Abstracts

Rationale: Complex network analysis and graph theory provide powerful tools in the study of epilepsy. Here, we use network analyses based on mutual information to identify differences in connectivity between baseline and seizure periods using electrocorticographic data acquired from epileptic patients undergoing seizure monitoring.Methods: We calculate the average mutual information between the voltage traces from a pair of electrodes for a range of time lags, which allows us to extract the strength, latency, and direction of the connection between that pair. For each patient we construct baseline networks using short time blocks collected over several days in order to examine the temporal evolution of these networks. We compare the connectivity at different time points by calculating measures of similarity between adjacency matrices and by comparing the networks’ topological metrics.Results: First, we find that the connectivity of the baseline networks is stable over time scales ranging from minutes to days. Next, with these baseline networks in hand, we then compare the topology of these networks to those computed during seizures. We divided each patient’s seizures into a pre seizure, early seizure, late seizure, and post seizure period, and construct a network for each section. We find significant changes in network structure between the stable baseline period and all seizure stages.Conclusions: Our preliminary data suggest that much of the observed differences in network topology during seizure activity is due to a decrease in connectivity during all stages of the seizure as measured by the network density and average degree.