IC-U-Net: A U-Net-based Denoising Autoencoder Using Mixtures of Independent Components for Automatic EEG Artifact
Chun-Hsiang Chuang1, Kong-Yi Chang2, Chih-Sheng Huang3
1Research Center for Education and Mind Sciences, College of Education, National Tsing Hua University, Hsinchu, Taiwan; Institute of Information Systems and Applications, College of Electrical Engineering and Computer Science, National Tsing Hua University, Hsinchu, Taiwan; Department of Education and Learning Technology, National Tsing Hua University, Hsinchu, Taiwan.
A new deep learning model, IC-U-Net, effectively removes artifacts from electroencephalography (EEG) signals. This method reconstructs brain activity, improving the reliability of EEG data for brain-computer interfaces and mobile brain imaging.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Electroencephalography (EEG) signals are crucial for neuroscience research and brain-computer interfaces (BCIs).
- Signal artifacts, such as eye blinks and muscle activity, frequently contaminate EEG data, leading to misinterpretations and reduced BCI performance.
- Developing robust artifact removal techniques is essential for accurate neural signal analysis.
Purpose of the Study:
- To introduce IC-U-Net, a novel deep learning model based on the U-Net architecture.
- To demonstrate the model's capability in removing pervasive EEG artifacts and reconstructing clean brain signals.
- To provide an automated, end-to-end solution for EEG artifact removal.
Main Methods:
- Developed IC-U-Net utilizing the U-Net architecture for EEG artifact removal.
- Trained the model on mixed brain and non-brain components derived from independent component analysis.
- Employed an ensemble of loss functions to capture complex signal fluctuations in EEG recordings.
Main Results:
- Validated IC-U-Net's effectiveness in a simulation study and four real-world EEG experiments.
- Demonstrated successful removal of diverse artifacts including eye movements, muscle activity, and line noise.
- Showcased the model's ability to reconstruct multi-channel EEG signals with high fidelity.
Conclusions:
- IC-U-Net offers a promising end-to-end solution for automatic EEG artifact removal.
- The model is applicable to various artifact types and supports multi-channel signal reconstruction.
- IC-U-Net facilitates accurate brain dynamics imaging, particularly in mobile settings.
