The noise in the EEG signals from the BIS #696
leonip (2025-03-03 11:34)
· 36 view(s)
· 1 comment(s)
Hi,
What are these high, spiky noises in the raw EEG data from the BIS Vista? How can they be removed?
They literally exist in every single case in this dataset, filtering barely works on them. They are a disaster for EEG analysis.
I'm not sure if this is noise that comes with the dataset or if it's caused by my data transformation process (from .vital format to .mat format).
Ross Thomson (2025-03-08 18:56)
I don't use Matlab and therefore am not familiar with the .mat format, but I think it's very unlikely the conversion process is generating this noise. Look at the original .vital file in VitalRecorder - you can easily see if the noise was present pre-transformation.
The timebase on the figures you provided isn't clear - is the scale in hours? The noise seems to occur regularly. A classic cause of this is the nurse rolling/suctioning the patient, although it seems to occur a bit too frequently for that. I don't use the BIS Vista device so can only speculate, but another possibility is that it is recalibrating or sending system status data, etc. at regular time points.
Irrespective of the cause of the noise - and it would be good to identify before you record from more patients - it should be fairly straightforward to remove it from these tracks. The noise is an order of magnitude larger than the data, so you can easily apply simple outlier filtering rules.
The timebase on the figures you provided isn't clear - is the scale in hours? The noise seems to occur regularly. A classic cause of this is the nurse rolling/suctioning the patient, although it seems to occur a bit too frequently for that. I don't use the BIS Vista device so can only speculate, but another possibility is that it is recalibrating or sending system status data, etc. at regular time points.
Irrespective of the cause of the noise - and it would be good to identify before you record from more patients - it should be fairly straightforward to remove it from these tracks. The noise is an order of magnitude larger than the data, so you can easily apply simple outlier filtering rules.