ML-TGNet: A Multi-Level Topology Guidance Network for Motor Imagery Decoding.

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) are crucial for neural rehabilitation, often using motor imagery electroencephalogram (MI-EEG) signals.
• Current BCIs frequently overlook brain dynamics, focusing on complex spatio-temporal features, which can lead to redundant information and reduced decoding performance.

Purpose of the Study:

• To develop a novel network, the multi-level topology-guidance network (ML-TGNet), that integrates topological brain synchronization information for enhanced MI-EEG feature extraction.
• To improve the decoding performance of BCIs by effectively capturing brain dynamics.

Main Methods:

• Designed ML-TGNet, a network incorporating a multi-level topology guidance module, a feature pool module, and a multi-branch decoding module.
• Leveraged topological brain synchronization information to guide feature extraction for motor imagery tasks.
• Validated the model on three public MI datasets: BCI Competition IV-2a, High Gamma, and OpenBMI.

Main Results:

• ML-TGNet achieved high classification accuracies: 82.33% on BCI Competition IV-2a, 96.42% on High Gamma, and 85.26% on OpenBMI.
• The proposed method outperformed existing state-of-the-art models in MI-EEG decoding.
• Demonstrated the effectiveness of incorporating brain synchronization information into deep learning models for EEG-based brain state decoding.

Conclusions:

• The study confirms the efficacy of using brain synchronization information to guide motor imagery decoding in BCIs.
• ML-TGNet offers a novel approach for EEG-based brain state decoding by integrating brain dynamics into deep learning architectures.
• This research opens new avenues for advancing neural rehabilitation technologies through improved BCI performance.