Assessment of Seizure Prediction Based on Synchronization Changes in the EEG Dynamics: Statistical Significance Versus Clinical Relevance
Abstract number :
1.051
Submission category :
Clinical Neurophysiology-Computer Analysis of EEG
Year :
2006
Submission ID :
6185
Source :
www.aesnet.org
Presentation date :
12/1/2006 12:00:00 AM
Published date :
Nov 30, 2006, 06:00 AM
Authors :
1,2Bjoern Schelter, 1,2Matthias Winterhalder, 1,2Hinnerk Feldwisch genannt Drentrup, 1,3Johannes Wohlmuth, 1,3Jakob Nawrath, 3Armin Brandt, 1,2Jens Timmer, and 2,3Andr
A reliable seizure prediction would offer new therapeutic strategies for epilepsy. An abnormal synchronization of neurons leading to epileptic seizures suggests multivariate time series analysis techniques detecting synchronized dynamics as a promising approach for seizure prediction when applied to invasive EEG recordings of epilepsy patients. The proper assessment of these techniques is a prerequisite for a seizure prediction device. Apart from statistical significance the clinical relevance is investigated., A quantity measuring phase synchronization (Mormann et al., Physica D 2000; 144:358-369) has been applied to an invasive EEG data pool of 14 patients with long term intracranial continuous EEG data of more than 100 days duration. The seizure prediction performance has been assessed for long intervention times of up to 4 hours and occurrence periods of up to 2 hours. The obtained results are compared to the performance of an unspecific random predictor (Schelter et al., Chaos 2006; 16: 013108). Moreover, additional measures are used to quantify the strain of especially false predictions on patients. Thus, the fraction of false predictions with respect to the total number of predictions as well as the time a patient is awaiting a seizure that will never occur are investigated to quantify the clinical relevance of the prediction techniques., For several patients a prediction performance superior to a random predictor could be shown. Sensitivities of up to 75% were observed when allowing 2 false predictions per day. To achieve this promising result the patient is awaiting seizures for 15%, i.e., 3-4 hours every day that will never occur. This fraction is rather low compared to other published results, which is the benefit of subdividing the prediction horizon in intervention time and occurrence period. A fraction of false alarms with respect to the total number of alarms of more than 50% on average was observed., Preictal changes in the synchronization of the EEG dynamics may offer a chance for epileptic seizure prediction. Besides statistical significance other parameters have to be taken into account quantifying the clinical relevance. Especially avoiding false predictions will reduce the strain on patients and, thereby, lead to a much higher clinical relevance., (Supported by the German Federal Ministry of Education and Research (BMBF grant 01GQ0420) and the German Science Foundation (Ti 315/2-2; Sonderforschungsbereich-TR3).)
Neurophysiology