Abstract Díaz et al. 2008

Major states of the sleep-wake cycle have been traditionally characterized by the conjunction of specific phenomenological parameters. Some of those criteria include spectral features (delta, theta and sigma bands) and EMG. Recently, a novel analysis based on the assessment of the coefficient of variation (CV) of the envelopes of neural signals has allowed inferring the degree of coupling and synchrony among neuronal oscillators (Díaz et al., DOI:10.1523/JNEUROSCI.4512-06.2007). Here we report how this method, as applied to the CV of the theta band envelope (CVtheta), could be used to recognize sleep-wake states.
As previously reported, a CV near 0.523 indicates Rayleigh fading interference, which characterizes a signal originated from independent oscillators in terms of frequency and phase, whereas a population of synchronous oscillators produces signals characterized by a lower envelope modulation and significantly lower CVs. Thus, CVtheta could be interpreted as measuring the level of synchrony (or coupling) among theta generators.
Chronically implanted rats (Sprague-Dawley, males, 300 g) were used to obtain continuous epidural recordings (3 channels, 250 Hz A/D rate, 8 bits, band pass: 0.3-30 Hz) as well as EMG (1 channel, 250 Hz A/D rate, band pass: 30-100 Hz). Full-day recordings were divided in 5 seconds epochs. After proper filtering, CVtheta was calculated for every epoch. For the same data, wake-sleep states were diagnosed by conventional methods. We found that CVtheta for the NREM state (avg=0.525; sdev=0.05) approximates the Rayleigh fading fingerprint value (CV=0.523). In contrast, for REM and AWAKE states, CVtheta was significantly lower (p<0.05). Furthermore, combining CVtheta with a second parameter (i.e. EMG amplitude or theta band amplitude), it was possible to generate 2D projections where three well defined clusters, corresponding to REM, NREM and AWAKE states, appeared.
That sleep dynamics could be subsumed into the temporal evolution of a single statistical parameter like CV shows how that this parameter not only serves to classify neural states but illuminates problem about the generation of neural signals at different levels of organization. In effect as CV has been used to characterized neural oscillations, from sensory epithelia oscillations to EEG, it appears possible measuring and understanding neural synchronization, at the mesoscopic and macroscopic levels, using a signal analysis tool that asses the degree of interference among signals emitters. FINANCIAL SUPPORT: FONDECYT Nº 1060250, 1061108.