Online detection and removal of eye blink artifacts from electroencephalogram. (August 2021)

• Record Type: Journal Article
• Authors: Egambaram, Ashvaany; Badruddin, Nasreen; Asirvadam, Vijanth S; Begum, Tahamina; Fauvet, Eric; Stolz, Christophe
• Is Part Of: Biomedical signal processing and control. Volume 69(2021)
• Publication Date: 2021-08
• Subjects: Canonical Correlation Analysis (CCA) -- Electroencephalogram (EEG) -- Eye blink artifact -- Modified Empirical Mode Decomposition (FastEMD); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102887
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: More than 97% removal accuracy and an average of 10–13 ms removal speed. Impressive reconstruction with CC value of artifact-free segments approaches 1. Insignificant distortion to artifact-free segments with RMSE, and η dB close to 0. Feasible for online removal of eyeblink artifacts with low neural distortion. Abstract: The most prominent type of artifact contaminating electroencephalogram (EEG) signals are the eye blink (EB) artifacts, which could potentially lead to misinterpretation of the EEG signal. Online identification and elimination of eye blink artifacts are crucial in applications such a Brain-Computer Interfaces (BCI), neurofeedback, and epilepsy diagnosis. In this paper, algorithms that combine unsupervised eye blink artifact detection (eADA) with modified Empirical Mode Decomposition (FastEMD) and Canonical Correlation Analysis (CCA) are proposed, i.e., FastEMD-CCA 2 and FastCCA, to automatically identify eye blink artifacts and remove them in an online setting. The average accuracy, sensitivity, specificity, and error rate for eye blink artifact removal with FastEMD-CCA 2 is 97.9%, 97.65%, 99.22%, and 2.1%, respectively, validated on a Hitachi dataset with 60 EEG signals, consisting of more than 5600 eye blink artifacts. FastCCA achieved an average of 99.47%, 99.44%, 99.74%, and 0.53% artifact removal accuracy, sensitivity, specificity, and error rate, respectively, validated on the Hitachi dataset too. FastEMD-CCA 2 and FastCCA algorithms are developed and implemented in the C++ programming language, mainly to investigate the processing speed that these algorithms could achieve in a different medium. Analysis has shown that FastEMD-CCA 2 and FastCCA took about 10.7 and 12.7 ms, respectively, on average to clean a 1-s length of EEG segment. As a result, they're a viable option for applications that require online removal of eye blink objects from EEG signals.