Optimised electrodes registration

đź’¬ Support & Discussions
1

Hey,

I want to create the trans file for my cohort. All the analysis will be run in the subject native space so I will have one trans file per subject. I am using mne.gui.coregistration() to import the subject’s electrode positions

Screen Shot 2023-03-14 at 16.03.06
Screen Shot 2023-03-14 at 16.03.06621Ă—578 114 KB

Then I pressed fit (ICP) to get the electrodes closer to the surface but as you can see some of them are basically in the air.

Screen Shot 2023-03-14 at 16.02.27
Screen Shot 2023-03-14 at 16.02.271439Ă—791 170 KB

How can I get all of them correctly registered and projected into the surface?

Thanks

2

Hi Dave,

For coregistration, you want to use the high-density scalp surface from mne_make_scalp_surfaces. The next step is to align the fiducial points (LPA, RPA and nasion) … they look completely off in your coregistration. Once that is aligned, you may further fine-tune using the electrode positions. Note that it’s not totally unreasonable if the positions are slightly above the scalp since the EEG electrodes slightly above, but your current co-registration does not look correct.

Mainak

3

Thanks @mainakjas,

I’ve run mne_make_scalp_surfaces for one subject and get the high-density scalp surface.
Re: aligning the fiducials, i am not quite sure what do you mean and how to actually do it. I re-do the montage and projected back in the new surface. Making the high-density scalp surface doesn’t seem to have resolve the issue

Screen Shot 2023-03-15 at 11.12.18
Screen Shot 2023-03-15 at 11.12.10

So I guess the problem can be the fiducial alignment. Any insight on how practically solve this?

Thanks

4

are you sure it’s the MRI of the subject? It really looks like a too small head

Alex

5

@agramfort Yep 100% sure it is the MRI of the subject. For being absolutely clear this is another subject to the one that I posted before but same story/problem

6

did the subject have some kind of bandage around the head?

Alex

7

Hi Dave,

Can you follow the steps here: https://www.youtube.com/watch?v=ALV5qqMHLlQ

You need to mark the same points on the MRI that were digitized during the EEG recording. See here:

https://www.fieldtriptoolbox.org/faq/how_should_i_report_the_positions_of_the_fiducial_points_on_the_head/

There may be some variation in how different labs report these points, but you need to mark the same landmark that was marked during the experiment.
Once the fiducials are aligned, the ICP algorithm is likely to be helpful for fine-tuning.

Mainak

8

Thanks @mainakjas and @agramfort. I followed the video tutorial but I think the problem is that the electrodes x,y,z coordinates do not match with the head shape. They seem to form a shape that is larger than the subject head

Screen Shot 2023-03-15 at 17.53.51
Screen Shot 2023-03-15 at 17.53.51469Ă—576 140 KB

Screen Shot 2023-03-15 at 17.53.43

THis is an open dataset so I have no idea how the data were collected. Would you have any idea on how to work around this problem?

Thanks

9

Your best bet is to contact the authors of the dataset asking them for more information. I would not recommend hacking and moving forward without understanding what is going on. The fiducials need to match as a first pass … otherwise something is wrong. You could also try another subject to check if it is a systematic problem in the dataset or a one-off.

Mainak