Epileptogenic Networks: Applying Network Analysis Techniques to Human Seizure Activity

Abstract

A key aspect of the treatment of patients with medically intractable focal epilepsy is surgical resection of the seizure focus. This requires the seizure onset zone to be localized to a specific area of brain tissue. Currently, this seizure locus is largely identified via expert visual assessment of the electroencephalogram (EEG) and electrocorticogram (ECoG). Therefore, localization could potentially be improved by (1) the incorporation of recording techniques with higher spatial and temporal resolution; and (2) by analyses that help to objectively and quantitatively identify the source of epileptic activity. Here, we discuss the application of signal processing and graph theory measures to human ECoG and depth electrode data. We demonstrate the application of a subset of these techniques, including cross correlation, coherence, and Granger causality, to analyze a recording of a seizure from a multi-electrode array (MEA) implanted in human cortex. The results are used to define epileptogenic networks, which can be validated against the visually identifiable development and propagation of the seizure. We then characterize each network and quantify the degree of similarity between networks in order to directly compare analysis techniques.