EOG artifact correction from EEG recording using stationary subspace analysis and empirical mode decomposition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Ocular artifacts are a common and significant problem in electroencephalography (EEG) data acquisition.
• These artifacts can interfere with the accurate diagnosis of neurobiological events.
• Existing methods for artifact removal may not be effective under all recording conditions.

Purpose of the Study:

• To propose and validate an effective approach for removing ocular artifacts from raw EEG recordings.
• To improve the accuracy of neurobiological event diagnosis by cleaning EEG data.
• To address challenges posed by limited electrode usage and non-stationary, non-independent artifact sources.

Main Methods:

• Blind source separation using stationary subspace analysis to isolate artifactual components.
• Adaptive signal decomposition with empirical mode decomposition (EMD) to denoise artifact components.
• Subtraction of artifact-only components from the original EEG signal to obtain clean data.

Main Results:

• The proposed method effectively removes ocular artifacts from both simulated and real EEG data.
• Demonstrated effectiveness even with a limited number of electrodes.
• Successfully handled highly non-stationary artifact-contaminated signals where source independence is not assumed.

Conclusions:

• The developed method provides a robust solution for ocular artifact removal in EEG.
• This technique enhances the reliability of EEG data for diagnosing neurobiological events.
• The approach is particularly valuable for challenging recording scenarios, improving diagnostic accuracy.