Topographic distribution of seizure patterns in patients with absence epilepsy

Background: Absence seizures, which are also known as petit mal seizures, are the most common type of seizures in pediatric epilepsy. They appear in several types of epilepsy and are characterized by impaired consciousness and 3-Hz spike-and-slow-wave complexes in the electroencephalogram (EEG). The treatment with anti-epileptic drugs (AEDs) is the result of a very delicate weighting, which leads to a trade-off between the side effects the drugs are causing and the disapperance of the seizures. the company Hypo-Safe A/S is currently developing a device, which hopefully will permit to reduce the number of EEG examinations needed to achieve the optimal medication. Objective: The project has two biomedical signal processing objectives: seizure on- set detection and automatic topographic seizure distribution description by means of statistical measures. Seizure onset detection is directly relevant for the user of the sub- cutaneously implanted Hypo-Safe EEG apparatus and for medical monitoring purposes. Automatic topographic distribution description by means of statistical measures is relevant for decision concerning placement of the apparatus. Methods: An absence seizure detection algorithm based on fractal dimension estimation was designed, implemented and tested together with a topographic evaluation of absence seizure patterns. Results: Excluding patients with symptomatic epilepsy it was possible to achieve a SE of 97% and a FDR of 0.15 FP/h on channel F4-F8. Similar performance could also be achieved in a few neighboring channels. Therefore for the other patients this area represents a very good location for placing the Hypo-Safe subcutaneous electrode. In patients with symptomatic epilepsy it is still possible to find a good location, but they must be assessed individually and the optimal position will change from patient to patient. Significance: This is the first study which evaluates the topographic distribution of absence seizure patterns using an appositely designed seizure detection algorithm