Artifact or real gamma power?

💬 Support & Discussions
1

Hello everyone,

I struggle with my EEG data. I found an electrode main effect for my data (rm-ANOVA), but nothing more. I used a memory paradigm with different stimuli, and I expected a memory and stimuli-type effect as well.
Now I wonder if there is an explanation for this. When I did a TFR-plot collapsed over all conditions, I found this sort of “bar” in most of the electrodes between 45 to 55 Hz. Could this explain the electrode effect and is this only noise?

MicrosoftTeams-image
MicrosoftTeams-image1755×834 86.7 KB

MicrosoftTeams-image (1)

Thanks a lot
Franziska

2

The horizontal lines around 50Hz is caused by a notch filter either you or the amplifier has applied to the signal. The power in this band is very small (around zero), as expected. BTW, I’m not sure why you use a color scale which extends to negative numbers – this doesn’t make sense if you are showing power.

The main effect of electrode is probably not caused by the presence of a notch filter. This effect means that PSDs are different over electrodes, and judging from your image the left and right outermost electrodes show much more activity than the more central channels. I think that this might be causing this effect.

3

Hi

thank you very much for your evaluation
Yes, totally makes sense to not use a scale that extends to negative values

I thought that the electrode effect might be caused by noise, because the activity found in the outermost electrodes does seem random. Maybe I am confused because I thought of finding a nice blub rather than kind of bursts. So you would think this is “real” gamma activity and not only noise?

Thank you so much
(confused PhD-student)

4

Hi Franziska,

I recommend starting with a dataset that is known to show gamma. Example, look at this fieldtrip tutorial and try to replicate it in MNE:

https://www.fieldtriptoolbox.org/tutorial/beamformingextended/

Once you have the pipeline and understand what needs to be done to detect gamma, then you can try on your own data. Otherwise you’re trying a little blind … there are many factors that affect what you’re seeing, did you baseline correct, did you average across sensors that show gamma etc. It’s hard to detect gamma because of the 1/f decay

Mainak

1 Like
5

Thanks for your tip, Mainak!