Implementation of a Convolutional Neural Network for Eye Blink Artifacts Removal From the Electroencephalography Signal

Marcin Jurczak; Marcin Kołodziej; Andrzej Majkowski

doi:10.3389/fnins.2022.782367

February 11, 2022Frontiers in NeuroscienceOpen Access

Implementation of a Convolutional Neural Network for Eye Blink Artifacts Removal From the Electroencephalography Signal

View Full Paper

Ask AI

Bookmark

Share

Why the study?

Eye blink artifacts spread across the head surface and disrupt EEG signal analysis, prompting implementation of a convolutional neural network to eliminate them.

Key Points

To implement a convolutional neural network combined with a signal augmentation method to effectively eliminate eye blink artifacts from electroencephalography data.
Trained a convolutional neural network (CNN) utilizing a proposed signal augmentation framework for electroencephalography (EEG) recordings.
Evaluated and compared CNN denoising performance against standard independent component analysis (ICA) and regression techniques across both synthetic and real EEG signals.
The CNN-based approach achieved superior artifact elimination compared to conventional ICA and regression methods on both simulated and real recordings.
Performance improvements were most pronounced at electrode sites situated across the central region of the head.

Structured PICO

P

Population

20 people (women and men, 19-25 years old) performing an N-back task for 60 minutes, plus generated artificial EEG/EOG signals.

I

Intervention

Convolutional neural network (CNN) for removing eye blink artifacts

C

Comparator

Independent component analysis (ICA) and linear regression (REG)

O

Outcome

Performance in removing eye-blink artifacts

A convolutional neural network outperforms traditional independent component analysis and regression methods for removing eye blink artifacts from EEG signals.

Limitations

A large number of teaching examples are needed to train the CNN, but the number of recorded EEG signal examples is often too small

Implication

May enhance EEG preprocessing accuracy; leaves open clinical translation pending prospective validation.

Abstract

Electroencephalography (EEG) signals are disrupted by technical and physiological artifacts. One of the most common artifacts is the natural activity that results from the movement of the eyes and the blinking of the subject. Eye blink artifacts (EB) spread across the entire head surface and make EEG signal analysis difficult. Methods for the elimination of electrooculography (EOG) artifacts, such as independent component analysis (ICA) and regression, are known. The aim of this article was to implement the convolutional neural network (CNN) to eliminate eye blink artifacts. To train the CNN, a method for augmenting EEG signals was proposed. The results obtained from the CNN were compared with the results of the ICA and regression methods for the generated and real EEG signals. The results obtained indicate a much better performance of the CNN in the task of removing eye-blink artifacts, in particular for the electrodes located in the central part of the head.