Automatic Detection of Epileptic Spikes in the Long Term Electroencephalogram Using Wavelet Transform

We describe a new strategy to automatically detect the epileptiform activities (IEDs) in the long term 18 channel human electroencephalogram (EEG). Our scheme for detecting epileptic spikes in the EEG is based on a multi resolution, multi-level analysis, which is fast and delivers satisfactory results. The signal on each channel is decomposed into six sub -bands using discrete wavelet transform and for detecting spikes, a two level analysis is then performed on these sub- bands which falls in the frequency range of 4-8 Hz and 8-16 Hz. We processed 18 channels of EEG data, which covered most of the areas of brain, thereby proving valuable localizing information clinically necessary for classification of epilepsies. One of the major advantages of the proposed scheme is that the threshold for different scales are computed adaptively to suit to different epileptic patients. In this method, seizure and non- seizure activities were selected from 22 patient's EEGs in consensus among experts. The results are clinically evaluated by the experts at the R Madhavan Nayar Center for comprehensive epilepsy research centre, thiruvanathapuram (SCTIMST) and we got the system accuracy as 90.5%. The proposed algorithm was implemented in MATLAB.