Scholarlink

<p>Cardiovascular diseases are a significant health issue that needs to be diagnosed at early stages to avoid mortality. The electrocardiogram (ECG) signal is frequently employed in medical heart condition diagnosis. Extracting information from ECG signals is an intricate and computationally demanding issue. However, manual detection of CVD is a time-consuming and difficult task for early diagnosis. To overcome this challenge, a transfer learning framework has been broadly applied to ECG signals for detecting CVD. In this research, a novel CVD-TransNet is introduced for classifying CVD using ECG signals. Initially, the ECG signals are denoised with Discrete Wavelet Transform (DWT) to reduce the noisy distortions. In the proposed CVD-TransNet, three pre-trained neural networks are concatenated to achieve highly accurate results via a transfer learning method. This transfer learning method is utilized to retrieve the deep features and concatenate these features for further classification. Finally, the Dilated Convolutional Neural Network (Di-CNN) is employed for classifying cardiac diseases into normal, heart attack, and arrhythmia. The competence of the proposed CVD-TransNet was measured with network metrics like specificity, precision, recall, accuracy, and F1-score. The experimental outcomes exposed that the proposed CVD-TransNet attains a high accuracy of 99.62% for the detection of CVD in its initial stages. The proposed Transfer learning-based CVD-TransNet model progresses the accuracy by 2.39%, 3.93%, 1.31%, and 0.06% better than Deep CNN, ANN, SVM-FFBPNN, and SGDM-TL Deep CNN respectively.</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

CVD- TransNet: Cardiovascular Disease Classification Using Transfer Learning Based Convolutional Neural Networks with Electrocardiogram Signal

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • J. Priyadharshini,
  • M. Madheswaran

摘要

Cardiovascular diseases are a significant health issue that needs to be diagnosed at early stages to avoid mortality. The electrocardiogram (ECG) signal is frequently employed in medical heart condition diagnosis. Extracting information from ECG signals is an intricate and computationally demanding issue. However, manual detection of CVD is a time-consuming and difficult task for early diagnosis. To overcome this challenge, a transfer learning framework has been broadly applied to ECG signals for detecting CVD. In this research, a novel CVD-TransNet is introduced for classifying CVD using ECG signals. Initially, the ECG signals are denoised with Discrete Wavelet Transform (DWT) to reduce the noisy distortions. In the proposed CVD-TransNet, three pre-trained neural networks are concatenated to achieve highly accurate results via a transfer learning method. This transfer learning method is utilized to retrieve the deep features and concatenate these features for further classification. Finally, the Dilated Convolutional Neural Network (Di-CNN) is employed for classifying cardiac diseases into normal, heart attack, and arrhythmia. The competence of the proposed CVD-TransNet was measured with network metrics like specificity, precision, recall, accuracy, and F1-score. The experimental outcomes exposed that the proposed CVD-TransNet attains a high accuracy of 99.62% for the detection of CVD in its initial stages. The proposed Transfer learning-based CVD-TransNet model progresses the accuracy by 2.39%, 3.93%, 1.31%, and 0.06% better than Deep CNN, ANN, SVM-FFBPNN, and SGDM-TL Deep CNN respectively.