Discrete Scale Invariance of Human Large EEG Voltage Deflections is More Prominent in Waking than Sleep Stage 2.

Todd Zorick,Mark A. Mandelkern

doi:10.3389/fnhum.2015.00638

Abstract

Electroencephalography (EEG) is typically viewed through the lens of spectral analysis. Recently, multiple lines of evidence have demonstrated that the underlying neuronal dynamics are characterized by scale-free avalanches. These results suggest that techniques from statistical physics may be used to analyze EEG signals. We utilized a publicly available database of fourteen subjects with waking and sleep stage 2 EEG tracings per subject, and observe that power-law dynamics of critical-state neuronal avalanches are not sufficient to fully describe essential features of EEG signals. We hypothesized that this could reflect the phenomenon of discrete scale invariance (DSI) in EEG large voltage deflections (LVDs) as being more prominent in waking consciousness. We isolated LVDs, and analyzed logarithmically transformed LVD size probability density functions (PDF) to assess for DSI. We find evidence of increased DSI in waking, as opposed to sleep stage 2 consciousness. We also show that the signatures of DSI are specific for EEG LVDs, and not a general feature of fractal simulations with similar statistical properties to EEG. Removing only LVDs from waking EEG produces a reduction in power in the alpha and beta frequency bands. These findings may represent a new insight into the understanding of the cortical dynamics underlying consciousness.

Highlights

Rarely explicitly stated, the dominant model for the analysis of human EEG signals for more than 50 years has been spectral analysis, implicitly viewing the time-dependent changes in cortical local field potentials as a set of dynamic and standing electrical waves originating from the top 5 mm of the cerebral cortex (Nunez and Srinivasan, 2006; Nunez, 2010)
Many lines of evidence originating in multielectrode array recordings of rodent cortex (Plenz and Thiagarajan, 2007; Gireesh and Plenz, 2008; Klaus et al, 2011), extending through human electrocorticography (Priesemann et al, 2013), electroencephalography (EEG; Poil et al, 2012; Palva et al, 2013), magnetoencephalography (MEG; Shriki et al, 2013; Yu et al, 2013), discrete scale invariance (DSI) and EEG large voltage deflections (LVDs) and magnetic resonance imaging (MRI; Kitzbichler et al, 2009) have resounding demonstrated that a fundamental organizing principle of neuronal activity in the cerebral cortex is via dynamic distributions of scale-free ‘‘neuronal avalanches’’
Given that the EEG is likely determined by time-dependent dynamical neuronal avalanches, we reasoned that such structure in the waking autocorrelation function (ACF) could indicate the loss of some measure of scale invariance in the neuronal avalanche distribution

Summary

Introduction

The dominant model for the analysis of human EEG signals for more than 50 years has been spectral analysis, implicitly viewing the time-dependent changes in cortical local field potentials as a set of dynamic and standing electrical waves originating from the top 5 mm of the cerebral cortex (Nunez and Srinivasan, 2006; Nunez, 2010). Such techniques have proven spectacularly successful in both research and clinical medicine (Schwilden, 2006; Arciniegas, 2011; Schiff et al, 2014). EEG and MEG signals themselves have been shown to be highly non-stationary, albeit amenable to analysis as ‘‘quasi-stationary’’ (Kaplan et al, 2005)

Methods

Results

Conclusion