基于脉搏时空解析建模的恶性心律失常定量识别方法研究

In the process of malignant arrhythmia recognition, aiming to some problems, e.g. the pulse periods and waveforms changes of patients are unclear, and the abnormal segments and pathological segments in pulse signal are difficult to distinguish, quantitative identification methods for malignant arrhythmia with engineering application prospect are proposed based on pulse signal time-space analytical model. The main research contents include: (1) On the basis of analyzing the time domain and spatial domain characteristics of pulse signal, a pulse time-space analytical modeling method is proposed to quantitatively describe the changes of pulse waveforms and periods when combing pulse decomposition model with pulse rhythm. (2) The cyclostationary characteristics of the pulse signal are integrated into the model parameter estimation process to improve the speed and accuracy of the parameter estimation, and an abnormal segments adaptive detection method is proposed by the difference of model parameters between the abnormal segments and pathological segments. (3) A quantitative method for analyzing the pathological differences of model parameters is studied after clarifying the intrinsic connection and regularity among the synchronous pulse and ECG signals, and a model parameter fusion method is employed for quantitative identification of malignant arrhythmia. (4) With the help of the developed test prototype, an objective evaluation system for the proposed methods is designed from the three dimensions of the algorithm itself, the clinical medical measurement and the practicality. It is believed that our results will enrich the methods which would be engaged to describe the change of dynamic pulse signal quantitatively and detect abnormal segments in pulse signal and will provide theoretical support for the research of malignant arrhythmia recognition, remote monitoring and early warning as well.

针对恶性心律失常导致的脉搏波周期和形态变化不明、干扰段与病理段难以区分等问题，本项目拟基于脉搏时空解析模型，提出具有工程应用前景的恶性心律失常定量识别方法。主要研究内容包括（1）分析脉搏信号时域和空间域特征，将脉搏解析模型与脉律结合，提出脉搏时空解析建模方法，定量描述脉搏波形态和周期变化；（2）将脉搏信号的循环平稳特性引入模型参数的估计过程，提高参数估计的速度和精度，并根据干扰段和病理段模型参数的差异，提出干扰段自适应检测方法；（3）依据同步脉搏心电信号间内在联系与规律，提出模型参数病理学差异的定量分析方法，以及基于模型参数融合的恶性心律失常定量识别方法；（4）结合已研制的试验样机，从方法开发本身、临床医学度量和方法实用性三个维度，设计系统对所提出方法进行客观评价。本项目研究成果将进一步完善和丰富动态脉搏信号定量描述与干扰段检测方法，并为恶性心律失常识别、远程监测与预警的研究提供理论支撑。

恶性心律失常是导致患者晕厥甚至死亡的一种心血管疾病，大部分患者在医院外发病，其具有突发性、偶发性的特点，增加了防治难度。恶性心律失常会引起心血管系统血流动力学和搏动节律的变化，现有基于脉率的诊断方法忽略了血流动力学参数的变化，准确率有待提高。因此，本项目以恶性心律失常定量识别为研究对象，将脉搏信号解析模型与脉律结合，研究脉搏信号时空解析建模方法，定量描述脉搏波形态和周期的变化；研究解析模型的参数估计方法，从模型中提取可用于极度心律过缓、极度心律过速、室性心律过速、心室扑动/颤动等恶性心律失常识别的量化参数；研究基于机器学习的恶性心律失常识别方法；研制样机验证所提出方法。本项目主要研究成果为：.（1）针对在线采集脉搏信号的噪声抑制问题，将微处理器数据更新过程与数学形态学滤波原理相结合，提出一种快速的数学形态学滤波法。实验结果表明：该方法在保持原方法精度不变的前提下，可用于脉搏信号单采样点实时滤波。.（2）针对在线采集脉搏信号数据量大的问题，提出一种基于时域关键特征点的脉搏信号压缩方法，并构建信号压缩与恢复流程。实验结果表明：该方法在保持高压缩比的同时，保留了时域波形的关键特征，可直接用于脉率等参数提取。.（3）针对脉搏波周期和形态的量化描述问题，根据脉搏波的形成机理与时域波形特征的关联，构建时空解析模型及其参数求解流程，通过实验获取最优化解析模型。实验结果表明：3-Lognormal函数模型为脉搏波量化描述的最优选择，模型参数可有效反映心血管系统相关的生理和病理信息。.（4）针对恶性心律失常识别问题，分析脉搏主波间期特征及脉搏时空解析模型参数的显著性变化，选择贡献率高的特征，基于决策树、极限学习、随机森林等机器学习算法实现恶性心律失常识别。实验结果表明：四种恶性心律失常识别平均准确率在95%以上。.（5）开发和升级可穿戴设备原型样机，对所提出的技术和算法进行验证，探索项目研究理论及成果的应用前景。

内容获取失败，请点击重试