Ballistocardiogram artifact removal in simultaneous EEG-fMRI using generative adversarial network.

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Signal Processing

Background:

• Simultaneous electroencephalogram-functional magnetic resonance imaging (EEG-fMRI) offers high temporal and spatial resolution for brain science research.
• Ballistocardiogram (BCG) artifacts significantly contaminate EEG data during fMRI acquisition, posing a challenge for accurate analysis.
• Existing BCG artifact removal methods often require additional reference signals or complex hardware, limiting their practical application.

Purpose of the Study:

• To develop an effective and accessible method for removing BCG artifacts from EEG data acquired during simultaneous EEG-fMRI.
• To improve the performance of artifact removal by optimizing individual modules within a generative adversarial network (GAN).
• To enhance the local representation ability of the network model for more reliable BCG artifact removal.

Main Methods:

• A novel modular generative adversarial network (GAN) architecture was designed for BCG artifact removal.
• A specialized training strategy was implemented to optimize the parameters of each network module.
• The proposed method was evaluated based on its ability to remove artifacts while preserving essential EEG information, without requiring extra hardware or reference signals.

Main Results:

• The proposed modular GAN demonstrated superior performance in removing BCG artifacts compared to existing methods.
• The network effectively retained crucial EEG signal information, ensuring the integrity of the neural data.
• The method proved reliable and did not necessitate additional reference signals or complex equipment.

Conclusions:

• The developed modular GAN provides an effective solution for BCG artifact removal in simultaneous EEG-fMRI.
• This approach enhances the quality of EEG data, facilitating more accurate brain science research.
• The method's accessibility and effectiveness make it a valuable tool for the neuroimaging community.