Real-time multi-channel seizure detection and analysis hardware

This paper presents a low power platform which performs continuous multi-channel detection and analysis of seizures for epilepsy patients. The detection unit, upon detecting a seizure, enables an analysis circuit that locally processes and transmits energy and frequency contents of the EEG data. Transmission traffic reduction of 256x is achieved by locally processing data and transmitting relevant information about the EEG rather than transmitting the raw EEG data itself. Multi-channel detection is accomplished by replicating an ultra-low power single-channel detection unit 16 times, each for a distinct EEG channel. These detection units pass their output into a multi-channel detection block that, when multiple channels flag a seizure, enables the analysis circuit. The proposed detection architecture removes false positives and reduces detection delays by as much as 16 sec when compared to the current method. The platform is implemented in 65 nm CMOS which contains 16 seizure detection modules with a seizure analysis unit and occupies 0.43 mm
²
. When simulated at 1.3 V, the seizure detection unit runs continuously and dissipates 0.04 μW at 256 Hz. The analysis unit consumes 0.36 μW at 1.85 KHz when powered on after a seizure detection.