An Automatic EEG Spike Detection Algorithm Using Morphological Filter

Abstract

Epileptic electroencephalogram data contains transient components and background activities. One of the transients is spike, which occurs randomly with short-duration. Spike detection in EEG is significant for clinical diagnosis of epilepsy. Since it is time consuming to scan spikes manually, an automatic spike detection method is necessary. In this paper, we introduce an automatic spike detection method in epileptic EEG based on morphological filter. Firstly, an average weighted combination of open-closing and close-opening morphological operator, which eliminates statistical deflection of amplitude, is utilized to extract spike component from epileptic EEG. Then, according to the characteristic of spike component, the structure elements are constructed with two parabolas, and a new criterion is put forward to optimize center amplitude and width of the structure elements. The proposed method is evaluated by simulated epileptic EEG data. Results show that background activity is fully restrained and spike component is well extracted. Finally, the method is applied to normal and epileptic EEG data which are actually recorded from nine testees. The average detection rate of spikes is 91.62% and no false detection for normal EEG signals