I'm new to Brainstorm. I came across a problem and hope somebody could help me.
I found a kind of noise segments that appeared regularly in almost each participant's recording and their noise recordings (with stimulation on and nobody in the scanning room).
Do I need to mark each of this kind of noise segements as bad, or leave them and calculate the noise covariance and go to next step? (I picked several of this kind of noise recordings, as shown in the picture)
Besides, honestly, I don't understand the significance of calculating noise covariance, even after reading the tutorial, although I know it's necessary for source estimation. If someone could explain this briefly to me.
The noise covariance is computed on empty room (for MEG) to have a measure and mode the instrument noise (sensor noise). The noise covariance matrix is used in MNE to whiten the sensor data to reduce the between-sensor correlation caused by instrumental noise. You can further check this references:
D. Engemann and A. Gramfort
"Automated model selection in covariance estimation and spatial whitening of MEG and EEG signals" Redirecting
(Check the Introduction)
If the perturbations do not go away after low pass filtering within your bandwidth of interest for brain data, or after designing a particular SSP based on these bad segments, I would exclude these segments as bad as the perturbation seems quite strong. Best would be to find the cause of this environmental artifact through, and suppress it, if possible.
Can't say I've seen anything like that before. I'd guess something electronic, maybe wireless. Environmental artefacts can be difficult to hunt down. It helps if you can find when it started, when it happens (time of day, week vs weekends), etc. But that's beside your current issue of dealing with it in existing data.
