Subcortical vs Cortical time-frequency source localization

Discussions
,
1

Hi Brainstorm Team!

I have developed understanding and practice for EEG source localization in the past few months.
I am able to go from pre-processed and epoched EEG data (event based, trial-design) in time domain to tomographies of different frequency bands at the cortex level.

Now, I have another EEG dataset to deal with but with the below differences:

  1. I have one trial only from each participant. This is not event/trial based design.
    The single trial lasts for around 3 minutes. I also have pre and post experience recorded for 2 minutes.

  2. I would like to include sub-cortical regions in my model - previously my localization efforts where focused only on the cortex.

For your reference, this is the pipeline I have used previously:

  1. Estimate the source for each trial using MNE (output is in pA.m)
  2. Estimate TF decomposition for each trial using morlet wavelet (to save time, consider averaging time-bins and frequency bands)
  3. Average TF maps within a subject.
  4. Perform normalization for the averaged map using ERD/ERS method.
  5. Average maps across subjects and plot the average time-frequency map.
  6. Alternatively, choose a scout and perform statistical analysis.

I would like to particularly understand what changes in this pipeline in this case?

2

Brainstorm has the tools to analyze continuous data as well.

To do so, you can use volume head models:
https://neuroimage.usc.edu/brainstorm/Tutorials/TutVolSource

The new pipeline mostly depends on the research goal and the hypothesis that you want to test on that data. Check the literature for similar experiment designs for inspiration.

As the experiment design is not multiple event-related trials, if the assumption is that brain activity during the unique long trial is sustained, you may want to split that trial in block when importing it, so the analysis can be similar to the event-related data. Although the time dimension in the average TF will not be tied to an event. Also, the meaning of baseline would be now different, baseline now means the signal before the long trial. Brainstorm allows to do baseline normalization with respect to other file rather to a time segment in the same file.

PS. Apologies for the delayed response, we just realized we had missed your message.