Critical dynamics in the human brain

Abstract

Introduction The brain spontaneously generates complex patterns of oscillatory activity whose dynamic signature is poorly understood. Identifying hidden structures in the electrophysiological time series could yield important insights into the function and mechanisms of large-scale neuronal interactions. Self-organized criticality (SOC) provides a general theoretical framework of how complex systems can evolve to a non-equilibrium state characterized by power-law scaling behavior [l]. This statistical property of the critical state stems from a non-linear process creating transient spatio-temporal correlations at wide scales and lends the system susceptible to change. We explore the possibility that the dynamical fingerprint of criticality in a family of complex systems is shared by oscillations in the human brain. Methods Spontaneous electrical brain activity from 8 normal subjects (aged 20-30 years, 1 female) was recorded using a whole-scalp magnetometer with 122 planar gradiometers. The subjects were seated in a magnetically shielded room and instructed to relax with eyes either open or closed in two separate 20-minute recording sessions. The measurements were replicated in 4 subjects, giving 12 data sets for each condition. The time-varying signal amplitude of narrow frequency bands was estimated by means of wavelet analysis. We focussed the analysis on the 8-13 Hz frequency range due to the high signal-to-noise ratio of the alpha rhythm. Results Wavelet transformation of the time series revealed amplitude fluctuations on wide time scales (Fig. 1). To detect power-law temporal correlations, characteristic for scale-invariant systems far from equilibrium, we employed the detrended fluctuation analysis [2]. The outcome was a remarkably invariant and persistent scaling behavior across subjects and conditions (Fig. 2), indicating a lack of ‘typical’ time scales for the duration and recurrence of oscillations. Least-squares fits in double logarithmic coordinates yielded self-similarity parameters