Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ECGMamba: Towards Efficient ECG Classification with BiSSM

Yupeng Qiang, Xunde Dong, Xiuling Liu, Yang Yang, Yihai Fang, Jianhong Dou

TL;DR

ECGMamba tackles the inefficiency of transformer-based ECG classification for long sequences by employing a bidirectional state-space model (BiSSM) within a Mamba-based block. The architecture blends an ECG encoder with a BiSSM-powered Mamba layer, augmented by residual connections, LayerNorm, and a two-layer Conv1D FFN to balance accuracy and inference speed. Experimental results on PTB-XL and CPSC2018 show competitive AUC, F1, and accuracy while substantially reducing computation compared with strong baselines, highlighting improved efficiency for long ECG sequences. The work offers a new research direction for efficient ECG signal analysis with potential clinical impact, while recognizing the need for broader validation and enhanced interpretability.

Abstract

Electrocardiogram (ECG) signal analysis represents a pivotal technique in the diagnosis of cardiovascular diseases. Although transformer-based models have made significant progress in ECG classification, they exhibit inefficiencies in the inference phase. The issue is primarily attributable to the secondary computational complexity of Transformer's self-attention mechanism. particularly when processing lengthy sequences. To address this issue, we propose a novel model, ECGMamba, which employs a bidirectional state-space model (BiSSM) to enhance classification efficiency. ECGMamba is based on the innovative Mamba-based block, which incorporates a range of time series modeling techniques to enhance performance while maintaining the efficiency of inference. The experimental results on two publicly available ECG datasets demonstrate that ECGMamba effectively balances the effectiveness and efficiency of classification, achieving competitive performance. This study not only contributes to the body of knowledge in the field of ECG classification but also provides a new research path for efficient and accurate ECG signal analysis. This is of guiding significance for the development of diagnostic models for cardiovascular diseases.

ECGMamba: Towards Efficient ECG Classification with BiSSM

TL;DR

ECGMamba tackles the inefficiency of transformer-based ECG classification for long sequences by employing a bidirectional state-space model (BiSSM) within a Mamba-based block. The architecture blends an ECG encoder with a BiSSM-powered Mamba layer, augmented by residual connections, LayerNorm, and a two-layer Conv1D FFN to balance accuracy and inference speed. Experimental results on PTB-XL and CPSC2018 show competitive AUC, F1, and accuracy while substantially reducing computation compared with strong baselines, highlighting improved efficiency for long ECG sequences. The work offers a new research direction for efficient ECG signal analysis with potential clinical impact, while recognizing the need for broader validation and enhanced interpretability.

Abstract

Electrocardiogram (ECG) signal analysis represents a pivotal technique in the diagnosis of cardiovascular diseases. Although transformer-based models have made significant progress in ECG classification, they exhibit inefficiencies in the inference phase. The issue is primarily attributable to the secondary computational complexity of Transformer's self-attention mechanism. particularly when processing lengthy sequences. To address this issue, we propose a novel model, ECGMamba, which employs a bidirectional state-space model (BiSSM) to enhance classification efficiency. ECGMamba is based on the innovative Mamba-based block, which incorporates a range of time series modeling techniques to enhance performance while maintaining the efficiency of inference. The experimental results on two publicly available ECG datasets demonstrate that ECGMamba effectively balances the effectiveness and efficiency of classification, achieving competitive performance. This study not only contributes to the body of knowledge in the field of ECG classification but also provides a new research path for efficient and accurate ECG signal analysis. This is of guiding significance for the development of diagnostic models for cardiovascular diseases.
Paper Structure (27 sections, 5 equations, 2 figures, 4 tables, 1 algorithm)

This paper contains 27 sections, 5 equations, 2 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep models for ECG classification
  4. State Space Models
  5. Method
  6. Basic Knowledge
  7. ECG Encoder
  8. Mamba Layer
  9. Mamba-based Block
  10. FFN
  11. Memory-Efficiency
  12. Computation-Efficiency
  13. Experiments and Results
  14. Basic Setting
  15. Datasets
  16. ...and 12 more sections

Figures (2)

  • Figure 1: Overall structure and detailed components of ECGMamba. FFN: Feed-Forward Network; Layer Norm: Layer Normalization; Conv1d: 1D convolution; BN: Batch Normalization; ReLU: Rectified Linear Unit; BiSSM: bidirectional SSM.
  • Figure 2: t-SNE visualization of ECGMamba.