EEG_GLT-Net: Optimising EEG Graphs for Real-time Motor Imagery Signals Classification

TL;DR

The paper addresses real-time EEG motor imagery classification by optimizing channel relationships through an EEG Graph Lottery Ticket approach, which prunes adjacency edges to produce subject- and model-specific graphs. The EEG_GLT method, built on spectral graph convolution, outperforms traditional Geodesic and PCC adjacency constructions, achieving higher accuracy (mean improvement around 13.39% over PCC) and substantial reductions in MACs (up to 97%). The study shows that adjacency matrix design can outweigh network architecture in determining performance and demonstrates practical viability for low-latency EEG_MI decoding. These findings offer a path to more efficient, subject-tailored BCI systems capable of real-time operation.

Abstract

Brain-Computer Interfaces connect the brain to external control devices, necessitating the accurate translation of brain signals such as from electroencephalography (EEG) into executable commands. Graph Neural Networks (GCN) have been increasingly applied for classifying EEG Motor Imagery signals, primarily because they incorporates the spatial relationships among EEG channels, resulting in improved accuracy over traditional convolutional methods. Recent advances by GCNs-Net in real-time EEG MI signal classification utilised Pearson Coefficient Correlation (PCC) for constructing adjacency matrices, yielding significant results on the PhysioNet dataset. Our paper introduces the EEG Graph Lottery Ticket (EEG_GLT) algorithm, an innovative technique for constructing adjacency matrices for EEG channels. It does not require pre-existing knowledge of inter-channel relationships, and it can be tailored to suit both individual subjects and GCN model architectures. Our findings demonstrated that the PCC method outperformed the Geodesic approach by 9.65% in mean accuracy, while our EEG_GLT matrix consistently exceeded the performance of the PCC method by a mean accuracy of 13.39%. Also, we found that the construction of the adjacency matrix significantly influenced accuracy, to a greater extent than GCN model configurations. A basic GCN configuration utilising our EEG_GLT matrix exceeded the performance of even the most complex GCN setup with a PCC matrix in average accuracy. Our EEG_GLT method also reduced MACs by up to 97% compared to the PCC method, while maintaining or enhancing accuracy. In conclusion, the EEG_GLT algorithm marks a breakthrough in the development of optimal adjacency matrices, effectively boosting both computational accuracy and efficiency, making it well-suited for real-time classification of EEG MI signals that demand intensive computational resources.

Figures (9)

• Figure 1: Our model: (a) Overall architecture (classifying EEG MI of one time point $\frac{1}{160}s$ of signals from 64 EEG electrodes). Note that EEG Graph adjacency matrix can be $A^{Geodesic}$, $A^{PCC}$ or $A^{EEG\_GLT}$, (b) Components inside the spectral graph convolution block, (c) Chebyshev spectral graph convolution
• Figure 2: EEG graph ($m_g$) pruning using Algorithm 1: At each $N_{ep}$ iteration, the bottom $p_g$% are pruned, reducing density from 100% until the lowest density $s_g$%. Solid lines indicate remaining edges, while red-dashed lines depict removed edges
• Figure 3: Geodesic Distance Adjacency Matrix ($A^{Geodesic}$)
• Figure 4: PCC Adjacency Matrix ($A^{PCC}$) of Subject $S_6$ and $S_{14}$
• Figure 5: Representations of $m_{g\_EEG\_GLT}$ for Subject $S_6$ at 13.39% Density. (a) Adjacency Matrix - Model A (Accuracy: 78.13%) (b) Graph - Model A (c) Adjacency Matrix - Model E (Accuracy: 73.55%) (d) Graph - Model E