Abstract
AbstractEEG signals are commonly analyzed in the frequency-domain. Analysis of the power spectrum commonly demonstrates a reproducible fall-off of high frequencies, sometimes referred to as the “1/f slope”. The 1/f slope comes about from a linear fit of the power spectral density in log-log form and takes typical values between (minus) 0–4. This slope has been recently hypothesized to correlate with the ratio of excitatory to inhibitory synaptic weights. However, we demonstrate models of excitatory-inhibitory (E/I) balance only explain small values of 1/f slope (< 2). We seek to construct a model to explain large slopes (> 2), which have been reported under conditions of anesthesia. Using simulations of clustered spike pulses, we find that pulse widths on the order of 1-10 ms result in a transition to steep 1/f slopes between 3–4. This trend also holds when measuring the 1/f slope from a spiking recurrent network model in a synchronous regime. We conclude that steep 1/f slopes result from synchronous spiking in bulk microcircuit activity. This suggests interpretations of steep 1/f slopes should consider spike synchrony in addition to other factors like E/I balance.
Publisher
Cold Spring Harbor Laboratory