Incorporating BSS to epileptic seizure predictability measure from scalp EEG

Epileptic seizure prediction has been explored by many researchers for decades. Most of the methods are based on the evaluation of the chaotic behavior of intracranial electroencephalographic (EEG) recordings. Here, a novel approach has been developed to predict the dynamical changes of the brain from the scalp EEG signals. Blind source separation (BSS) has been successfully used to separate the EEG signals into their constitute components including the seizure sources. Then the chaotic behavior was evaluated by measuring the Short-term Largest Lyapunov exponent (STLmax). The simultaneous intracranial and scalp EEG recordings were used to compare our approach with the traditional method using intracranial recordings. Similar prediction results were obtained from the scalp and intracranial recordings. Also different BSS algorithms were applied to compare their performance of source separation.