A Multifrequency Brain Network-Based Deep Learning Framework for Motor Imagery Decoding.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Motor imagery (MI) decoding is crucial for brain-computer interfaces (BCIs) and stroke rehabilitation.
• Accurate decoding of electroencephalography (EEG) signals for MI is a significant research focus.

Purpose of the Study:

• To propose a novel multifrequency brain network-based deep learning framework for enhanced motor imagery decoding.
• To improve the accuracy of classifying different MI tasks from EEG signals.

Main Methods:

• Constructing a multifrequency brain network from multichannel MI-related EEG signals, with each layer representing a specific brain frequency band.
• Utilizing the Filter Bank Common Spatial Pattern (FBCSP) algorithm for spatial domain filtering and feature extraction.
• Designing a multilayer convolutional neural network (CNN) to extract and exploit topological information within the multifrequency brain network.

Main Results:

• Achieved state-of-the-art results on the BCI competition IV dataset 2a (83.83% accuracy, 0.784 kappa).
• Obtained high accuracy on the BCI competition III dataset IIIa (89.45% accuracy, 0.859 kappa).
• Demonstrated effective classification of different MI tasks from multichannel EEG signals.

Conclusions:

• The proposed multifrequency brain network-based deep learning framework effectively decodes motor imagery tasks from EEG signals.
• The framework shows significant potential for applications in stroke patient neural system remodeling and rehabilitation training.