Feature Extraction of EEG Signals for Epileptic Seizure Prediction

Feature extraction of electroencephalography (EEG) signals is crucial for epileptic seizure prediction. In this study, nine scalp EEG temporal feature descriptors based on the positive zero-crossing interval length series and statistical analysis are proposed for seizure prediction. Experimental results have shown that all feature descriptors present statistical significance for discriminating pre-ictal and inter-ictal EEG epochs in a majority of subjects. Moreover, the combination of our approach and support vector machine achieves the best performance with the sensitivity 92.75%, specificity 69.77%, Kappa 16.30%, and accuracy 86.50%.