Scholarlink

<p>Cardiovascular diseases (CVDs) cause global mortality, highlighting the need for effective strategies. Artificial intelligence (AI)-based applications have been used to diagnose myocardial infarction (MI) using automatic ECG signals. This study mainly introduces an effective MI diagnostic technique that incorporates advanced techniques. Signal processing, denoising, and analysis are performed sequentially to extract relevant features from electrocardiogram (ECG) signals. Clinicians can gain insight by understanding the patterns of these features. The relation of features is represented in a graph-based dataset. A novel compact attentional graph neural network (CA-GNN) model has been developed to classify cardiovascular diseases affected patients. The proposed model achieves an accuracy of 98.32% in the PTB dataset, indicating its robustness in diagnosing MI. However, to show the effectiveness and generalizability, the model was further experimented on two other datasets: MIT-BIH and PTB-XL; and achieved accuracies of 96.89%, 92.74%, respectively. This extension highlights the advantages of the proposed framework for diagnosing different types of cardiovascular disease abnormalities. In addition, this compact framework accelerates the training process, reduces the computation cost, and improves efficiency. This study also investigated a semisupervised autoencoder model to assess the performance of unlabeled data. The model obtains an accuracy of 89.43% on unlabeled PTB datasets and minimizes the loss to 0.095. The outcome of this study conveys an effective automated computer diagnostic system in cardiovascular classification abnormalities and paves the way for future endeavors to practice the classification of heart diseases from ECG signals.</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 dual-stage framework for cardiovascular abnormalities diagnosis from ECG signals using CA-GNN and semi-supervised autoencoders

  • 期刊论文

  • 发表日期：

    2025年12月19日

查看全文
  • Nur Mohammad Fahad,
  • Md. Abdur Rahman,
  • Md. Jakarea,
  • Mirjam Jonkman,
  • Sami Azam

摘要

Cardiovascular diseases (CVDs) cause global mortality, highlighting the need for effective strategies. Artificial intelligence (AI)-based applications have been used to diagnose myocardial infarction (MI) using automatic ECG signals. This study mainly introduces an effective MI diagnostic technique that incorporates advanced techniques. Signal processing, denoising, and analysis are performed sequentially to extract relevant features from electrocardiogram (ECG) signals. Clinicians can gain insight by understanding the patterns of these features. The relation of features is represented in a graph-based dataset. A novel compact attentional graph neural network (CA-GNN) model has been developed to classify cardiovascular diseases affected patients. The proposed model achieves an accuracy of 98.32% in the PTB dataset, indicating its robustness in diagnosing MI. However, to show the effectiveness and generalizability, the model was further experimented on two other datasets: MIT-BIH and PTB-XL; and achieved accuracies of 96.89%, 92.74%, respectively. This extension highlights the advantages of the proposed framework for diagnosing different types of cardiovascular disease abnormalities. In addition, this compact framework accelerates the training process, reduces the computation cost, and improves efficiency. This study also investigated a semisupervised autoencoder model to assess the performance of unlabeled data. The model obtains an accuracy of 89.43% on unlabeled PTB datasets and minimizes the loss to 0.095. The outcome of this study conveys an effective automated computer diagnostic system in cardiovascular classification abnormalities and paves the way for future endeavors to practice the classification of heart diseases from ECG signals.