Help with EEG-MRI Coregistration Using Template Electrode Locations in MNE

Hello MNE Community,

I am encountering challenges with coregistering EEG channels to individual MRIs that have been preprocessed using FreeSurfer. Unfortunately, I do not have access to individual channel locations and must rely on template electrode location coordinates. I’m utilizing the mne.coreg function and followed the automated coregistration pipeline, but the accuracy doesn’t seem satisfactory.

I used the automated coregistration method as provided by the MNE tutorial here:
https://mne.tools/stable/auto_tutorials/forward/25_automated_coreg.html

Upon visualizing the alignment using the following code, I noticed that the original sensor positions appear to have some electrodes inside the head surface, while the projected sensors are correctly aligned on the surface.

mne.viz.plot_alignment(
info=eeg.info,
subject=subject,
subjects_dir=subjects_dir,
trans=trans,
dig=‘fiducials’,
eeg=[‘original’, ‘projected’],
surfaces=‘head-dense’,
mri_fiducials=True,
coord_frame=‘head’
)

My questions are:

  1. Which set of coordinates (original or projected) is used for source reconstruction in MNE?
  2. If the original EEG coordinates are crucial for accurate source reconstruction, how can I properly adjust the coregistration?
  3. Given that I lack individual channel location data, are there recommended methods or settings within MNE to improve the coregistration automatically?

Any insights, suggestions, or best practices to resolve these coregistration issues would be greatly appreciated.

(1) I would assume it uses the projected position.
(2) and (3) If you don’t have a measure of the electrode position, there is no magic to be done here to reconstruct a missing piece of information. You’ll have to go with the template position which might or might not match what you had experimentally (e.g. cap more positioned towards the front or the back of the head).

Mathieu

1 Like