结合静态小波变换的自适应窗口可除颤节律检测  

Detection of Shockable Rhythm Using Adaptive Window and Stationary Wavelet Transform

作　　者：罗美荣 杨丰[1] 詹长安[1] Luo Mei-rong;Yang Feng;Zhan Chang’an(School of Biomedical Engineering,Southern Medical University,Guangzhou Guangdong 510515,China)

机构地区：[1]南方医科大学生物医学工程学院

出　　处：《航天医学与医学工程》2019年第6期539-546,共8页Space Medicine & Medical Engineering

基　　金：国家自然科学基金(61771233)

摘　　要：目的为克服固定时间窗口分段导致R波截断的缺点,提出一种新的分段方法——固定R波个数的自适应窗口分段法,用于可除颤节律检测。方法首先用波谷波峰法定位ECG信号R波,以5个R波长度为窗口大小自适应分段ECG信号,得到每段信号能量、复杂度、时间长度时域特征;其次再对每段信号进行静态小波变换,获取每层小波系数与原始分段信号的相关系数时频特征;最后混合时域、时频域两类特征,输入到支持向量机、k-近邻、随机森林分类器进行信号分类,实现可除颤节律检测。结果在CUDB和VFDB两个开源数据库上对新算法进行验证比较,其准确率最高分别为:98.12%、97.19%;敏感度最高分别为:97.20%、95.88%;特异性最高分别为:98.72%、97.96%。结论新算法能够很好地实现可除颤节律的检测。Objective To overcome the shortcoming of R wave truncation caused by fixed time-window ECG segmentation,a new segmentation method was proposed.The time-window was adaptive to the fixed number of R-waves and the features of segmented signal could be used to detect the shockable rhythm.Methods The peaks and troughs were used to locate the R-waves in the ECG signal and every 5 R-waves were taken as the window size for one ECG segment.The time-domain features of segment energy,complexity and duration were obtained.The segmented ECG episodes were then decomposed using stationary discrete wavelet decomposition for each episode.The correlation coefficient between each wavelet coefficient and its original ECG episode was used as time-frequency feature.Finally,the above features were used as input to support vector machines,k-Nearest Neighbors and random forests algorithm for classification of shockable and non-shockable rhythms.Results Based on the open source databases(CUDB and VFDB),the new algorithm achieved the best accuracy of 98.12%and 97.19%,the best sensitivity of 97.20%and 95.88%and the best specificity of 98.72%and 97.96%respectively.Conclusion The new algorithm can effectively detect shockable rhythm.

关 键 词：可除颤节律 R波 分段 静态小波变换 支持向量机 K-近邻 随机森林 

分 类 号：R318[医药卫生—生物医学工程] R857[医药卫生—基础医学]