Automatic EEG eyeblink artefact identification and removal technique using independent component analysis in combination with support vector machines and denoising autoencoder

Abstract

This study proposes a novel combination of independent component analysis (ICA) in conjunction with support vector machine (SVM) and denoising autoencoder (DA), for the first time, for removal of eyeblink artefacts from the corrupted electroencephalography (EEG). At first the eyeblink corrupted EEG signals are decomposed into independent components (ICs) using ICA, the corrupted-ICs are then identified using SVM as a classifier. From the corrupted-ICs, the artefacted segment is identified with a second SVM classifier and corrected by the pre-trained DA. Finally, inverse-ICA operation is applied on the remaining ICs and the corrected ICs to obtain the artefact-free EEG signal. The proposed methodology modifies only the portion corrupted with artefacts, and does not alter the uncorrupted part, thereby preserving the neural information in the original EEG. The proposed methodology was implemented to remove eyeblinks from the EEG data collected from the publicly available EEGLab data set. The results reveal that the proposed methodology is superior to the other recently reported methods in terms of the mutual information and average correlation coefficient. Further, the proposed method is automatic and does not require any intervention of the operator, whereas the other methods require intervention of the user.

Artifact Detection and Correction in EEG data: A Review

“…The more recent entries in Table I imply a growing popularity of deep learning techniques comes at the expense of traditional approaches and expert knowledge. However, we note that recent papers successfully drew on the rich history and knowledge developed within the EEG preprocessing community to build hybrid approaches that synthesize deep learning, ICA frameworks [12], or features borrowed from EEG prognostication [11]. We believe that hybrid frameworks are an interesting future direction of work in this domain and uniquely situated to combine the strengths of multiple approaches that will advance the current state-of-the-art.…”

“…Gilbert et al trained several classifiers (LDA, SVM, KNN) to distinguish between signal and noise independent components [4], and as previously mentioned, [9] trained a CNN classifier to distinguish between noise and signal components. Notably, these methods involve some global loss of information when the signal is reconstructed [12].…”

“…Phadikar et al suggested a hybrid model that uses SVMs to detect noise components in the ICA deconstructed signals and a denoising autoencoder to remove artifacts from the ICA components rather than the raw EEG [12]. By denoising the ICA components, instead of excluding them from the reconstruction all together, the reconstruction was found to be more accurate.…”

“…The F1 score and accuracy are appropriate for the assessment of tasks with balanced outcome class labels, which is not common in artifact annotation settings; a Artifact correction algorithms are more challenging to assess compared to detection algorithms as (barring simulated data) the ground truth is unknown. When artifacts are simulated, and access to the artifact free waveform is available, metrics such as normalized mean square error (NMSE) and root mean square error (RMSE) are used [12], [13]. When the data is not simulated the same metrics are calculated using artifact free EEG data collected under similar circumstances (i.e.…”

Automated Feature Extraction on AsMap for Emotion Classification Using EEG

“…Various research works have been conducted to remove artifacts from EEG signals [ 26 ]. Recently, automatic artifact removal techniques have gained much popularity [ 27 , 28 ]. After removal of artifacts, the most important task is feature extraction.…”