Automatic EEG eye movement artifacts removal using Independent Component Analysis

Abstract

Background. Eye movement artifacts contained in EEG recordings hamper a lot the automatic processing and analysis of EEG signal. Therefore, the removal of such artifacts is important stage for any further signal processing. There are artifacts removal methods based on using wavelet transformation, regression analysis in the time and frequency domain, Principal component analysis and Independent component analysis. Methods. The novel method of automatic EEG eye movement artifacts removal based on Independent Component Analysis was proposed. The method utilizes the TDSEP algorithm for blind source separation. Own criteria for artifact components detection were used. The method was implemented with the Python programming language and tested on EEG signals recorded from two healthy volunteers. Results. Comparison of the effectiveness of the method was conducted with the participation of two experts. They were asked to review the EEG fragments before and after artifacts removal and evaluate the quality of artifacts removal. The average value of assessing the quality of artifacts removal was 4.83 for TDSEP based algorithm and 4.58 for FastICA based algorithm. Conclusion. The proposed method is more effective then method based on FastICA algorithm and using it for automatic EEG eye movement artifacts removal is expedient.