Automatic detection and spatial clustering of interictal discharges in invasive recordings

Interictal epileptiform discharges (spikes) represent electrographic marker of epileptogenic brain tissue. Besides ictal onsets, localization of interictal epileptiform discharges provides additional information to plan resective epilepsy surgery. The main goals of this study were: 1) to develop a reliable automatic algorithm to detect high and low amplitude interictal epileptiform discharges in intracranial EEG recordings and 2) to design a clustering method to extract spatial patterns of their propagation. For detection, we used a signal envelope modeling technique which adaptively identifies statistical parameters of signals containing spikes. Application of this technique to human intracranial EEG data demonstrated that it was superior to expert labeling and it was able to detect even small amplitude interictal epileptiform discharges. In the second task, detected spikes were clustered by principal component analysis according to their spatial distribution. Preliminary results showed that this unsupervised approach is able to identify distinct sources of interictal epileptiform discharges and has the potential to increase the yield of presurgical examination by improved delineation of the irritative zone.