Unable to use children MRI templates for source reconstruction (BEM surfaces)

  • MNE version: 1.7.0
  • operating system: Windows 10 or Ubuntu 22

Dear MNE-users,

We are trying to do source reconstruction on EEG data recorded in children participants (~8yo). We do not have access to the participants’ own anatomical MRI so we plan to use templates. Unfortunately, no children template datasets are available through MNE (unlike fsaverage for adults, or infants templates until 2yo).

We found the following MRI datasets: NITRC: Neurodevelopmental MRI Database: Tool/Resource Filelist, from the same database used for the infants datasets offered through MNE. They include freesurfer reconstructions but without the “bem” subfolder, which is crucial to compute the head and forward models. They include NIFTI files with masks of the different head compartments but this format are not usable as is.

I am currently failing to populate this freesurfer bem subfolder, specifically with the bem surfaces in the .surf format. I manage to produce the source space (src) file or the head surfaces.

I attempted the following methods:

  • Using the mne.bem.make_watershed_bem function and toying with the atlas and preflood arguments: on all age groups, it produced intersecting surfaces which forbids to go further with the BEM solution.

  • mne.bem.make_flash_bem did not seem like an option with these datasets

  • I tried through Brainstorm to use the FieldTrip method to produce the BEM surfaces from the T1.mgz file: it also failed to produce coherent outer skin surfaces.

  • I tried to take advantages of the “BEM segmentations” available in this datasets in the NIFTI format with the kind of Freesurfer code under. It produces surfaces that are not really consistent with I would expect, and with a wrong orientation when I visualize them on MNE (figure below).
    Figure 2024-06-03 104544

    mri_convert $dirbem'AVG'$subj'_segmented_BEM5.nii.gz' segmented_head.mgz

    mri_binarize --i segmented_head.mgz --match 5 --o outer_skin.mgz
    mri_tessellate outer_skin.mgz 1 outer_skin
    mris_extract_main_component ./outer_skin ./outer_skin2
    mris_convert ./outer_skin2 $dirout'outer_skin.surf'

Now I am quite lost and I do not know where to go from there. Has anyone managed to use other MRI templates than fsaverage or infants for their source reconstructions needs? Has anyone use children MRi templates in MNE?

Thank you for your help!