Seizure detection with single-channel EEG using Extreme Learning Machine

The electroencephalography (EEG) is the most essential tool for the diagnosis and the treatment of the epilepsy. It allows observing events strongly associated with epilepsy or epileptic spikes and locating the brain regions that cause the symptoms of epilepsy. This paper presents an automated classification of EEG signals for the detection of epileptic seizures with Single-Channel using the wavelet transform and the Extreme Learning Machine. The aim is to create a system with reduced computation time and resources with the minimum number of required electrodes. The decision making process is comprised of three steps: (a) Preprocessing, (b) feature extraction based on the wavelet transform, and (c) classification by the Extreme Learning Machine. The proposed algorithm has been tested on three different data sets from the CHB-MIT scalp EEG database using only the FT10-T8 channel. The proposed method achieves a classification accuracy of 94.85%.