Evolutionary coherence on EEG signals for epileptic seizure detection

Abstract

Electroencephalogram (EEG) signal for epileptic seizure is nonstationary by nature. The onset of epileptic seizure is determined by the increase in synchronicity of firing neurons, and the spreading of epileptic seizure could be traced with investigating on the evolution of synchronicity across channels. However, there are only a few previous studies on utilizing evolutionary coherence in detecting epileptic seizure EEG events. Besides that, these researches also mostly focus on only a few channels for mere simple and quick comparison. There is also a lack of research in comparing coherence analysis from different non-parametric approaches. Therefore, this research aims to analyze the brain connectivity in EEG epileptic seizure using nonstationary coherence by applying specifically SLEX coherence, wavelet coherence and STFT coherence. The algorithm is tested on a real epileptic seizure patient with focal epilepsy seizure at the left temporal lobe. The coherence obtained is further plotted using Circos software package, which is advantageous in mapping complex links and relationships. In conclusion, evolutionary coherence on EEG signals for epileptic seizure detection has been performed using STFT, wavelet and SLEX coherence. It was found that wavelet and SLEX coherence are capable of epileptogenic focus localization and seizure prediction, with wavelet coherence showing slightly better performance.