Detection Of Epileptic Spike Wave In EEG Signals Based On Morphological Component Analysis

ObjectiveEpilepsy is the result of central neurons group sudden excessive repetitive firing. It has serious harm to the chronic diseases of the human nervous system, so epilepsy accurate analysis has important practical significance. According to statistics, about80%of epilepsy patients have abnormal EEG Performance following epilepsy characteristic wave:sporadic spikes, sporadic spikes, spike and wave or sharp slow-wave integrated, multi-spike wave group. This paper used Morphological Component Analysis (MCA), which based on appropriate selected dictionaries, to detect the spike wave in Electroencephalogram (EEG) signals.Methods1. Experimental dataThis paper, experimental data from http://www.eeg.p1/epi/, selected11cases which are already diagnosed epilepsy patients (three of them have had surgery). Each case selected at least1length of9seconds EEG signals.33lengths were selected.2. Detection of epileptic spike wave in EEG signals based on Morphological Component Analysis(1)Select the appropriate dictionary:it is a critical step in MCA to select the appropriate dictionary. Based on the characteristics of the epileptic EEG background signal and the spike wave signal, the selection has been pre-set function as a dictionary, optional dictionary is set out as follows:①db4wavelet transform as the extraction of epileptic EEG spikes signal dictionary and Gabor transform as the extraction of epileptic EEG background signal dictionary.②Curvelet transform as the extraction of epileptic EEG spikes signal dictionary and Gabor transform as the extraction of epileptic EEG background signal dictionary.③db4wavelet transform as the extraction of epileptic EEG spikes signal dictionary and DCT as the extraction of epileptic EEG background signal dictionary.④Curvelet transform as the extraction of epileptic EEG spikes signal dictionary and DCT as the extraction of epileptic EEG background signal dictionary. (2) extract different forms of epileptic EEG spikes based on Morphological component analysis:According to the appropriate dictionary, it can extract different forms of spike wave signal and the background signal.3. Detection of epileptic spike wave in EEG signals based on wavelet transformIn order to ensure the effectiveness of epileptic EEG spike detection method based on MCA, selected wavelet transform to detect epileptic EEG spike. In this paper, selected db4wave and Mexihat wave which are similar with spike wave to extract the epileptic EEG spike as the primarily wave of wavelet transform.Results1. Detection of epileptic spike wave in EEG signals based on Morphological Component Analysis(1) Db4wavelet transform as the extraction of epileptic EEG spikes signal dictionary and Gabor transform as the extraction of epileptic EEG background signal dictionary: the detection rate is80.00%, the rate of accuracy is85.43%, the rate of omission is20.00%, the rate of false is13.63%.(2) Curvelet transform as the extraction of epileptic EEG spikes signal dictionary and Gabor transform as the extraction of epileptic EEG background signal dictionary:the detection rate is88.18%, the rate of accuracy is86.61%, the rate of omission is10.91%, the rate of false is13.63%.(3) Db4wavelet transform as the extraction of epileptic EEG spikes signal dictionary and DCT as the extraction of epileptic EEG background signal dictionary:the detection rate is89.09%, the rate of accuracy is90.71%, the rate of omission is10.91%, the rate of false is9.09%.(4) Curvelet transform as the extraction of epileptic EEG spikes signal dictionary and DCT as the extraction of epileptic EEG background signal dictionary:the detection rate is95.45%, the rate of accuracy is94.59%, the rate of omission is4.55%, the rate of false is5.45%.2. Detection of epileptic spike wave in EEG signals based on wavelet transform(1)Db4wavelet transform:the detection rate is81.82%, the rate of accuracy is81.08%, the rate of omission is18.18%, the rate of false is19.09%. (2)Mexihat wavelet transform:the detection rate is82.72%, the rate of accuracy is77.78%, the rate of omission is17.27%, the rate of false is23.64%.Conclusions:1. The detection of epileptic spike wave in EEG signals based on Morphological Component Analysis has better results when the DCT is selected as the extraction of epileptic EEG background signal dictionary, the Curvelet transform is selected as the extraction of epileptic EEG spikes signal dictionary.2. Compared with the wavelet transform, The detection of epileptic spike wave in EEG signals based on Morphological Component Analysis can be more effective to detect the spike wave from the epileptic EEG signal.