About the source report

1

Dear Brainstorm Experts:

Since the source data I'm currently studying reported an activation area at the bottom of the skull (the red circle), but there is no such ROI area on Destrieux Atlas.

圖片2
圖片21943×1125 470 KB
圖片1
圖片11629×1126 429 KB
1
1813×1126 725 KB

I would like to ask you, is this a problem of noise?
Or should I add some steps to the pre-processing process?

Thank you very much for your help.

Yours,
James

2

There are no Destrieux parcels in the central parts of the FreeSurfer surfaces, because they do not correspond to any real anatomical structure. The FreeSurfer output is represented as two separate and closed hemispheres, while the cortex is in reality not closed. Some parts of the these surfaces are artificial, and do not appear in the anatomical atlases.
The other half of the response is that the Destrieux atlas is a parcellation of the cortex, while some of the vertices you point at are probably part of subcortical structures (artificially integrated to the FreeSurfer cortex surface).

If you want a scout in these regions, just create it manually.
Or use a volume head model and a volume atlas.

Be very careful when interpreting very deep sources. The minimum norm solution reconstructs non-zero values everywhere you have dipoles in your source model. See all the forum posts about dba, and read the articles related to the tutorial "Deep cerebral structures":

The X shape over your "baseline" (which you probably set to 2-2.5s) is suspicious. What you have over this time window you used to center your recordings is not a "baseline", there are clearly things happening in the brain, there. Otherwise, it would average out to zero in your ERP.

3

Dear Francois:

Thank you for your reply.
However, regarding the choice of baseline.
Since I extracted the time period:1899~3899ms from a section of Epoch(4900ms) for analysis, I use the 1899~2399ms as my baseline correction, I am not quite sure if it is correct to do so?

1
11995×996 61.4 KB

Thank you very much for your help.

Yours,
James

4

Since I extracted the time period:1899~3899ms from a section of Epoch(4900ms) for analysis, I use the 1899~2399ms as my baseline correction, I am not quite sure if it is correct to do so?

Therefore there is some fair amount of time-locked brain processing expected during 1.9s-2.4s, this X shape of the signals is expected. It is up to you to decide whether this a good idea to use this time window as baseline to center your EEG signals. Referring to what is done in similar experiments in the literature is probably your best guess.

To compute this ERP, I would have two alternative approaches to suggest:

  1. import a segment [-0.7s,4.2s] (a few hundred ms extra for time-frequency analysis, see: https://neuroimage.usc.edu/brainstorm/Tutorials/Epoching#Epoch_length), and remove the DC offset estimated from [-0.5s, 0s].
  2. high-pass filter the signal (threshold somewhere between 0.3Hz and 2Hz, depending on the temporal dynamics of the brain processes your are expecting to observe), and not do any additional baseline correction on the EEG data.

I'm not an ERP expert, and these might not be appropriate advices. The design of your processing pipeline should depend mostly on your experimental hypotheses and design, trying also to stay as close as possible to your reference literature.