Scholarlink

<p>The present study proposes a new approach to predict the outcome of external defibrillation in out-of-hospital cardiac arrest (OHCA) patients using sophisticated nonlinear feature extraction and classification via a probabilistic neural network (PNN) of electrocardiogram (ECG) signals. We analyzed 251 samples of ECG data collected from 195 unsuccessful and 56 successful resuscitations in OHCA patients. Our proposed methodology utilized the following five advanced nonlinear dynamics features: Sample Entropy, Approximate Entropy, Detrended Fluctuation Analysis, Correlation Dimension, and Largest Lyapunov Exponent. These features were selected such that the complex dynamics of ECG signals are not missed by traditional linear analyses and subtle patterns. Then, classification has been performed using the extracted features with a PNN model, which gave great performances with a training accuracy of 97.16% and a testing accuracy of 93.33%. With very good discriminative ability, the model resulted in a high area under the curve (AUC) of 0.98 and 0.96 for both the training and testing sets, respectively. This constitutes a significant advance in the ECG-based prediction of the outcome of defibrillation and could enable better-informed decision-making in emergency cardiac care. Our approach seems clinically applicable given its high accuracy and robustness, though further validation with larger-scale studies would be desirable.</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

Enhanced prediction of defibrillation success in out-of-hospital cardiac arrest using nonlinear ECG features and probabilistic neural network classification

  • 期刊论文

  • 发表日期：

    2025年5月28日

查看全文
  • Hamed Shamsi,
  • Alireza Golkarieh,
  • Houman Nouri,
  • Amirhosein Dolatabadi

摘要

The present study proposes a new approach to predict the outcome of external defibrillation in out-of-hospital cardiac arrest (OHCA) patients using sophisticated nonlinear feature extraction and classification via a probabilistic neural network (PNN) of electrocardiogram (ECG) signals. We analyzed 251 samples of ECG data collected from 195 unsuccessful and 56 successful resuscitations in OHCA patients. Our proposed methodology utilized the following five advanced nonlinear dynamics features: Sample Entropy, Approximate Entropy, Detrended Fluctuation Analysis, Correlation Dimension, and Largest Lyapunov Exponent. These features were selected such that the complex dynamics of ECG signals are not missed by traditional linear analyses and subtle patterns. Then, classification has been performed using the extracted features with a PNN model, which gave great performances with a training accuracy of 97.16% and a testing accuracy of 93.33%. With very good discriminative ability, the model resulted in a high area under the curve (AUC) of 0.98 and 0.96 for both the training and testing sets, respectively. This constitutes a significant advance in the ECG-based prediction of the outcome of defibrillation and could enable better-informed decision-making in emergency cardiac care. Our approach seems clinically applicable given its high accuracy and robustness, though further validation with larger-scale studies would be desirable.