|
From: <bri...@nc...> - 2017-02-22 19:06:48
|
We see the same thing - if the electrode gets any pre-processing like re-referencing, filtering, and/or baseline correction, it will probably look like good EEG data. This flat line channel becomes a negative cms/drl signal, so it looks like perfectly reasonable EEG data. We run a high-offset check before any pre-processing of the raw data to identify these sites in an automated way and force them to be interpolate in the first faster step. Brian -----Original Message----- From: "Richard Macatee" <ma...@ps...> Sent: Wednesday, February 22, 2017 9:38am To: fas...@li... Subject: [Faster-eeg-list] Contaminated Channel Detection - Identifying removed channels Hello, Occasionally, our lab has to pop out a particular electrode that is causing problems with getting stable impedance readings for the rest of the electrodes on the cap. I assumed FASTER's contaminated channel detection algorithm would automatically identify these channels given that their values would have very little variance and poorly correlate with surrounding channels. However, I've noticed for multiple subjects FASTER will not identify these electrodes as dead despite the raw values being abnormally constant (e.g., -258 +/- 4 microvolts throughout the duration of recording) and poorly correlated with surrounding electrodes - in order to interpolate these electrodes, I have to enter them in the 'known bad channels' field. Any idea why the algorithm isn't identifying these removed electrodes as contaminated? Thanks everyone! -Ricky |
