Projecting to default anatomy, with deep brain mixed model

Hi everybody,

I have individual scalp EEGs and individual subject-specific T1 MRIs. I am aiming to do source reconstruction with BEM head model, and compute connectivities between the following scout groups: all scouts in Desikan-Killiany surface atlas, hippocampus DBA constrained surface, and amygdala/thalamus DBA unconstrained volumes. I'd like to check my rough pipeline for this purpose, past you guys, as there are some warnings, and this is uncharted territory in the tutorials. TutorialDBA does not seem to cover projecting sources from individual to default mixed cortex+subcortical surfaces yet, nor computing connectivities between surface and volume scouts, so hopefully this discussion will help some future users also.

Steps:

  • New protocol, import individual anatomy, compute MNI coordinates, generate BEM surfaces for individual anatomy
  • Like in TutorialDBA, make cortex_mixed18000V surface, but for BOTH the default anatomy and individual anatomy separately.
  • Like in TutorialDBA, open this surface, make a new atlas with source model options, and set to DeepBrain mode.
  • Functional data, review raw file, Add electrode positions, EEG preprocessing, import in database (epoched), Compute head model (custom source model) OpenMEEG BEM
  • Compute sources 2018, with MNE
  • Project sources to default anatomy cortex_mixed surface
    This works but produces a warning ‘BST> Warning: Normals could not be calculated for 76 vertices. Rightclick on the surface and select "Clean surface" to remove these vertices.’
  • Now, having L clicked one of the group analysis source files or the default anatomy cortex_mixed surface, make a new volume atlas with amygdala+thalamus, and a new surface atlas with Desikan-Killiany + hippocampus surfaces.
  • Alternatively to above step, I think it's equivalent to make new volume/surface atlases from subject-specific source files, then in Scout menu, project the atlas to cortex_mixed.
  • Then, extract the scout time series from group analysis source files for the surface atlas and volume atlas separately, and compute connectivity outside of Brainstorm (or combine the surface and volume scout time series somehow).

Questions:
Why does project sources produce that warning message?
Is projecting to default cortex_mixed surface supported by Brainstorm, and a valid procedure?
Incidentally, projecting within Process 1 does not seem to have the option of projecting to cortex_mixed.

Thank you

The projection of the sources to a template are available in two configurations:

  • using FreeSurfer/BrainSuite cortex surfaces for which you have a correspondence estimated for each vertex between the subject and the template.
  • using volume atlases: in this case, we simply use the SPM affine normalization function to MNI space.
    Projecting between structures that are not correctly registered (ie. all your subcortical regions) is not correctly supported.
    https://neuroimage.usc.edu/brainstorm/Tutorials/CoregisterSubjects

Note that you probably don’t need to project your sources here. I think you should compute the connectivity metrics for each subject separately, and then compare the results across subjects. This doesn’t involve any spatial projection, which is needed only when you want to average full source maps across subjects.

Computing connectivity metrics across signals of different types is not recommended. Source signals from surface models with constrained orientations have only one component oscillating around zero. Source signals from unconstrained/volume models have 3 spatial components, and for estimating connectivity for these models we don’t have any clear solution available in Brainstorm.

@leahy @Sylvain @hossein27en: What is the current status of connectivity metrics on unconstrained source models?

Thank you for that, Francois.

To clarify, I can make subject-specific volume atlases (e.g. for amygdala with DeepBrain options), by growing scouts. I can also make subject-specific surface atlases (e.g. hippocampal surfaces with DeepBrain options) in the same way. However, any subcortical atlas I make cannot be successfully be projected across different subjects or from a common MNI space to all the subjects. Is that right?

Will be very interested to find out if there are any updates on connectivity metrics on unconstrained source models.

However, any subcortical atlas I make cannot be successfully be projected across different subjects or from a common MNI space to all the subjects. Is that right?

Using mixed head models, no.
But with a volume head model, you can:

  1. Follow these instructions to create a grid in your MNI template and use it for all the subjects: https://neuroimage.usc.edu/brainstorm/Tutorials/CoregisterSubjects#Volume_source_models
  2. Create a volume scout in any of the subject, or in any of the projected sources (in a subject, right-click on the source results > Project source grid): https://neuroimage.usc.edu/brainstorm/Tutorials/TutVolSource#Volume_scouts
  3. This scout can be readily used for all the other subjects, as they all use the same source grid. But the menus are missing to project the scout to other subjects. You can however go around this limitation by copying it manually from one subject to another. Open the source results for one subject, select your scout, menu Scout > Export to Matlab > “s”, close the figure, open results for another subject, create an empty volume atlas, menu Scout > Import from Matlab > “s”