A Lightweight CNN-Attention-BiLSTM Architecture for Multi-Class Arrhythmia Classification on Standard and Wearable ECGs

2025-11-11

AI Summary

To address the clinical need for early and accurate arrhythmia detection, this paper proposes a lightweight, multi-lead-compatible deep learning model capable of jointly processing both 12-lead and single-lead ECG signals. Methodologically, the architecture integrates 1D convolutional layers for local time-frequency feature extraction, bidirectional LSTMs to capture long-range temporal dependencies, and a dual-channel–temporal attention mechanism to emphasize discriminative signal segments; a class-weighted loss function is further adopted to mitigate label imbalance. With only 0.945 million parameters, the model achieves an F1-score of 0.892 on the CPSC 2018 dataset—surpassing state-of-the-art baselines—while maintaining low inference latency and minimal memory footprint. The primary contributions are: (i) the first lightweight unified architecture enabling seamless single- and multi-lead ECG modeling; and (ii) a practical, edge-deployable solution that balances high accuracy, strong generalizability, and real-time capability for wearable cardiac monitoring.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Early and accurate detection of cardiac arrhythmias is vital for timely diagnosis and intervention. We propose a lightweight deep learning model combining 1D Convolutional Neural Networks (CNN), attention mechanisms, and Bidirectional Long Short-Term Memory (BiLSTM) for classifying arrhythmias from both 12-lead and single-lead ECGs. Evaluated on the CPSC 2018 dataset, the model addresses class imbalance using a class-weighted loss and demonstrates superior accuracy and F1- scores over baseline models. With only 0.945 million parameters, our model is well-suited for real-time deployment in wearable health monitoring systems.

Problem

• Classifying cardiac arrhythmias using lightweight deep learning architecture
• Addressing class imbalance in ECG data with weighted loss
• Enabling real-time arrhythmia detection on wearable ECG devices

Innovation

• Combines CNN attention and BiLSTM for ECG classification
• Uses class-weighted loss to handle data imbalance
• Lightweight model with under one million parameters