A low-power implantable event-based seizure detection algorithm

Closed-loop neurostimulation has shown great promise as an alternate therapy for over 30% of the epileptic patient population that remain non-responsive to other forms of treatment. We present an event-based seizure detection algorithm that can be implemented in real-time using low power digital CMOS circuits to form an implantable epilepsy prosthesis. Seizures are detected by classifying and marking out dasiaeventspsila in the recorded local field potential data and measuring the inter-event-intervals (IEI). The circuit implementation can be programmed post-implantation to custom fit the thresholds for detection. Hippocampal depth electrode recordings are used to validate the efficacy of a designed hardware prototype and thresholds are tuned to produce less than 5% false positives from recorded data.