Measuring the dynamic balance of integration and segregation underlying consciousness, anesthesia, and sleep

A dynamic balance between integration and segregation in functional brain networks is necessary for consciousness. A new fMRI-based measure, the integration-segregation difference (ISD), was developed to capture both aspects. Using this metric, changes from conscious wakefulness to loss of responsiveness induced by the anesthetic propofol showed a profound shift toward segregation in the whole brain and all subnetworks during anesthesia. Brain networks displayed similar sequences of disintegration and subsequent reintegration during loss and return of responsiveness. Machine learning models identified awake versus unresponsive states and their transitions with up to 93% accuracy. Metastability was more closely linked with integration, while complexity was linked with segregation. Analysis of a sleep dataset revealed similar findings, demonstrating that the integration-segregation balance can differentiate among various conscious and unconscious states.

Study at a glance

Abstract

Consciousness requires a dynamic balance of integration and segregation in functional brain networks. An optimal integration-segregation balance depends on two key aspects of functional connectivity: global efficiency (i.e., integration) and clustering (i.e., segregation). We developed a new fMRI-based measure, termed the integration-segregation difference (ISD), which captures both aspects. We used this metric to quantify changes in brain state from conscious wakefulness to loss of responsiveness induced by the anesthetic propofol. The observed changes in ISD suggest a profound shift to segregation in both whole brain and all brain subnetworks during anesthesia. Moreover, brain networks displayed similar sequences of disintegration and subsequent reintegration during, respectively, loss and return of responsiveness. Random forest machine learning models, trained with the integration and segregation of brain networks, identified the awake vs. unresponsive states and their transitions with accuracy up to 93%. We found that metastability (i.e., the dynamic recurrence of non-equilibrium transient states) is more effectively explained by integration, while complexity (i.e., diversity and intricacy of neural activity) is more closely linked with segregation. The analysis of a sleep dataset revealed similar findings. Our results demonstrate that the integration-segregation balance is a useful index that can differentiate among various conscious and unconscious states.