Group analysis sources: subject coregistration or scouts in def. anatomy?

Discussions
, , ,
1

Hello, Community! I am currently working on a project involving source-level time-frequency analysis of EEG data. I have collected data from approximately 30 participants, and the following is my analysis pipeline:

  1. Preprocess the data at the channel level using EEGLAB. (Clean_asr, filtering between 1-40Hz, and ICA for removing eye movement and cardiac field artifacts)
  2. Epoch the data as needed and perform single-trial rejection.
  3. In Brainstorm, create a FEM head model using T1+T2 images (using Simnibs 4 and Charm).
  4. Generate the lead field matrix (co-registration with Polhemus Fastrak .pom electrode positions).
  5. Perform minimal norm imaging using sLORETA.

I have three types of event markers (for data from all 30 participants) and intend to conduct two types of statistical analyses:

1+2 vs 3 / 1 vs 2

Therefore, my analysis will involve paired t-tests.

However, I want to perform this analysis at the source level.

It seems that there are two methods for this:

  1. Define ROI (around 10-20) as Scouts from individual source levels, conduct time-frequency analysis in these regions, and perform statistical analysis on the resulting time-frequency maps.
  2. Subject coregister the individual source level to the default FreeSurfer anatomy and perform the time-frequency analysis within the FreeSurfer anatomy.

Which approach would you recommend? While theoretically, there should be no difference between the two methods, I expect there may be differences in the results.

2

Under this approach, the Scouts to compare will certainly be comprised of different numbers vertices, and probably slightly different location, thus differences among subject will be activity and scout size (and location). Moreover, the sign of the scout time series may also change among subjects.

Coregistering subjects to a default anatomy (aka template) is more in line with the approaches reported in literature. It removes the differences in scout definition.

Check this page, with special attention to the Group analysis sections:
https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows

3

Thank you a lot! I will read tutorial again and revise my pipeline.