Non-linear models in focus localization, seizure detection and prediction

Jonas Henriksen (Invited author)

Research output: Contribution to conferencePaperResearchpeer-review


One of the most devastating problems for epilepsy patients is the unpredictable nature of seizures. Not knowing when or where a seizure occurs has severe consequences in social interaction, ability to work, driving a car, go swimming etc. Traditionally the patient and the doctor work together to reduce the amount of seizures, but if only the seizures could be predicted many patients would live a better life even with the same amount of seizures. Most of the research in this area is based on continuous EEG-recordings. Doing the right analysis on the EEG-signal is essential. In the presentation I will introduce different methods developed in collaboration between DTU and Rigshospitalet for seizure prediction and detection. Prediction of seizure onset with nonlinear signal processing has shown promising results in recent years. The goal is to make an attention-device that gives a message when a seizure is approaching. The primary obstacle is the lack of sufficient large databases to make a patient-specific algorithm rather than a “one-size-fits-all” approach. At Rigshospitalet, a research project is carried out that aims at collecting enough data to be able to do this. The next couple of years will probably show whether we should put our hopes up for a seizure prediction algorithm. Seizure detection is a more established field of nonlinear EEG-analysis. Some kinds of epilepsy give none or few clinical symptoms and are thus difficult for the treating physician, relatives, and the patient him- or herself to know when and how often there is seizure activity in the brain. It is therefore interesting to make an objective and automatic detection of the quantity of seizure activity, which is not reliable on competence or fatigue by the epileptologist. With the best algorithm it has been possible to obtain a 100 % detection rate with false alarms amounting to merely 0.02 per hour. While these numbers only apply to seizures of a certain type, a general algorithm has achieved an 83 % detection rate with the same number of false alarms per hour.
Original languageEnglish
Publication date2009
Publication statusPublished - 2009
EventDanish Epilepsy Society & the Danish Society of Clinical Neurophysiology Annual Meeting - Odense, Denmark
Duration: 1 Jan 2009 → …


ConferenceDanish Epilepsy Society & the Danish Society of Clinical Neurophysiology Annual Meeting
CityOdense, Denmark
Period01/01/2009 → …


Dive into the research topics of 'Non-linear models in focus localization, seizure detection and prediction'. Together they form a unique fingerprint.

Cite this