房颤f波提取、分解方法及f波子波应用研究

Atrial Fibrillation (AF) is the most common super-ventricular arrhythmia. Recent studies show that persistent AF is sustained by multiple localized rotors. That means fabrillatory (f) wave in surface electrocardiogram (ECG) can be decomposed into several daughter-f waves which reflect electrical activity of localized rotors. In this study, we will focus on two key points of f wave analysis, f wave extraction and f wave decomposition. Computation of f wave's new features and f wave decomposition require f wave with less distortion. However, f waves extracted by existing single-lead extraction algorithms are with serious local distortion. In order to obtain f wave without distortion, we will carry out research on single-lead f wave extraction following two ways, namely, signal reconstruction and statistical estimation. To obtain daughter-f wave through decomposition of multi-lead f waves, blind source separation theory would be taken as a key tool. Based on simulation model of ECG forward problem, we will found the model of the relationship between epicardial potential and body surface potential. Based on this model, we will non-invasively estimate the location of localized rotors in atria using daughter-f wave's separation vector. In addition, we will carry on research on daughter-f wave's clinical applications, mainly on predicting outcome of electrical cardioversion for AF patients.

心房颤动是一种最常见的室上性心律失常。最近的研究表明，大多数持续性房颤患者的心房内同时有多个颤动源共同起作用，体表心电信号中的f波是心房内各颤动源对应的f波子波的线性叠加。本课题重点研究f波分析过程中的两个关键环节，即f波提取和f波分解问题。针对现有的单导联f波提取算法所提取的f波存在局部严重失真的问题，拟分别从信号重构和统计估计的角度展开研究，尽可能达到无失真提取f波的目标，以满足f波新特征量计算和f波分解的需求；为了由多导联体表f波分解获得各颤动源对应的f波子波，拟采用盲源信号分离的相关方法实现f波子波的分离；同时，拟从心电正向问题仿真模型出发，建立心房外膜电势与体表电势的映射模型，在此基础上，探索利用f波子波分离向量实现颤动源位置的无创伤定位；另外，本课题还将积极探索实践f波子波的应用，拟展开f波子波在房颤电复律治疗效果预评估中的应用研究。

本项目在房颤心电信号预处理算法、房颤f波提取算法、房颤f波分解算法以及心电采集专用设备等方面均有深入研究。在预处理方面，提出了多特征融合判断的伪迹识别方法和基于形态变换的单导联QRS复合波检测方法。在房颤f波提取算法方面，提出了两种f波重构算法：分别基于信号外推理论和ARMA模型，这两种算法均显著降低了所提取f波的失真程度，明显优于现有算法。在房颤f波分解算法方面，提出了基于PCA的分解算法，设计出了中心频率和频谱集中度相结合的f波子波识别规则。在心电采集专用设备研制方面，研制出了高度集成的心电采集专用芯片（集成了所有的模拟前端电路），在功耗、体积、精度方面均有明显优势，并基于此芯片研制出了穿戴式的心电信号采集仪。基于本项目的研究成果，共发表了4篇SCI期刊论文，2篇国际会议论文，并获得了2项中国发明专利授权。