基于子空间分析法的脑中风微波检测研究

发布时间：2018-03-28 22:13

本文选题：脑中风　切入点：微波　出处：《东华大学》2017年硕士论文

【摘要】：随着现在生活水平不断提升,脑中风发病人数激增,脑中风已经成为除癌症之外威胁人类生命的一大杀手,脑中风及时得到检测和治疗会大大提高存活率和治愈率。所以脑中风的早期诊断尤为重要,同时也得到了海内外研究人员的高度关注。脑中风微波检测是利用微波技术对脑中风进行检测,具有及时快速,低成本,高有效性并且安全性能良好的优点。目前人体微波无损检测主要采用基于电磁逆散射原理的微波成像技术,难以有效地应用于具有复杂结构组织,中风病灶与正常组织介电特性相差不大的脑中风检测,本文将模式识别的分类检测方法用于脑中风检测,采用基于子空间分析法的脑中风检测分类器模型,提出了基于天线对交叉点的中风病灶定位算法,有效而且快速地检测并定位脑中风。本文首先介绍了微波检测脑中风的理论基础,包括脑部复杂的结构组织以及微波检测的基本原理,进行微波检测系统的设计,主要由产生激励信号的调制高斯脉冲、发送和接收天线、数据收集模块和数据分析模块四个部分组成。其次,本文基于时域有限差分法(FDTD,Finite-Difference Time-Domain)算法创建人脑电磁计算仿真模型,获取微波透射S参数作为仿真数据样本,并对两类不同样本,即含有血块的脑部数据和正常脑部数据,进行标识,组成样本库;然后进行特征提取,建立子空间线性分类函数,利用交叉验证,设计和训练子空间分类器,来区分这两类样本,并利用主角度序列法,对分类子空间基向量进行优化;进一步,建立子空间分类器识别血块是否在天线对连线上,进而利用天线对交叉定位原理,进行脑中风病灶的定位。然后搭建实验平台利用实验数据验证脑中风检测和定位方法,实验平台利用超宽带天线收发射频信号,根据脑部介电特性采用材料替代脑部具体组织,构建脑部头型。微波收集模块采用罗德施瓦茨(ZVL)矢量网络分析仪收集S参数。通过脑中风检测实验系统证明了本方法对实验数据具有较高的分类准确性。最后总结全文,并进行展望。经仿真和实验系统证明,基于子空间分类方法的脑中风检测和定位方法,具有较高的分类正确性,是一种有效的脑中风检测方法,适用于入院前预诊断,对于中风病人早期诊断和治疗具有重要的应用意义。
[Abstract]:With rising living standards and a surge in the number of stroke cases, stroke has become a major killer of human life besides cancer. Timely detection and treatment of stroke can greatly improve survival and cure rates, so early diagnosis of stroke is particularly important. At the same time, it has received great attention from researchers at home and abroad. Microwave detection of stroke is a method of detecting stroke using microwave technology, which has the advantages of prompt, fast and low cost. At present, microwave imaging technology based on electromagnetic inverse scattering principle is mainly used in human microwave nondestructive testing, which is difficult to be effectively applied to complex structures. In this paper, the classification method of pattern recognition is applied to the detection of stroke, and the model of cerebral stroke detection classifier based on subspace analysis is used. In this paper, an algorithm for locating stroke focus based on antenna intersection is proposed to detect and locate stroke effectively and quickly. Firstly, the theoretical basis of microwave detection of stroke is introduced in this paper. Including the complex structure of the brain and the basic principles of microwave detection, the design of the microwave detection system is mainly composed of the modulation of Gao Si pulse, which generates the excitation signal, to send and receive the antenna. Data collection module and data analysis module are composed of four parts. Secondly, based on FDTDX Finite-Difference Time-Domain-based algorithm, a human brain electromagnetic computing simulation model is established, and microwave transmission S parameters are obtained as simulation data samples. Two different kinds of samples, brain data containing blood clots and normal brain data, are identified to form a sample database, and then feature extraction is carried out, subspace linear classification function is established, and cross-validation is used. The subspace classifier is designed and trained to distinguish the two kinds of samples, and the classification subspace basis vector is optimized by the main angle sequence method. Furthermore, the subspace classifier is established to identify whether the blood clot is on the antenna line. Then the principle of cross-localization of the antenna is used to locate the cerebral apoplexy focus. Then the experimental platform is built to verify the method of stroke detection and location using experimental data. The experimental platform uses ultra-wideband antenna to receive and receive RF signals. According to the dielectric properties of the brain, the material is used to replace the specific brain tissue. The microwave collection module uses Luo De Schwartz ZVLV vector network analyzer to collect S parameters. The experimental system of cerebral stroke detection proves that this method has high classification accuracy for experimental data. Finally, the full text is summarized. The simulation and experimental results show that the method of stroke detection and location based on subspace classification method has higher classification accuracy and is an effective method for stroke detection, which is suitable for pre-hospital diagnosis. It is of great significance for the early diagnosis and treatment of stroke patients.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN015;R743.3

【参考文献】