Source estimation without T1 image for cluster permutation (group analysis)

💬 Support & Discussions
,
1

Dear MNE community,

I want to perform a cluster permutation test comparing time-frequency (morlet waves) EEG data from 40 subjects performing a task in two different sessions and then visualize the significant clusters.

However, I have not collected individual T1 images from the subjects, thus I want to do the source estimation using the ‘fsaverage’ model.

For doing the source estimation for just one subject I have to perform the following steps:

  1. Coregister ‘fsavarage’ with my subjects’ contrasts (epoch_session_1, epoch_session_2)
  2. Compute the source space
  3. Compute BEM solution
  4. Compute the forward model
  5. Calculate the noise covariance matrix
  6. Create and apply the inverse operator
  7. Compute the adjacency matrix

But if I have 40 subjects, should I perform source estimation (all steps above) for all my 40 subjects individually?

Or can I do it only once after concatenating epochs (or evokeds???) of all 40 subjects?

I’d be happy to explain further details and thankful for any feedback

Best regards,

Bruno

  • MNE-Python version: 0.22.0
  • operating system: * Ubuntu 20.04.2 LTS
2

Any comment would be really appreciated. I still don’t know how to proceed.

3

Hello, sorry for the late response, I guess we were all a little busy with the sprint

Since noise levels typically vary between recordings, and since the noise covariance plays an integral role during source estimation, I would suggest you calculate the noise covariance and inverse for each participant individually, and average across all participants only at the very end. If you wish to calculate any statistics, e.g. contrast two experimental conditions, you will need those individual solutions anyway.

Hope this helps!

1 Like