Arrhythmia Classification Using Graph Neural Networks Based on Correlation Matrix

2024-10-14

IEEE International Conference on Bioinformatics and Biomedicine

AI Summary

This study addresses automatic arrhythmia classification from electrocardiogram (ECG) signals. We propose an end-to-end graph neural network (GNN) framework that constructs a physiology-informed graph structure directly from the Pearson correlation matrix of ECG features—used as the adjacency matrix—thereby eliminating handcrafted graph topology design. The method jointly learns time-frequency domain ECG representations, feature correlation patterns, and topological embeddings via GNNs within a unified architecture. Our key contribution is the first application of correlation-based adjacency matrices for interpretable, mechanism-guided graph construction in arrhythmia analysis, significantly enhancing model interpretability and generalizability. Evaluated on public benchmark datasets, the approach achieves precision and recall exceeding 50% across all arrhythmia classes, demonstrating both effectiveness and robustness. This work establishes a novel paradigm for physiology-driven, interpretable GNN modeling in cardiac signal analysis.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

With the advancements in graph neural network (GNN), there has been increasing interest in applying GNN to electrocardiogram (ECG) analysis. In this study, we generated an adjacency matrix using correlation matrix of extracted ECG features and applied a GNN to classify arrhythmias. Our work was compared with existing approaches from the literature. The results demonstrated that precision and recall for all arrhythmia classes exceeded 50%, suggesting that proposed method can be considered an approach for arrhythmia classification.

Problem

• Classify arrhythmias using graph neural networks
• Apply correlation matrix for ECG signal analysis
• Improve precision and recall in arrhythmia detection

Innovation

• Graph Neural Networks
• Correlation Matrix
• Arrhythmia Classification