Abstracts

RATIONALE: Non-linear analysis of the EEG can detect pre-ictal changes several minutes before seizure onset in mesial temporal lobe epilepsy (MTLE). Previous studies were carried out on EEG periods that preceded seizure onset for less than one hour. Little is known concerning the modifications of the EEG dynamics during longer time expanses, and correlations with electro-clinical state. Here we evaluate the specificity and sensibility of our EEG analysis method. METHODS: We analyzed long epochs of intracranial or scalp EEG, ranging from 4 to 24 hours continuously, for 3 patients, for a total of 120 hours. Data were obtained from EEG-video recording sessions of patients suffering from intractable MTLE. As a non-linear indicator of dynamics, we applied the similarity measure (NeuroReport, 1999; 10, 2149-2155), that compares the signal dynamics of successive tested windows to that of a 'reference' period, for each of the 27-32 electrodes. In a parallel way, we inspected EEG and video to allow a correlation between the dynamical changes and electrical (slow waves, spikes...) or/and clinical changes. We also evaluated the reproducibility of the method by studying long EEG periods containing or not a seizure during different days of recording in the same patient. RESULTS: For long periods distant from seizures during stable physiological state (awake or sleeping) with moderate inter-ictal epileptic activity, changes of the similarity index are below the threshold of statistical signification (5?). Only transitions between physiological states or major enhancement of inter-ictal epileptic activity can generate sustained significant deviations of the similarity index. Use of different 'reference' states improves the specificity of our method. The intra-individual reproducibility appears to be good with a possible correlation between the importance of the ictal EEG abnormalities and the ability to anticipate the seizure. CONCLUSIONS: Sensitivity of the method allows to faithfully and synthetically reflect the EEG dynamics. Specificity of the method confers potential for current clinical use in order to anticipate seizure onset in real time condition.