Reconstructing EOG From EEG Timeseries: A Spatial Filtering Approach.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Unobtrusive mental state monitoring relies on neurophysiological signals like electroencephalography (EEG) and electrooculography (EOG).
• Current wearable systems for EEG+EOG recording can be inconvenient, necessitating streamlined sensor configurations.
• Integrating EOG recording into frontal-only EEG montages is desirable for practical applications.

Purpose of the Study:

• To propose and evaluate a spatial filtering approach for extracting EOG signals from a reduced set of frontal EEG electrodes.
• To investigate the efficacy of linear least squares (LLS) and Least Absolute Shrinkage and Selection Operator (LASSO) methods for EOG signal estimation.
• To demonstrate the potential of these methods to replace actual EOG signals in automated oculographic event detection.

Main Methods:

• A spatial filtering approach was developed to extract EOG signals from frontal electroencephalography (EEG) electrodes.
• Two distinct schemes, linear least squares (LLS) and LASSO, were examined within a common signal analytic framework.
• The methods were validated using experimental data to assess their ability to estimate EOG activity from EEG signals.

Main Results:

• Both LLS and LASSO data-driven techniques reliably estimated EOG activity from EEG signals using minimal training data.
• The LASSO-based technique demonstrated generalizability across subjects.
• Empirical evidence suggests the estimated EOG signals can substitute for actual EOG signals in detecting oculographic events like blinks and saccades.

Conclusions:

• A spatial filtering method effectively extracts EOG signals from frontal EEG electrodes, enhancing wearable system convenience.
• Data-driven techniques like LLS and LASSO provide reliable EOG estimation from EEG.
• This approach facilitates unobtrusive mental state monitoring and automated oculographic event detection.