Scholarlink

<p>Malignant cardiac arrhythmias, such as ventricular fibrillation (VF) and ventricular tachycardia (VT), are major contributors to sudden cardiac death (SCD). Accurate differentiation between shockable and non-shockable arrhythmias plays a critical role in timely and effective therapeutic intervention, including defibrillation. This study presents a novel Deep Multi-Scale Convolutional Neural Network (DMS-CNN) framework designed for automated classification of arrhythmias from electrocardiogram (ECG) signals. The proposed approach employs wavelet-based filtering for noise suppression and multi-scale signal decomposition, followed by deep convolutional feature extraction and classification. The model was trained and evaluated using two publicly available datasets, namely the Creighton University Ventricular Tachyarrhythmia Database (CUDB) and the MIT-BIH Malignant Ventricular Arrhythmia Database (VFDB). Experimental results demonstrate that the proposed DMS-CNN achieved a classification accuracy of 98.95% using 2-second ECG segments, outperforming traditional classifiers such as LS-SVM, standard CNN, and multi-threshold-based decision models in terms of sensitivity and specificity. These findings validate the effectiveness of integrating wavelet-based preprocessing with multi-scale deep learning to enhance arrhythmia classification performance and ensure robust detection across variable ECG patterns.</p>

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

A Novel Deep Multi-scale Convolutional Neural Network for Shockable and Non-shockable Arrhythmia Classification

  • 期刊论文

  • 发表日期：

    2025年11月6日

查看全文
  • P. Anitha,
  • R. Sethumadhavi,
  • Sheik Yousuf Tharvaj,
  • Supraja Eduru,
  • T. Y. Satheesha,
  • Ganta Bhagyalakshmi

摘要

Malignant cardiac arrhythmias, such as ventricular fibrillation (VF) and ventricular tachycardia (VT), are major contributors to sudden cardiac death (SCD). Accurate differentiation between shockable and non-shockable arrhythmias plays a critical role in timely and effective therapeutic intervention, including defibrillation. This study presents a novel Deep Multi-Scale Convolutional Neural Network (DMS-CNN) framework designed for automated classification of arrhythmias from electrocardiogram (ECG) signals. The proposed approach employs wavelet-based filtering for noise suppression and multi-scale signal decomposition, followed by deep convolutional feature extraction and classification. The model was trained and evaluated using two publicly available datasets, namely the Creighton University Ventricular Tachyarrhythmia Database (CUDB) and the MIT-BIH Malignant Ventricular Arrhythmia Database (VFDB). Experimental results demonstrate that the proposed DMS-CNN achieved a classification accuracy of 98.95% using 2-second ECG segments, outperforming traditional classifiers such as LS-SVM, standard CNN, and multi-threshold-based decision models in terms of sensitivity and specificity. These findings validate the effectiveness of integrating wavelet-based preprocessing with multi-scale deep learning to enhance arrhythmia classification performance and ensure robust detection across variable ECG patterns.