An Ultra Low Power System for Personalized, Wearable Seizure Detection

Page, Adam ;   Mohsenin, Tinoosh ;   Oates, Tim ;   Hopp, Jennifer

In this work we explore the use of a variety of machine learning classifiers for designing an ultra-low power, personalized seizure detection system using multi-channel, scalp EEG. The key objective is to design a low-power system capable of running locally at the sensor-side while still achieving high detection performance. All feature and classifier pairs were able to obtain F1 scores over 80% with 100% onset sensitivity when tested on 10 subjects. Among the classifiers explored, logistic regression (LR) proved to be the best solution for the application of personalized seizure detection. Both FPGA and ASIC implementations of the system with LR are presented and deliver the smallest area and power footprint.