A low-power "near-threshold" epileptic seizure detection processor with multiple algorithm programmability

Abstract

In this paper, we developed and implemented a low-power multiple-algorithm processor for detection of epileptic seizures with high efficacy. Four different algorithms --The Coastline, Hjorth parameter, Energy and Non-linear energy metric based algorithms --are used in our analysis. The Boolean logical combinations of these algorithms allow programmability according to the patient specific needs and improve efficacy by almost 10%. The system implemented in 65-nm TSMC technology consumes about 13.1 W (measured) at scaled supply voltage of 400mV. Programmability using Boolean logic helps in improving the efficacy resulting in almost an error-free detection.