An Effective Connectivity Model Based on Excitation-Inhibition Imbalance to Classify States of the Epileptogenic Network

In this paper, we present an effective connectivity model based on the imbalance between neuronal excitation and inhibition to classify epileptogenic network states. We used electroencephalographic intracranial signals from refractory epilepsy patients. Autoregressive-moving-average models were adjusted to these signals to estimate effective connectivity. We estimated ‘in-degree’ and ‘out-degree’ patterns associated with inhibitory and excitatory processes. These patterns were used as variables or features in supervised classifiers to classify ictal, pre-ictal and basal states of the epileptogenic network. We obtained values of Area Under the Curve (AUC) larger than 0.99 in the distinction of these pathological brain states.