Scholarlink

<p>Cardiovascular diseases (CVDs) are a major public health concern. Electrocardiograms (ECG) are very often used as a tool for CVD diagnosis. Cardiologists scrutinize ECG, searching for signs for diagnosis. Faced with a growing number of patients, computer-aided analysis of ECGs has been suggested as a second opinion to help cardiologists in their analysis. This work develops an algorithm for ECG computerized analysis. In the method in this paper, we first filter the ECG signal. The filtered signal then undergoes the Hilbert Huang Transform (HHT) on one hand, and both Fuzzy Entropy (FE) and Hurst Exponent (HE), on the other hand. The results of the latter step are fed into a dual-channel (DC) SqueezeNet convolutional neural network (CNN) that classifies the original ECG signal into either atrial fibrillation, atrial flutter, normal sinus, other rhythms, or noisy recording. The tool’s performance thus developed, was assessed using the 2017 PhysioNet/Computing in Cardiology (CinC) dataset and the MIT-BIH atrial fibrillation Database (MIT-BIH AFDB). Experimental results showed that the proposed algorithm yielded an F1 score of 98.1% for atrial fibrillation, 99.5% for normal sinus rhythms, 98.64% for other rhythms, and 97.38% for noisy recording when using the 2017-CinC dataset. As for the MIT-BIH AFDB dataset, we got an F1 score of 98.65% for atrial fibrillation, 97.2% for atrial flutter, and 99.68% for normal sinus rhythm. These results are encouraging and demonstrate the effectiveness of the computerized analysis tool for ECG signals. It can successfully be used as a second opinion and therefore help cardiologists.</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

Automated ECG Analysis with Dual-Channel SqueezeNet Using Hilbert Huang Transform, Fuzzy Entropy, and Hurst Exponent

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • Richel T. Nguimdo,
  • Alain Tiedeu

摘要

Cardiovascular diseases (CVDs) are a major public health concern. Electrocardiograms (ECG) are very often used as a tool for CVD diagnosis. Cardiologists scrutinize ECG, searching for signs for diagnosis. Faced with a growing number of patients, computer-aided analysis of ECGs has been suggested as a second opinion to help cardiologists in their analysis. This work develops an algorithm for ECG computerized analysis. In the method in this paper, we first filter the ECG signal. The filtered signal then undergoes the Hilbert Huang Transform (HHT) on one hand, and both Fuzzy Entropy (FE) and Hurst Exponent (HE), on the other hand. The results of the latter step are fed into a dual-channel (DC) SqueezeNet convolutional neural network (CNN) that classifies the original ECG signal into either atrial fibrillation, atrial flutter, normal sinus, other rhythms, or noisy recording. The tool’s performance thus developed, was assessed using the 2017 PhysioNet/Computing in Cardiology (CinC) dataset and the MIT-BIH atrial fibrillation Database (MIT-BIH AFDB). Experimental results showed that the proposed algorithm yielded an F1 score of 98.1% for atrial fibrillation, 99.5% for normal sinus rhythms, 98.64% for other rhythms, and 97.38% for noisy recording when using the 2017-CinC dataset. As for the MIT-BIH AFDB dataset, we got an F1 score of 98.65% for atrial fibrillation, 97.2% for atrial flutter, and 99.68% for normal sinus rhythm. These results are encouraging and demonstrate the effectiveness of the computerized analysis tool for ECG signals. It can successfully be used as a second opinion and therefore help cardiologists.