Abstracts

RATIONALE: The non-linear time series analysis of EEG data has been shown to provide information about he spatiotemporal dynamics of epileptogenic network prior to the seizure onset. Coarse-grained entropy rates (CER - Palus, Physica D 3,1996) provide a characterisation equivalent to that of Kolgomorov -Sinai entropy, or the sum of positive Lyapunov exponent. Coarse-grained information rates (CIR) quantify common information or dependence of two signals and are able to identify the "direction of information flow" i.e.,to distinguish driven from driving systems. The aim of our study was to used CER and CIR in candidates for epilepsy surgery in order to identify changes in spatiotemporal EEG dynamics and synchronisation which precede the seizure onset. METHODS: The intracranial (subdural) and scalp EEG data of three presurgical candidates were acquired for analysis using Stellate Monitor software or Schwarzer s Brainlab. The 24hours video/EEG recordings were performed on a 32 channel BMSI or Schwarzer system . Our study is based on analysis of 3.5 hours of ictal and interictal EEG recording. RESULTS: We have found significant decreases of CER in preictal subdural EEG . In addition coarse grained information rates show that preictal increase of synchronisation or dependence of EEG signals from different channels is better pronounced than decrease od dynamical complexity of individual signals. CONCLUSIONS: The "direction of information flow" established by the conditional information rates shows the spatiotemporal dynamic interaction between the epileptogenic and other brain regions. Supported by grant IGA MZ CR NF 6258