【摘要】 体外自动除颤器(AED)是挽救心脏骤停(SCA)患者生命的重要设备。可电击复律心律自动判别算法(SAA)是AED的核心技术。本研究在构建包括8 s的2 024段可电击复律心律(SHR)心电图(ECG)和7 884段不可电击复律心律(NSHR)ECG数据集的基础上，提出了一种基于机器学习的SAA。首先提取ECG的时域、频域、复杂度相关的32个特征，经筛选得到6个有效特征；之后用支持向量机实现SHR和NSHR自动分类。根据500次按患者随机分组的实验，敏感度、特异性、准确率的均值±标准差分别为97.62%±0.18%、99.15%±0.04%、98.79%±0.08%。所提出的SAA符合美国心脏病协会对AED中SAA敏感度超过90%,特异性超过95%的要求，可作为AED算法模块进行SHR的自动判别。更多还原

【Abstract】 Automatic external defibrillator(AED) is an important device in saving patients with cardiac arrest(SCA). Shockable advice algorithm(SAA) is the key technology of AED. In this work, our model was trained with a data set including 2 024 segments of shockable rhythms(SHR) and 7 884 segments of non-shockable rhythms(NSHR). We proposed a SAA based on machine learning. Combining 6 effective features selected from 32 features such as time domain, frequency domain and complexity, support vector machine was employed to classify SHR and NSHR. After 500 experiments, the mean value ±standard deviation of sensitivity of the algorithm was 97.62±0.18%, the specificity was 99.15±0.04%, and the accuracy was 98.79±0.08%. The results showed that the SAA proposed in this paper met the requirements of American Heart Association for SAA performance in AED, and it can be used as an AED algorithm module for automatic discrimination of SHR.更多还原