Skip to Main Content
This paper presents a set of four features to be used in the detection of seizure in the electroencephalograms (EEGs) of newborns. The features are designed with the aid of recent advances in modelling of the newborn EEG. The performance of the features is analysed with a database of 500 epochs of newborn EEG (250 background/250 seizure). The covariance of the features is also analysed to indicate the redundancy of the feature set. The results show significant differences in the features between seizure and background EEG. The covariance between the features suggests that there is little redundant information between the features.