Automated Detection of Neonatal EEG Asymmetry

Accurate automatic detection of neonatal brain abnormalities has been of great interest for diagnosis and prognosis. This is mainly due to the fact that current manual processes involved in identifying such abnormalities are time intensive and require skilled clinicians. Among different abnormalities in newborn EEG signals, the presence of asymmetry is of great importance as studies have shown strong correlation between this abnormality and adverse neurodevelopmental outcomes in childhood. Existing techniques for automated detection of asymmetry in EEG signals are based on the Fouriertransform (FT) which, as proven in this research, are only sensitive to the frequency contents of the signal under analysis and not to the way they may vary in time. As newborn EEG signals are known to be non-stationary, FT-based techniques may lead to non-accurate results when used to measure asymmetry in such signals. As an alternative, in this research a new technique is proposed which is based on time-frequency analysis of EEG signals. In this technique, time-frequency correlation is used as a measure for asymmetry. The proposed methodology is tested on both simulated and real newborn EEG signals and its performance is compared with that of the existing techniques. Results show that proposedmethodis more sensitive to asymmetry in both real and simulated EEG signals. Also it can be used for automatic detection of asymmetry abnormality in the Neonatal intensive care unit.