Automatic Identification of Interictal Epileptiform Discharges with the Use of Complex Networks

Abstract

The identification of Interictal Epileptiform Discharges (IEDs), which are characterized by spikes and waves in electroencephalographic (EEG) data, is highly beneficial to the automated detection and prediction of epileptic seizures. In this paper, a novel single-step approach for IEDs detection based on the complex network theory is proposed. Our main goal is to illustrate how the differences in dynamics in EEG signals from patients diagnosed with idiopathic generalized epilepsy are reflected in the topology of the corresponding networks. Based on various network metrics, namely, the strongly connected component, the shortest path length and the mean jump length, our results show that this method enables the discrimination between IEDs and free IEDs events. A decision about the presence of epileptiform activity in EEG signals was made based on the confusion matrix. An overall detection accuracy of $$98.2\%$$ was achieved.