【作者】 王刚；

【导师】 饶妮妮；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2008， 博士

【摘要】 房颤是临床最常见的一种心律不齐病症，在通常人群中的发病率为0．4％到1．0％，并且随着人群年龄的增加，其发病率逐渐提高。临床医学表明，房颤是诱发血栓，导致心血管疾病的重要因素。除导致血流动力学改变和生活质量下降外，房颤对人体有严重的致残和致命的威胁。因此，房颤的监护和治疗正逐渐成为当今国际心电生理研究的热点。在临床监护中，用“无损”的方法从房颤病人的体表心电图中得到房颤信号是分析和描述房颤特征十分重要的一个环节，房颤信号的提取起着关键性的作用。在临床治疗中，射频导管消融(radiofrequency catheter ablation，RFCA)治疗房颤的安全性和有效性在过去的10年中得到了验证。然而，相关的治疗仪器市场却被国外品牌(Carto和Ensite)所占据，射频导管消融治疗仪器的开发是国内临床医学界亟待解决的技术难题之一。本论文围绕房颤的监护和治疗研究了以下两个方面的内容：基于盲源分离的房颤信号的提取；RFCA软件系统中图像配准。研究内容分列如下：1在盲源分离的基础上引入基于二阶统计量的盲源提取技术来提取房颤信号。由于盲源分离技术对体表12导联心电数据进行分离将得到12组源信号，必须对分离后的所有12组信号做进一步判断(例如频谱分析法等)才能决定房颤信号，这样不仅计算量大而且提取精度差，不利于临床监护。为了解决这个问题，我们提出一种新的基于二阶统计量的盲源提取算法，可直接提取出房颤信号。理论分析和实验仿真证明了该算法的有效性。2提出一种基于四阶统计量的盲源提取算法来提取房颤信号。房颤病人的体表心电图是房颤波、心室波和其它干扰的混合。心室波的幅度较大，是正峭度的超高斯信号；房颤波的幅度较小，是负峭度的亚高斯信号，但具有绝对值最大的负峭度；其他干扰可以近似为高斯噪声。因此，我们利用信号的四阶统计量提取房颤信号。实验仿真证明了该算法的有效性。与盲源分离算法相比，盲源提取算法只提取一个所需信号，从而更适合临床监护。3研究心脏三维标测中的配准算法，并介绍了我们与国内医疗仪器厂商合作开发射频导管消融治疗仪器的一些研究进展。在图像配准中，我们引入了基于12个自由度的仿射变换模型和相应的迭代最近点算法来实现配准，并研究了如何把该算法与手术操作结合起来。结合OpenGL，我们用Visual C++6．0软件实现了该配准方法。仿真结果验证了该算法的有效性。更多还原

【Abstract】 Atrial fibrillation (AF) is the most common sustained arrhythmias encountered byclinicians and occurs in approximately 0.4%-1.0% of the general population.Itsprevalence increases with age.There is increasing awareness that AF is major cause ofemolic events and cerebrovascular accidents.Symptoms such as occasionally disablinghaemo-dynamic impairment and a decrease in life expectancy are among the untowardeffects of AF,resulting in an important morbidity and mortality.In this respect,theclinic AF patient monitor and its treatment has been the subject of arousing interest andintensive clinical research in recent years.In clinic patient monitor,the non-invasive approach for the AA estimation in AFepisodes is a key step in the analysis and characterization of AF.In clinical treatment,safety and efficacy of radiofrequency catheter ablation (RFCA) guided by 3-dimension(3-D) mapping system in patients with AF have been verified in the past ten years.However,the civil market of RFCA machines has been 100% penetrated by foreignproducts,i.e.,Ensite3000 and Carto system.It has been in great need to develop RFCAsystem with our own intellectual property rights.This dissertation,focused on AF monitor and treatment,includes the following twoparts:the AF signal extraction from surface 12-lead Electrocardiograph (ECG) and theimage registration in the software of RFCA system.The main contents are organized asfollows:1 A second order statistics (SOS) based blind source extraction (BSE) algorithm ispresented to extract AF signal.The early method based on BSS utilized all theinformation from 12 lead,and could obtain 12 source signals including AF signal.Thenthe AF signal would be selected with the help of power spectrum analysis.However,theselection sometimes goes wrong and may not work well in monitor.Here we presente anew method based on BSE to settle this problem.The efficacy is verified by theoreticalanalysis and simulation results.2 A fourth order statistics (FOS) based BSE algorithm is proposed to extract AFsignal.Note that the 12-lead ECG of AF patients are composed of the independent sources of atrial activiey (AA),ventricular activity (VA) and other nuisance signals.With respect to non-Gaussianity,VA presents high values within the heart beat (QRScomplex) and low values in the rest of the cardiac cycle.Hence,the histogram analysisof VA reveals a super-Gaussian behavior with high positive kurtosis value.On the otherhand,AA behaves as a sub-Gaussian random process with low negative kurtosis value.The other nuisance signals whose kurtosis approximates zero can be regarded asGaussian noises.Then we can use the HOS based BSE algorithm to extract AF signal.The validity and performance of this algorithm are confirmed by extensive computersimulations and experiments on real-world data.Compared with BSS,BSE onlyextracts one desired signal,and thus is more suitable to clinical monitor.3 Regarding the research of the image registrastion in the RFCA system,some resultsof how we do the RFCA system with the civil company are presented.In thecooperation,we focused on how to show and registrate the images in software.Regarding the registration,we introduced affine transformation model using twelve freedegrees and the corresponding iterative closest point algorithm.We also did research onhow incorporate the registration algorithm into the clinical operations.With the help ofOpenGL,the algorithm was realized via Visual C++ 6.0.Simulation results and theanimal experiment verified this algorithm.更多还原