Scholarlink

The life style of human being is prone to disease like heart strokes. Classification of cardiac arrhythmias at early stages has become highly essential. From electrocardiogram (ECG) signals we can identify the symptoms and timely intervention and treatment becomes easy for the patient. In this work, a novel deep learning architecture MSARTNet (Multi-Scale Attention-based Residual Transformer Network) is proposed for interpretable arrhythmia classification. This study combines three different features such as multi-scale convolutional layers, attention mechanisms, and transformer encoders to capture both local morphological and long-range temporal dependencies from ECG signals. MIT-BIH Arrhythmia Dataset is used in this work. MSARTNet able to gains satisfactory performance with an accuracy of 98.3% and F1-score of 0.975.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-Scale Attention-based Residual Transformer Network (MSARTNet) for ECG-based Arrhythmia Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jay Sureshchandra Raval,
  • V. N. Kamalesh,
  • Raj Kumar Patra

摘要

The life style of human being is prone to disease like heart strokes. Classification of cardiac arrhythmias at early stages has become highly essential. From electrocardiogram (ECG) signals we can identify the symptoms and timely intervention and treatment becomes easy for the patient. In this work, a novel deep learning architecture MSARTNet (Multi-Scale Attention-based Residual Transformer Network) is proposed for interpretable arrhythmia classification. This study combines three different features such as multi-scale convolutional layers, attention mechanisms, and transformer encoders to capture both local morphological and long-range temporal dependencies from ECG signals. MIT-BIH Arrhythmia Dataset is used in this work. MSARTNet able to gains satisfactory performance with an accuracy of 98.3% and F1-score of 0.975.