Abstracts

Rationale: Epileptic seizures had been considered unpredictable and sudden events until a few years ago. Recent results in literature showed that these seizures are likely to be stage of an epileptogenic process rather than an unpredictable and sudden event, and a predictable preictal phase has been hypothesized to occur between interictal and ictal periods. Indeed different studies endorsed this new hypothesis. Most of the researches have been focused on intracranial EEG analysis, whereas just few studies were carried out on scalp EEG. Our aim was to study the predictability of different kinds of seizures recorded from scalp EEG using the Adaptive Threshold Seizure Warning Algorithm (ATSWA). This technique is based on the estimation of the Short Term Maximun Lyapunov Exponent (STLmax) and on entrainment between critical electrodes. Methods: ATSWA has been implemented in a specific programming language (Matlab®) and tested over scalp EEG recordings from 3 patients (A, B and C, males, aged 50, 41 and 43 years) with frontal lobe epilepsy and from 1 patient (D, male, 25 years) with juvenile absence epilepsy. The mean duration of the recordings was 40 min. for patients A, B and D, and it was 5 hours for patient C. Nine seizures were analyzed. The technique exploited the first seizure of each recording for training in order to predict the next ones. Five of the 9 analyzed seizures could be predicted (1 for patients A, B, D, and 2 for patient C). Finally, the trend of the convergence of STLmax profiles allowed for the automatic detection of the electrodes involved in the process leading to the seizure. Results: ATSWA succeeded in issuing a warning before every seizure, with a warning horizon of 5.0min. for patient A, 12.0 for B, 21.8 and 101.8 for the two seizures of patient C, and 7 min. for patient D. As for the selection of the critical electrodes for patients A, B and C, the technique automatically selected the frontal ones, in agreement with the analogical EEG analyses. Frontal electrodes were also automatically selected for the patient D whose EEG showed generalized (absence) seizures. Conclusions: ATSWA seems to be able to detect changes in the dynamics of scalp EEG in partial as well as in generalized seizures, and to infer information about the epileptogenic area. Further studies are required in order to quantify the performance of the technique over long recordings including many seizures.