Zhang Haqi, Pan Xiaotian, Zhou Shan, Zhang Weiwei, Chen Jing, Pan Limin
Institute of Intelligent Media Computing, Hangzhou Dianzi University, Hangzhou, China.
Shangyu Institute of Science and Engineering Co. Ltd., Hangzhou Dianzi University, Shaoxing, China.
Front Cardiovasc Med. 2025 Jul 16;12:1550422. doi: 10.3389/fcvm.2025.1550422. eCollection 2025.
Out-of-hospital cardiac arrest (OHCA) represents a critical challenge for emergency medical services, with the necessity for rapid and accurate prediction of defibrillation outcomes to enhance patient survival. This study leverages a dataset of 251 ECG signals from OHCA patients, consisting of 195 unsuccessful and 56 successful resuscitation attempts as categorized by expert cardiologists. We extracted six crucial features from each ECG signal: heart rate, QRS complex amplitude, QRS complex duration, total power, low-frequency power (0.04-0.15 Hz), and high-frequency power (0.15-0.4 Hz). These features were derived using standard temporal and frequency domain methods. Subsequent analysis focused on selecting the most predictive features, with QRS complex amplitude, total power, and low-frequency power showing the highest discriminative ability based on their Area Under the Curve (AUC) values. A Support Vector Machine (SVM) classifier, trained on these selected features, demonstrated a prediction accuracy of 95.6%, highlighting the efficacy of combining targeted ECG signal features with machine learning techniques to forecast defibrillation success accurately. This approach provides a non-invasive, rapid, and reliable method to support clinical decisions during OHCA emergencies. Future research aims to expand the dataset, refine feature extraction techniques, and explore additional machine learning models to further enhance prediction accuracy. This study underscores the potential of ECG-based feature analysis and targeted machine learning in improving resuscitation strategies, ultimately contributing to higher survival rates in OHCA patients.
院外心脏骤停（OHCA）对紧急医疗服务来说是一项严峻挑战，需要快速准确地预测除颤结果以提高患者生存率。本研究利用了一个来自OHCA患者的包含251个心电图信号的数据集，该数据集由专家心脏病学家分类的195次复苏失败尝试和56次成功复苏尝试组成。我们从每个心电图信号中提取了六个关键特征：心率、QRS波群振幅、QRS波群持续时间、总功率、低频功率（0.04 - 0.15Hz）和高频功率（0.15 - 0.4Hz）。这些特征是使用标准的时域和频域方法得出的。后续分析着重于选择最具预测性的特征，基于曲线下面积（AUC）值，QRS波群振幅、总功率和低频功率显示出最高的判别能力。在这些选定特征上训练的支持向量机（SVM）分类器显示出95.6%的预测准确率，突出了将有针对性的心电图信号特征与机器学习技术相结合以准确预测除颤成功的有效性。这种方法提供了一种非侵入性、快速且可靠的方法，以支持OHCA紧急情况期间的临床决策。未来的研究旨在扩大数据集、改进特征提取技术，并探索其他机器学习模型以进一步提高预测准确率。本研究强调了基于心电图的特征分析和有针对性的机器学习在改善复苏策略方面的潜力，最终有助于提高OHCA患者的生存率。
