Muscle artifacts

Discussions
1

One of my recent patients was apparently really tense during his MEG scan. This led to high-frequency “motion-type” artifacts throughout the run. Is there a way to clean out the signal instead of rejecting the whole run? Note, I am interested in normal high-freq brain signals so LPF is not exactly an option.

Any suggestions on what to do?

Thanks,
YP

2

Can you send me a snap shot of the noise?

Beth

3

Hey Beth,

Here is an image of one of the epochs with the contamination. The channels with the most high-frequency artifacts are highlighted in red. They mostly correspond to the left fronto-temporal channels!

dep06_bline_contaminated.jpg1024×640 272 KB

4

I think you are correct, it seems to be muscle. We have had some success removing this type of artifact with the SSP process. I would suggest trying the following:
-mark spike-type events. This can be facilitated by opening a topography window and finding the offending topography - click on the ‘spiky’ events and try to determine what the artifact looks like. Then when you are choosing the events, find those that match this offending topography. You may also want to filter the data [30,150]Hz for visualizing the events and topography.
-choose a time window that captures these ‘events’. This is typically [-6,6]ms.
-run the custom SSP process on these events. (use the defined window and filter here) Note you will need a large number of events.
-inspect the SSP component (in the SSP review window) to be sure you are only removing the offending topography.

Please let me know if anything is less than clear.
Beth

5

Thanks Beth.

I have a few issues with implementing the SSP.

  1. I am working with 2 second epochs instead of my continuous signal. Is that a problem?
  2. For some reason, the custom detection for events isn’t necessarily working. I tried to manually define the artifacts to test out the effects of the algorithm (although, it’ll be tedious to do this for nearly 200 segments)
  3. Also, I’m not sure where to define the -6ms to 6ms time window
  4. The compute SSP is still grayed out in the visualization panel.
  5. Is there an undo button on Brainstorm?

Thanks,
YP

6
  1. You will want to perform this detection on the raw data - not imported epochs.
  2. The custom detection will not work properly on epoch data, again you will want to use raw, continuous data, or mark them by hand.
  3. Once you have your events marked, you will compute the SSP -> Compute SSP: Generic, enter your event name, time window=[-6,6]ms, frequency band=[30,150]Hz, Sensor types=MEG
  4. The SSP interface does not work on imported data.
  5. There is no undo button.

Beth

7

Hi ebock,

Does this method work on scalp EEG with 19 electrodes?

Regards,
Mubeen J

8

Does this method work on scalp EEG with 19 electrodes?

You might have better luck with ICA:
https://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy#Artifact_cleaning_with_ICA