Artifacts Removal From Epileptic EEG Signal Based on Independent Components Analysis Method

Electroencephalography (EEG) signal recording cannot be secluded from noise, affecting the measured signal for such critical application as epilepsy diagnosis. Often, EEG signal corrupt with unintended artefacts, including physiological (Electromyography (EMG), Electrocardiography (ECG), and Electrooculography (EOG)) and non-physiological ones (electrode skin contact and cable movement). During epilepsy diagnosis, the neurologist fails to determine the seizure type adequately or to expect seizure phases duration due to the existence of the contaminated EEG signal during the seizure. The main challenge in this paper is to decline the wrong decisions handled by the specialist, prove high-quality measurement of the EEG signal, and extract only relevant information from the EEG signal. Independent Component Analysis (ICA) method for a real epileptic EEG dataset will be investigated for artefact identification and removal. Thirteen epileptic patients were included in this analysis with diverse gender, generations, time measurements, seizure types, and artefacts. Moreover, to prove the efficiency of the developed approach and the quality measurement of acquired EEG signal, an accentuated validation step in terms of Signal to Noise Ratio (SNR) demonstrates the impact of the ICA on the recorded epileptic EEG signals, mainly for the channels Fp1 and Fp2 with an SNR value achieved equal to 14.7 dB and 15.5 dB respectively.