Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

BCCN/BFNT AG-Seminar

Tuesday, 18.06.2013 17 c.t.

Self- organized criticality as a universal brain state from wakefulness to deep sleep? Results from intracranial depth recordings in humans

by Dipl. Phys. Viola Priesemann
from Max Planck Institute for Brain Research, Frankfurt, Germany

Contact person: Theo Geisel

Location

Ludwig Prandtl lecture hall

Abstract

Neural activity differs from wakefulness to deep sleep. In contrast, a single attractor state, called self-organized critical (SOC), was proposed to govern brain dynamics because it supposedly allows for optimal information processing. But is the human brain SOC for each vigilance state despite the variations in neuronal dynamics? We characterized neuronal avalanches - spatiotemporal waves of enhanced activity - from intracranial depth recordings in humans. We showed that avalanche distributions closely followed a power law - the hallmark feature of SOC systems - for each vigilance state. However, avalanches clearly differed with vigilance states: slow wave sleep showed larger avalanches, wakefulness intermediate, and rapid eye movement (REM) sleep smaller ones. Our SOC model together with the data suggested (1) that these differences were mediated by global but tiny changes in synaptic strength and (2) that the changes with vigilance states reflect transition within the subcritical regime close to criticality. In contrast to criticality, a subcritical regime of operations allows for a more stable mode of operation, and keeps a safety margin to the supercritical regime, which is linked to epilepsy. To back up this study, we analyzed for SOC models how spatial subsampling affects the avalanche distributions. We show that for the above study sampling was sufficiently dense, but we also show that subsampling can heavily distort the observed avalanche distributions. These distortions depend on the number and distance between sampled sites, and can lead to misclassifications of a system. The distortions are also model-specific. This in turn allows to exploit systematic subsampling for model selection in the context of SOC and beyond.

back to overview