Feasibility of seizure risk prediction using intracranial EEG measurements in dogs

Patients with refractory epilepsy would greatly benefit from an accurate seizure forecasting system. This paper introduces a seizure prediction algorithm based on a random forest classifier that uses features computed from continuous intracranial electroencephalographic (iEEG) measurements in dogs with naturally occurring epilepsy. Results suggest that the proposed model can distinguish between interictal (baseline) and preictal (pre-seizure) periods and provide an intuitive measure of seizure risk that may have practical utility.