Automated real-time detection of epileptic seizures

Epilepsy is one of the most common neurological disorders and is characterized by recurrent spontaneous seizures, often monitored using electroencephalography (EEG) or intracranial electrocorticography (iEEG). The
ability to effectively detect seizures in real-time could potentially reduce the disabling unpredictability of seizures, by issuing a warning to the patient and/or its caregivers. Furthermore, automatic activation of therapies (pharmacological, electrical, optical or thermal stimulation) upon detection of seizures may significantly advance the treatment of epilepsy and potentially benefit patients suffering from medication side effects or refractory epilepsy.We propose an automated system for real-time detection of seizures tested in three
different animal models of epilepsy (genetic, status epilepticus-induced and albumin-induced).Cortical activity was continuously recorded through two epidural electrodes connected to an implanted telemetry transmitter. The acquired signals were filtered, segmented and underwent extraction of features, to be classified by an artificial neural network (ANN). For classifier training, a dataset of seizure and non-seizure recordings was comprised and represented by 22 extracted features. Forward selection analysis led to the identification of a
5 feature subset, allowing optimal trade-off between robust ANN classification and reduced computational time. A graphical user interface was created for simple execution of both real-time and off-line data analysis and seizure detection. System performance was assessed by analyzing over 2800 hours of raw iEEG recordings from 15 animals. Performance evaluation revealed overall sensitivity and positive predictive value above 98% in unedited signals containing noise, artifacts and interictal discharges.