Background: EEG data are susceptible to electrooculographic (EOG) and electromyography (EMG) artifacts (i.e., jaw clenching, teeth squeezing and forehead movements). Due to their non-stationary nature, these artifacts greatly obscure the information and power spectrum of EEG signals. Current EOG and EMG artifact correction methods are too time-consuming when applied to low-density EEG and have been focusing on offline processing or handling one single type of EEG artifact. A software-only real-time method for correcting multiple types of EEG artifacts of high-density EEG remains a significant challenge.
Methods: We demonstrate a novel approach for automatic online real-time EEG artifact correction of EOG and EMG artifacts. The method was tested on three healthy subjects using 63 EEG channels (Brain Products GmbH) and a sampling rate of 1,000 Hz. Captured EEG signals were imported in MATLAB with the labstreaminglayer interface allowing buffering of EEG data. EMG artifacts were detected by channel variance and adaptive thresholding and corrected by using channel interpolation. Real-time independent component analysis (ICA) was applied for correcting EOG artifacts.
Results: Our results demonstrate that the algorithm effectively reduces EMG artifacts, such as jaw clenching, teeth squeezing and forehead movements, and EOG artifacts (horizontal and vertical eye movements). The average computing time to correct EOG and EMG artifacts from 500 ms of 63 EEG channel data is ∼365 ms depending on the convergence of ICA and the type and intensity of the artifacts.
Conclusion: The study proposed a Matlab-based automatic online algorithm for correct- ing EOG and EMG artifacts of high-density EEG signals at high sampling rates in real-time, while preserving neuronal brain activity information. Its performance can improve EEG- neurofeedback during fMRI measurements and may be suitable for other BCI’s.