A Channel-Projection Mixed-Scale Convolutional Neural Network for Motor Imagery EEG Decoding.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Motor imagery electroencephalography (EEG) decoding is crucial for brain-computer interfaces (BCIs).
• Existing deep learning methods using decomposed EEG spectrums can miss vital spatial and temporal information, leading to suboptimal decoding performance.
• This limits the effectiveness of BCIs for motor-disabled patients.

Purpose of the Study:

• To propose an end-to-end EEG decoding framework that utilizes raw multi-channel EEG signals as input.
• To enhance motor imagery decoding accuracy by employing a novel channel-projection mixed-scale convolutional neural network (CP-MixedNet) and amplitude-perturbation data augmentation.
• To provide a more effective BCI solution for motor-disabled individuals.

Main Methods:

• Developed CP-MixedNet, an end-to-end deep learning framework processing raw multi-channel EEG.
• The framework includes blocks for learning spatial-temporal representations, capturing mixed-scale temporal information, and classifying EEG tasks.
• Amplitude-perturbation data augmentation was used to further improve performance.

Main Results:

• The proposed CP-MixedNet framework demonstrated improved decoding accuracy on two public EEG datasets (BCI competition IV 2a and High gamma dataset).
• The method effectively captures essential spatial dependencies and multi-scale temporal information from raw EEG signals.
• Experimental results showed competitive performance compared to state-of-the-art methods.

Conclusions:

• The proposed CP-MixedNet framework offers a promising solution for enhancing motor imagery decoding in BCIs.
• Utilizing raw EEG and advanced deep learning architectures can overcome limitations of previous methods.
• This approach has the potential to significantly improve BCI applications for motor-disabled patients.