Cleaning of Cross-talk in sensors adjacent to disabled ones

Hi team,
In few subject's artefact description, the following is mentioned
"Crostalk from disabled sensor MRP52 causes correlated noise in adjacent sensors: MRO11 MO12 MRP41 MRP51 MRP53. Can be cleaned with SSp from empty room recording. "
I am not able to figure out how to clean the crosstalk. Can someone help me with the steps ?

Hi @Rasmita, please provide more details:

  • Is this a dataset provided by Brainstorm?
  • Where is the "quote" coming from?

Hi Raymundo,
This data is from OMEGA dataset.
The quote is from subject_artefact description for each recording.


Hi Rasmita,
This post on the BIC forum addresses this question.

1 Like

Hi Marc,
This is the PSD after performing band pass filtering:

I followed the steps outlined in the BIC post. I notice MRP41,51 and 53 (affected by cross-talk) to have reduced amplitude as compared to neighboring time traces after applying SSP from empty-room recording. I am not sure, if this is right.

After applying the SSP projection; I again computed the PSD and found MRT sensors to have higher powers and deviated from the ensemble. Is this right ?

Also, what would you advice if for one subject had multiple empty room recordings ? I thought of using the latest ones.

Yes that's the trade-off. As stated in that post: "Cleaning causes both a large loss in power and a considerable SNR increase in affected sensors."

I see you also have ICA. This is probably a result from how the SSP was computed vs the ICA. In this unusual case where the SSP is computed independently on the empty-room, you'd need to do it first, and then compute ICA taking into account the existing projectors.

That's usually because something was happening during the first one, or a sensor was bad and defluxed. But yes you can usually ignore the details and use the last one.

Thanks Marc! This helps.