High Variability Periods in the EEG: A New Temporal Metric that Reflects Brain States

Abstract

Here we define a new metric to characterize temporal patterns of amplitude variability in the EEG signal and demonstrate that this feature of the signal varies significantly with brain state. The metric uses the standard deviation of waveform amplitude in a short moving time window of 3-20 seconds with 50% overlap. We define High Variability Periods or HVPs as segments when the moving standard deviation of the waveform amplitude is continuously higher than a cutoff defined as the 25th percentile of variability. HVPs typically last on the order of tens of seconds and are punctuated by Low Variability Periods or LVPs of shorter duration. We further show that the HVP characteristics differ between various conditions. HVPs in the resting state eyes closed condition are substantially and significantly shorter in duration and smaller in area relative to eyes open. In addition, in recordings from monkeys, HVPs disappear under anesthesia and do not reappear in early periods of recovery from anesthesia suggesting long term changes to the signal. Altogether this demonstrates that HVP metrics have discriminatory power and may therefore be important in predicting brain states and outcomes. Finally, they are fast to estimate and can be tracked in time, and therefore suited for near real-time monitoring in low-electrode configuration systems.