 1.5.0 documentation

Hope this helps

Alex

4

Thanks @agramfort , I fixed it following the guide.

M