Model-based Bayesian filtering of cardiac contaminants from biomedical recordings

Area of Science:

• Biomedical Signal Processing
• Computational Neuroscience
• Cardiovascular Physiology

Background:

• Electrocardiogram (ECG) and magnetocardiogram (MCG) signals are significant sources of noise in other biomedical recordings.
• Existing denoising methods often struggle to effectively remove complex cardiac artifacts.
• Bayesian filtering has shown promise for ECG signal denoising.

Purpose of the Study:

• To evaluate the efficacy of a Bayesian filtering framework for removing cardiac contaminants from diverse biomedical signals.
• To demonstrate the framework's utility beyond ECG denoising, including its application to electroencephalogram (EEG) and electromyogram (EMG) signals.
• To assess the method's capability in canceling maternal cardiac signals from fetal ECG/MCG recordings.

Main Methods:

• Application of a proposed Bayesian filtering framework.
• Testing on simulated and real-world biomedical signal datasets.
• Inclusion of electrocardiogram (ECG), magnetocardiogram (MCG), and ballistocardiographic (BCG) artifacts.

Main Results:

• The Bayesian filtering framework effectively removes cardiac contaminants like ECG, MCG, and BCG artifacts.
• Successful denoising was achieved across multiple biomedical recordings, including EEG and EMG.
• Maternal cardiac signal cancellation from fetal ECG/MCG was demonstrated.

Conclusions:

• The Bayesian filtering framework offers a versatile and effective solution for removing cardiac noise from various biomedical signals.
• This approach significantly improves the quality of recordings such as EEG, EMG, and fetal ECG/MCG.
• The method shows broad applicability in biomedical signal processing for artifact reduction.