Subgraph “Backbone” Analysis of Dynamic Brain Networks during Consciousness and Anesthesia

General anesthesia rapidly alters brain network connectivity, and standard graph-theoretical methods that assume static networks are poorly suited to capture these changes. A new technique was introduced to track the temporal evolution of network modules. Multichannel EEG was recorded from 18 surgical patients anesthetized with propofol or sevoflurane. Networks were reconstructed by treating each EEG channel as a node and correlated activity between channels as a link. The frequency of subgraphs with a defined number of links was analyzed; subgraphs with high occurrence probability were called network "backbones." Constitutive, variable, and state-specific backbones were identified across consciousness, induction, maintenance, and recovery. This approach enables a granular, dynamic description of network evolution.

Study at a glance

Abstract

General anesthesia significantly alters brain network connectivity. Graph-theoretical analysis has been used extensively to study static brain networks but may be limited in the study of rapidly changing brain connectivity during induction of or recovery from general anesthesia. Here we introduce a novel method to study the temporal evolution of network modules in the brain. We recorded multichannel electroencephalograms (EEG) from 18 surgical patients who underwent general anesthesia with either propofol (n = 9) or sevoflurane (n = 9). Time series data were used to reconstruct networks; each electroencephalographic channel was defined as a node and correlated activity between the channels was defined as a link. We analyzed the frequency of subgraphs in the network with a defined number of links; subgraphs with a high probability of occurrence were deemed network "backbones." We analyzed the behavior of network backbones across consciousness, anesthetic induction, anesthetic maintenance, and two points of recovery. Constitutive, variable and state-specific backbones were identified across anesthetic state transitions. Brain networks derived from neurophysiologic data can be deconstructed into network backbones that change rapidly across states of consciousness. This technique enabled a granular description of network evolution over time. The concept of network backbones may facilitate graph-theoretical analysis of dynamically changing networks.