Real-time detection of EEG spikes using neural networks

Abstract

A digital signal processor based microcomputer system is designed for real-time detection of EEG spikes. The system acquires 16 channel EEGs, displays them in 5 second epochs, detects EEG spikes, identifres them on the screen and stores their times of occurrence au in real-time. Spike detection is achieved by a two-level neural network system analyzing 100 msec of multichannel EEG in a sliding window. 1 n the first level, spikes are Ukntified in individual EEG channels by 16 identical neural network modules computed in the digital signal processor. In the second level, outputs of the first level modules are integrated by a second neural network module for thefinal detection. Results show that neural network based EEG spike detection systems can be implemented for real-time clinical operation using current digital signal processor technology.