An energy-based detection algorithm of epileptic seizures in EEG records.

ABSTRACT

A simple algorithm to automatically detect segments with epileptic seizures in long EEG records has been developed. The main advantages of the proposed method are the simple algorithm used and the lower computational cost. The algorithm measures the energy of each EEG channel by a sliding window and calculates some features of each patient signal to detect the epileptic seizure. It is also able to distinguish between seizures and noise artifacts. Nine invasive EEG records acquired by Epilepsy Center of the University Hospital of Freiburg were analyzed in this work. In 90 segments studied (39 with epileptic seizures) the sensitivity obtained with the method is 87.18 %. The algorithm is appropriate to detect epileptic seizures, with high sensitivity, in long EEG records to decrease the time used by physicians and specialists in visual inspections.