Removal of BCG artifacts using a non-Kirchhoffian overcomplete representation

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Signal Processing

Background:

• Simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) are powerful tools for studying brain activity.
• Ballistocardiogram (BCG) artifacts, arising from the heart's mechanical activity, contaminate EEG signals acquired during MRI.
• Accurate removal of BCG artifacts is crucial for preserving neural signals and enabling reliable single-trial analysis.

Discussion:

• A novel nonlinear unmixing approach is proposed to extract BCG artifacts from simultaneous EEG-fMRI data.
• This method utilizes an overcomplete basis identified from EEG data acquired with a custom multipath electrode cap.
• The approach infers non-Kirchhoffian latent variables, distinguishing BCG artifacts from neural activity, which is strictly Kirchhoffian.

Key Insights:

• The proposed method effectively removes BCG artifacts without attenuating neural signals essential for classification.
• Comparison with conventional methods like independent component analysis and optimal basis sets demonstrates superior performance.
• The nonlinear unmixing approach leads to enhanced single-trial classification accuracy in auditory oddball experiments.

Outlook:

• This technique offers a promising avenue for improving the quality of EEG data acquired during simultaneous fMRI.
• Further validation in diverse experimental paradigms and patient populations is warranted.
• The method has the potential to advance our understanding of neural dynamics and trial-to-trial variability in complex cognitive tasks.