Large peaks in resting state EEG source time course

Support & Discussions
, , ,
1

Dear MNE Users,

I am doing source localization with resting state EEG, and have noticed the source time courses in my data often show a few very large peaks. Below is an example of a source estimate and its root mean square (RMS) time course, with the red horizontal line indicating a threshold of 18 median absolute deviations (MAD). In the zoomed-in panels you can see that peaks normally last ~200-650 ms. These source space peak latencies sometimes align with small residual artifacts in the sensor space. I’m wondering if these peaks could be due to residual artifacts that are being amplified in the source space, or if there are other issues I should be considering.

I’ve tried varying the lambda parameter value and the distributed source method I used (eLORETA, sLORETA, dSPM, MNE) to help rule out inadequate regularization or method-specific issues, and the peaks look the same. On a vertex level, a large proportion (or all) vertices are usually involved in creating these peaks.

Have others working with resting state EEG run into this issue? If these peaks in the source space are due to residual eye blinks etc., I’m wondering why there would only be a few peaks per file and why all vertices tend to be affected.

Thank you in advance for your help,

Isabella

image
image1702×846 221 KB

2

Hi Isabella,

I haven’t come across this issue yet. So just a few thoughts.

You mentioned that there are some residual artefacts in sensor space and that those artefacts somtimes align with the peaks.

I would explore those artefacts a bit more in sensor space to get an idea of their origin:

  • Which electrode shows the strongest artefact?
  • Which frequency bands?
  • If you remove those artefacts in sensor space (filter, etc.), do the peaks in source space disappear?

If you have simultaneous eye tracking or electrodes below the eyes, you could try to correlate with eye movements/pupil dilation, etc., but this is almost a research project on it’s own :slight_smile:

Hope this is helpful,

Cheers,

Carina

3

Hi Carina,

Thank you for your reply. I think these peaks in the source space were primarily due to artifacts. After adding an additional artifact removal step to my preprocessing pipeline, the data in the sensor space look cleaner and these large peaks in the source space are mostly attenuated or gone.

Best,

Isabella

1 Like
4

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.