Document Actions

You are here: Home / News / Research News / Automatically recognizing impending epileptic seizures

Automatically recognizing impending epileptic seizures

Scientists in Freiburg have developed new methods for the recognition and classification of epileptic seizures (May 2008).

Epileptic patients suffer from the sudden occurrence of seizures, which are brought on by the simultaneous discharge of a large number of nerve cells in the brain. Each seizure occurs out of the blue - when a storm of neural activity is seen to brew within the brain, the individual who is affected will not even notice it.

Scientists associated with Ralph Meier and Ad Aertsen, at the Bernstein Center for Computational Neuroscience of the University of Freiburg, have now developed a method with which brainwaves of patients can automatically be measured and evaluated simultaneously. Since changes in neural activity generally appear some seconds before the first external indication of a seizure, this method allows patients and clinic personnel to be forewarned of upcoming attacks. In the future, one additionally hopes to develop implants which may influence the brainwaves and serve to counteract the beginning phases of an attack. For such systems, a mandatory prerequisite is the recognition, in time, of these upcoming convulsions.


Activity at the beginning of an epileptic seizure to be observed in the EEG and at the specific sites of the different electrodes (pink-coloured spots).

The Freiburg procedure for the evaluation of data is based on electroencephalography (EEG). With the help of electrodes placed on the scalp, changes in the voltage of the brain are measured through the skull. In the event of an epileptic seizure – and depending on the particular type of convulsion – there may be intensified discharges in certain frequency ranges or unusual discharge patterns may occur. In an EEG from a healthy individual, oscillations also occur within different frequency ranges, which reflect specific conditions of the brain like when sleeping, dozing or in a state of excitement. The goal of the scientists from Freiburg is to reliably differentiate healthy oscillatory patterns from such discharges which are associated with epilepsy.

To date, with the help of mathematical algorithms, attempts have already been made to evaluate an EEG automatically. However, not every process is suitable for each form of convulsion. To guarantee an optimal recording of all types of seizures, the scientists associated with Meier have made use of various mathematical evaluation procedures in parallel. “Our method requires no individual adjustment, but is instead suitable for all types of seizures”, explains Meier.

Meier and his colleagues made use of an approximately 1400 hour, long-term EEG, demonstrating a total of 91 verified attacks, in order to examine the efficiency of this procedure. Almost all of these attacks were recognized in time by the procedure. Only about once every two hours did their system result in a faulty recording indicative of an attack, which was then not seen to subsequently occur. Consequently, the process reveals a better accuracy in recognition than the other currently available methods to date. In addition, the system could differentiate between the course of various types of seizures and thereby serves further to contribute to the diagnosis of epilepsy. “In principle, the programme is ready for clinical application, although there are still a couple of technical hurdles to be cleared involving the routine connections necessary for clinical data capture”, as stated by Meier.




Prof. Dr. Ad Aertsen

Detecting epileptic seizures in long-term human EEG: A new approach to automatic online and real-time detection and classification of polymorphic seizure patterns.

Bernstein Center for Computational Neuroscience
Albert-Ludwigs-Universität Freiburg
Schaenzlestrasse 1
79104 Freiburg i.Br.
Tel: +49 (761) 203 2786

Ralph Meier, Heike Dittrich, Andreas Schulze-Bonhage, Ad Aertsen (2008).  published online May 8, 2008