【作者】 杨晓利；

【导师】 汤井田；

【作者基本信息】 中南大学 ， 生物医学工程， 2011， 博士

【摘要】 可电击复律心律(shockable rhythm, ShR)是引起心脏性猝死(sudden cardiac death, SCD)的主要原因,ShR包括室性心动过速(Ventricular Tachycardia,VT)和心室颤动(Ventricular Fibrillation,VF)。早期电除颤是治疗VF有效的方法。自动除颤仪(Automated External Defibrillator, AED)可增加院外SCD的存活率。而AED实现其功能的关键一步是快速准确地自动检测出ShR。本文对基于熵的ShR检测进行了研究,具有重要的理论和实际意义。本文基于不同的心律对应着不同的系统复杂程度的思想,对ShR检测展开研究,重点研究了基于近似熵、样本熵和多尺度熵的ShR检测方法。论文通过理论分析,模拟实验和实际应用等手段,分析了室性心律失常信号的非线性特性,研究了基于HHT与双谱分析相结合的ShR去噪方法,以及有效地区分心室扑动(Ventricular Flutter, VFL)、心脏停博(Asystole, ASYS)等信号的基于熵的ShR检测方法,并与其它基于复杂度的信号检测方法进行了对比。首先,论文研究了室性心律失常信号的非平稳性。从平稳信号的定义出发,研究了ShR的均值、方差、自相关函数和功率谱估计,确定了ShR的非平稳性。针对ShR的非平稳特性,用时频分析方法分析ShR的频率内容,以及频率随时间变化的规律,对比了多种时频分析方法的应用效果,结果表明HHT时频分析法时频聚集性强。并通过HHT时频分析精确地归纳出了ShR的频率与时间的关系。接着,本文研究了基于HHT—双谱分析的ShR去噪方法,首先在时域进行EMD分解,选择所需的IMF进行重构完成初步去噪,并应用SNR和RMSE对重构结果进行评价；再把去噪后的信号进行双谱分析,在双谱域用PSNR对信号去噪进行评价。研究结果表明HHT—双谱分析法利用有用信号频段EMD重构和双谱分析达到了双重消噪的目的,为后续的心电特征点的准确识别奠定了基础。然后,基于不同的心律可能对应着不同的系统复杂程度的思想,尝试用度量序列复杂性和统计量化的非线性动力学参数--熵来检测ShR,取得了良好的效果。提出了基于熵的ShR检测方法,包括基于近似熵的ShR检测法,基于样本熵的ShR检测法和基于多尺度熵的ShR检测法,并通过CUDB和VFDB数据库的所有数据进行检测,结果表明该方法能快速有效的进行ShR检测。针对基于近似熵的ShR检测法对于VFL和ASYS的检测效果良好且基于标准化斜率绝对值标准差算法对其它心律有较高的检测率,提出了基于标准化斜率绝对值标准差-熵结合的ShR检测法并进行算法实现,并通过CUDB和VFDB数据库的全部数据进行验证,结果表明基于标准化斜率绝对值标准差-熵结合的ShR检测法取得了比单一方法更好的检测效果。最后,总结了全文的主要内容和创新工作,并指出了研究工作的不足之处和今后工作的建议。更多还原

【Abstract】 Shockable Rhythm (ShR), which includes Ventricular Tachycardia (VT) and Ventricular Fibrillation (VF), is the main cause of Sudden Cardiac Death (SCD). Early defibrillation is the only effective way for treating VF. Automated External Defibrillator (AED) can increase the survival rate of SCD outside hospital. Moreover, the key step for AED to implement its function is to detect ShR quickly, accurately and automatically. ShR detection algorithm based on entropy, which has important theoretical and practical significance, has been studied.Different cardiac rhythms correspond to different level of system complexities, based on which ShR detection algorithm was studied, mainly involving approximate entropy, sample entropy and multiscale entropy. Through theoretical analysis, simulation experiment and practical application, the nonlinearity of signals in ventricular arrhythmia was analyzed, ShR denoising algorithm based on HHT and bispectrum analysis was studied, and ShR detection algorithm based on entropy for effectively distinguishing Ventricular Flutter (VFL), Asystole (ASYS) was also studied; meanwhile, which was compared with other algorithms based on complexity.Firstly, the nonstationarity of the signals in ventricular arrhythmia was studied. Based on the stationarity signals, the mean, variance, autocorrelation function and power spectrum estimation of ShR was studied, and nonstationarity of ShR was determined. Focusing on the nonstationarity of ShR, the frequency spectrum of ShR and its variation profile over time was analyzed by time-frequency analysis algorithm. Multiple time-frequency analysis algorithms were compared, from which HHT time-frequency algorithm has been found with the characteristic of high time-frequency aggregation, through which the relationship between frequency and time of ShR has been accurately summarized.Subsequently, based on HHT-bispectra analysis algorithm, de-noising processing to ShR signal was conducted. Firstly, decomposition and reconstruction in the time domain by EMD was implemented, the IMF (Intrinsic mode functions) needed was selected for reconstruction, and evaluation was conducted in accordance with SNR and RMSE, secondly, bispectra analysis was implemented in denoised signals; the denoised signals were evaluated by SPNR in bispectra domain, which is unavailable in previous algorithms. The study suggested that HHT-bispectra analysis algorithms using EMD (Empirical Mode Decomposition) reconstruction and bispectra analysis for avaliable signal frequency band achieved double denoising, which lay the foundation for accurately identifying the characteristic point in ECG in future.Different cardiac rhythms may correspond to different system complexities, so entropy, a nonlinear dynamic parameters used for measuring sequence complexity and as statistical quantification, was used for detecting ShR, which acquired good effect. ShR detection algorithm based on entropy including sample entropy and multiscale entropy was proposed, which can quickly and effectively detect ShR, based on verification by CUDB and VFDB database. In view of good effect in VFL and ASYS detection by ShR detection algorithm based on approximate entropy, and high detection rate to sinus rhythm based on standard deviation of standard slope absolute value, combining algorithm for instance, ShR detection algorithm based on standard deviation of standard slope absolute value-approximate entropy has been proposed; through verification by CUDB and VFDB database, it has demonstrated that combining algorithm has better detection effect than single algorithmFinally, the main context and innovation in this paper was summarized, inadequacy in study was proposed, and recommendation for further work was given.更多还原