Real-Time Epileptic Seizure Detection on Intra-cranial Rat Data Using Reservoir Computing Pieter Buteneers, Benjamin Schrauwen, David Verstraeten, and Dirk Stroobandt Ghent University, Belgium Abstract. In this paper it is shown that Reservoir Computing can be successfully applied to perform real-time detection of epileptic seizures in Electroencephalograms (EEGs). Absence and tonic-clonic seizures are detected on intracranial EEG coming from rats. This resulted in an area under the Receiver Operating Characteristics (ROC) curve of about 0.99 on the data that was used. For absences an average detection delay of 0.3s was noted, for tonic-clonic seizures this was 1.5s. Since it was possible to process 15h of data on an average computer in 14.5 minutes all conditions are met for a fast and reliable real-time detection system. 1 Introduction Epilepsy is a neurological disorder of the brain where the patient is disturbed by mostly recurring seizures. Around 1% of the world’s population suffers from this illness [1]. Although a cure for this disorder has not yet been found, medication is in most cases sufficient to block the seizures. To determine whether the applied medication is working, a doctor needs to determine the remaining epileptic activity, and thus also the seizures, on hours of recorded EEG data. Especially in the case of absence epilepsy, where the number of seizures is a very good indication, an automatic detection system is highly de- sired. Additionally anti-epileptic drugs are known for their side-effects. To avoid these side-effects one could apply a closed-loop system, where the medication is applied in real-time when a seizure occurs. A third application of a fast and reliable detection system is as a warning system. The environment of the patient can then be alerted when a seizure occurs so they are able to help and protect the patient in this unpleasant and dangerous episode. In this study we show how Reservoir Computing (RC) can be successfully applied to detect epileptic seizures on EEG in real-time. RC is a training method for recurrent neural networks where only a simple linear readout function is trained for a randomly created network or reservoir. The seizures of two different types of generalized epilepsy are detected: absence epilepsy and tonic-clonic epilepsy. Spike Wave Discharges (SWDs) are the EEG patterns that occur when a patient with absence epilepsy is having a seizure. These absences last from several seconds to a few minutes and are generally very M. K¨oppen et al. (Eds.): ICONIP 2008, Part I, LNCS 5506, pp. 56–63, 2009. c  Springer-Verlag Berlin Heidelberg 2009