Critical and Ictal Phases in Simulated EEG Signals on a Small-World Network.

Victor Castro,Andon Placzek,Jennie Q Lou,Robert M Worth,Louis R Nemzer,Francis Motta,Gary D Cravens

doi:10.3389/fncom.2020.583350

Abstract

Healthy brain function is marked by neuronal network dynamics at or near the critical phase, which separates regimes of instability and stasis. A failure to remain at this critical point can lead to neurological disorders such as epilepsy, which is associated with pathological synchronization of neuronal oscillations. Using full Hodgkin-Huxley (HH) simulations on a Small-World Network, we are able to generate synthetic electroencephalogram (EEG) signals with intervals corresponding to seizure (ictal) or non-seizure (interictal) states that can occur based on the hyperexcitability of the artificial neurons and the strength and topology of the synaptic connections between them. These interictal simulations can be further classified into scale-free critical phases and disjoint subcritical exponential phases. By changing the HH parameters, we can model seizures due to a variety of causes, including traumatic brain injury (TBI), congenital channelopathies, and idiopathic etiologies, as well as the effects of anticonvulsant drugs. The results of this work may be used to help identify parameters from actual patient EEG or electrocorticographic (ECoG) data associated with ictogenesis, as well as generating simulated data for training machine-learning seizure prediction algorithms.

Highlights

The human brain must remain sensitive to new stimuli and coordinate spatially distant information processing modules to function optimally in a continuously changing environment
It is thought that epilepsy occurs when the human brain is unable to dynamically operate in or near this critical state, where certain biological neuronal networks work near phases of pathological synchronization and insensitivity
It goes without saying that no simulation, especially with orders of magnitude fewer nodes, can reflect all of the complexity of the human brain, and we have focused on aspects more relevant for future machine learning algorithms for predicting seizures in patients with epilepsy

Summary

Introduction

The human brain must remain sensitive to new stimuli and coordinate spatially distant information processing modules to function optimally in a continuously changing environment. A biological system, such as the human brain, in the critical state near a phase transition, is maximally sensitive to external influences (Larremore et al, 2011), and is most efficient in amplifying small perturbations Such a state allows for long-range coordination between brain regions, with a theoretical correlation length that diverges to infinity. Epilepsy is one of the most common central nervous system (CNS) diseases (Zack and Kobau, 2017), affecting ∼50 million individuals worldwide, including both men and women of varying ages. It is a chronic neurological disorder characterized by a persisting predisposition to generate epileptic seizures and by the resultant neurobiological, cognitive, psychological, and social consequences. Most patients have little or no warning before a seizure occurs, and this unpredictability can have profound impacts on in their lifestyle, including restrictions on driving, or constraints on employment opportunities

Objectives

Methods

Results

Discussion

Conclusion