Strange EEG Source Activation Z-Scored Values

Discussions
1

Hi Brainstorm Community,

I'm a bit puzzled by some numbers I'm getting and I'm wondering if anyone else has encountered this and/or knows where I went wrong. When looking at source activation from an ERP task, if I Z-score the data using my baseline period (e.g. -400 to -100), my Z-values around ERP times (e.g. ~100ms, 200-400ms, etc) are commonly 20-40 and even as high as 60 (figure below).

How I've handled the data so far:

When importing from raw recordings I removed DC offset
I then averaged between conditions but within participants
I then average across participants to create a single averaged ERP for each condition
I then sourced that condition's averaged ERP and z-scored relative to the baseline

example.png900×1013 134 KB

Thanks for any help!
Anthony

2

Hi Anthony,

I don’t think these values are too strange. You have a very flat baseline, therefore the variance over your baseline is close to zero.
After 100ms, you have values that are relatively high with respect with this baseline (divided by the baseline variance, which is close to zero).

What could be improved in your pipeline is the order of the processing:

  1. For each subject: average EEG recordings, then compute the sources. The noise covariance can be very different for each subject, you need to configure your subject with one channel file per subject and compute one inverse per subject:
    http://neuroimage.usc.edu/brainstorm/Tutorials/NoiseCovariance#Other_scenarios
  2. Z-score normalize each subject (to account for difference of amplitude across subjects)
  3. Average the normalized values across subject.

I’m currently writing a new page about these things, it’s not finished bu maybe it can help you a bit:
http://neuroimage.usc.edu/brainstorm/Tutorials/Workflows

Cheers,
Francois

3

Thanks Francois!