An independent component analysis-based approach on ballistocardiogram artifact removing

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) acquisition combines high temporal and spatial resolution for brain activity investigation.
• MRI scanning introduces artifacts into EEG data, particularly the ballistocardiogram (BCG) artifact from cardiac activity, which is challenging to remove.
• Existing artifact removal strategies are limited by the unpredictable nature of BCG artifacts and the stochastic behavior of common Independent Component Analysis (ICA) algorithms.

Purpose of the Study:

• To address the challenge of reliably removing ballistocardiogram (BCG) artifacts from electroencephalography (EEG) signals acquired during simultaneous EEG-fMRI.
• To enhance the robustness and reliability of artifact removal using Independent Component Analysis (ICA) by mitigating its inherent stochastic behavior.

Main Methods:

• Utilized the FastICA algorithm, running it multiple times with varied initial conditions on simultaneous EEG-fMRI data.
• Applied clustering analysis in the component signal space to assess the reliability of the independent components identified by FastICA.
• Developed a novel approach for evaluating the consistency and trustworthiness of artifact-related components derived from ICA.

Main Results:

• The proposed method, involving repeated FastICA runs and subsequent clustering, offers a more reliable way to identify and assess artifactual components, specifically BCG artifacts.
• This approach helps overcome the stochastic nature of standard ICA, leading to more consistent results in artifact detection and removal.
• Demonstrated that clustering component signal spaces provides a quantitative measure for evaluating the reliability of estimated independent sources.

Conclusions:

• Repeated application of FastICA combined with signal space clustering provides a robust method for assessing the reliability of independent components, crucial for artifact removal in simultaneous EEG-fMRI.
• This technique enhances the utility of ICA for cleaning EEG data contaminated by challenging artifacts like BCG, improving the quality of combined EEG-fMRI analyses.
• The findings suggest a significant advancement in artifact management for multimodal neuroimaging, paving the way for more accurate investigations of brain function.