State-dependent precursors of seizures in correlation-based functional networks of electrocorticograms of patients with temporal lobe epilepsy.

Accurate prediction of epileptic seizures will open novel therapeutic possibilities for patients with intractable epilepsy. We attempted to identify precursors of seizures in the functional networks of electrocorticograms by applying graph theory. Long-term electrocorticograms for periods of 39-76 h from three patients with temporal lobe epilepsy were investigated using pair-wise cross-correlations. Time-varying network properties suggested that there were several distinct brain states. Although functional networks during seizures could be characterized as having a regular topography, no consistent characteristics of functional networks were found immediately prior to seizure onsets. However, it was found that seizures under an identical state were followed by similar transients of the network properties. These results suggest that network properties themselves could not serve as reliable predictors of seizure onset. Yet, some significant pre-seizure changes in the parameters tested appear likely to depend on the brain state. To predict seizures, it may be necessary to take into consideration the states of the brain. In addition to stationary network properties we characterized in the present study, dynamic interactions of epileptic activities with the network might be taken into account to predict the spread of a seizure.

RCT Entities:   Population Interventions Outcomes