Detect bad trials with EOG: threshold does not work?

Bugs
1

Hi Francois,

I need your help again! I attempted to used the EOG channel to reject trials, as shown in the attached image. A higher resolution dropbox link is also available here:
http://dl.dropbox.com/u/4202951/EOG_rejection_error.jpg

The threshold operator seems to have an unpredictable behavior, and the image shows just an example. Setting a threshold -100 to 100μV should not have rejected the trial. However the trial is rejected after clicking ok. Any ideas?

Also, after exporting the data into Matlab and ploting the EOG data, the time series has a shift in the y-axis. This shift does not explain the incorrect trial rejection, since the time series is still within the -100 to 100 limits. Still, why the EOG data have this DC shift in Brainstorm?

Thanks and merry Christmas!

Dimitrios

2

Hi Dimitrios,

Detection:
I’m sorry, this is my fault, I never documented this option window…
It’s not working because it’s actually not meant to be used this way.
It works the same way the MNE averager works: you define two peak-to-peak thresholds for your signal, one to detect that there is actually signal recorded on the channel, one to detect that it is within a reasonable range of value.
So if you want to check that your signal is in the range of [-100, +100]μV, you should use the values: [0, 200]μV in this option window
I added a comment that explains that in the window.

Advanced detection:
We added much nicer tools to detect and clean artifacts than this one, that work only with the continuous viewer.

  1. Right-click on subject > “review RAW file”
  2. Drag’n’drop raw file in the process box
  3. Process : Detect > Detect…
    4.a) To tag as BAD the imported trials during which an event occur, call your events “bad_eog” or "bad_whatever"
    4.b) To create a projector to remove the artifact: process SSP > Compute SSP: …
  4. The original file is NOT modified, the SSP and events are save only in the “Link to raw file” => Right-click on this link > Import to database

EOG:
A shift with respect with what? Is you screenshot supposed to illustrate this shift?

Happy holidays
Francois

3

Hi Francois,

Excellent! It now makes sense and of course works! I attach a new image that shows:
Top: average of 147 EEG trials, which is too noisy
Middle: average of EEG trials with no blinks between -100 and 800ms
Bottom: average of EEG trials with no blinks between -100 and 300ms

Clearly, the Bottom is better for the initial response, even though it is too noisy afterwards. I will stress to our users the importance of avoiding eyeblinks!

Merry Christmas!
Dimitrios