Multi agent systems for interpretation of EEG signals

Abstract

Electroencephalogram (EEG) has been the cheapest, most popular, and convenient brain imaging technique for a broad spectrum of applications. More importantly, EEG technology has influenced the future trends in cognitive systems in Artificial Intelligence. Devices to capture EEG signals range from those used in laboratory settings to wearable devices for personal usage such as entertainments, attention monitoring, meditation, and some clinical applications. The cost of personal scale EEG devices are increasingly becoming affordable. Most EEG devices are designed not only to capture the EEG signal but also to offer some basic, low-level interpretation of the captured EEG data. However, such interpretations are rather incomplete, unexplainable, and unreliable without feedbacks from an expert neurologist. Therefore, development of computer-based solutions to interpret EEG records has been a research challenge. We have conducted a research to build Multi Agent System solution, EEGMA, to bring the effect of neurologists’ interpretation for EEG signals. EEGMA reads EEG signal from an EEG headset and compute parameters, namely, most dominant frequency, continuity of frequencies, and all frequency distribution, eyeblink strength, and epoch size of an EEG session, and define four agents. These agents deliberate on parameters based on expert neurologist’s knowledge of interpretation of EEG signals. This process is analogues to neurologist or a group of neurologists deliberating on the above parameters to give better interpretation for an EEG session so that mental image of a person could be explained. EEGMA has been developed using SPADE platform to implement the MAS solution to analyze and interpret EEG signals which come from an EEG device. EEGMA has been evaluated by comparing its performance with neurologists’ interpretation of EEG signals. According to the results EEGMA has shown 70% accuracy in interpretation of EEG signals. EEGMA can be used by end users with some interest in EEG technology, EEG researchers, Neurologists, and developers of EEG-based solutions.