基于双目视觉的脉搏信息及血压波形测量方法研究

发布时间：2018-08-09 21:05

【摘要】：据《中国心血管病报告2016》概要中统计,心血管疾病死亡率占居民疾病死亡之首位,并呈逐年上升趋势,心血管疾病的防治亟待解决。脉搏搏动与血压变化源于心脏搏动,临床研究表明,在心血管疾病发病早期,某些异常信息已蕴涵在脉搏信号及动脉血压波形中。若能精确检测全面的脉搏信息,无创连续监测血压波形,就可以获得更多的生理、病理信息,有助于对心血管疾病的有效预防及诊断治疗,对于临床医学和基础医学均具有重要意义。本文在综述了血压测量方法及脉搏信号检测与分析方法的国内外研究现状的基础上,分析了现有各种方法的优缺点,得出了现有的脉搏信号检测方法检测的脉搏信息不够全面、血压波形的无创测量方法尚存在较多缺陷的结论,提出了基于双目视觉测量原理检测脉搏信号,进而实现无创血压波形测量的研究思路。首先,研制了双目视觉脉搏检测系统。采用柔性探头将桡动脉脉搏的搏动转换为探头薄膜的形变,用工业相机捕捉薄膜的变化信息,实现脉搏搏动可视化,以图像的方式检测全面的脉搏信号。针对相机自带图像采集软件的不足,基于相机的应用程序编程接口,在Microsoft Visual C++2008环境下二次开发了双相机同步采集软件,在双口千兆以太网卡及MP425数据采集模块的辅助下,利用相机的外触发模式,实现双相机稳定同步采集及自动连续保存。选择合适的高精度标定板,并自制背光源,应用张正友标定方法对工作在小视场、微距环境下的双目视觉系统进行标定,标定结果精度较高,满足脉搏幅度测量的要求。其次,基于双目视觉脉搏检测系统采集的动态脉搏图像序列,提取多点脉搏波。根据脉搏搏动引起的探头薄膜变形导致图像灰度在帧间发生变化的特点,采用常用的基本图像统计量提取脉搏波,并进行比较,提出了基于图像熵获取多点脉搏波的方法。该方法效率高,波形稳定性好,能准确获取脉搏搏动趋势,效果优于基于特征图案面积变化的方法。为了进一步获得脉搏搏动的实际幅值,提出了基于特征点纵向位移的多点脉搏波提取方法。提取脉搏图像中特征图案的质心作为特征点,引入“弓”字型扫描检测思想,将左右图像中的特征点进行匹配;基于双目视觉理论计算特征点的空间位置,继而提取纵向位移;根据多个特征点的纵向位移在帧间的变化,获取能精确反映实际脉搏幅值及脉搏搏动趋势的多点脉搏波。波形效果优于基于特征图案面积变化的方法及基于图像熵的方法。此外,对该方法的准确性和有效性进行了静态和动态验证。在多特征点纵向位移的基础上,获取动态脉搏三维形态。根据同一时刻多个特征点的空间位置,重构该时刻的薄膜三维形态,结果符合实际情况;分析探头薄膜在脉搏搏动下的受力形变情况,根据同一时刻多个特征点的脉搏幅值,重构该时刻由于脉搏搏动引起的薄膜形变;基于脉搏图像序列,获取随时间变化的动态脉搏三维形态,将脉搏形象立体地呈现出来。在脉宽、脉长基本概念的基础上,定义了计算方法,并基于动态脉搏三维形态计算脉宽及脉长。最后,基于脉搏波获取血压波形。构建模拟人体血液循环系统的血压模拟平台,用于对比分析血管的管壁径向位移波形与管内压力变化波形的关系,得出了双目视觉脉搏检测系统与压力变送器检测的信号波形一致性较好的结论,并据此提出了基于脉搏波获取血压波形的研究方法。在分析了血管内压力与管壁径向位移之间的线性关系之后,推导出从脉搏幅值计算血压的数学表达式;确定脉搏中心作为获取血压波形的最佳位置,初步估算血压波形。采用简单线性相关分析方法确定与血压值相关性较强的脉搏特征参数,采用多元线性回归分析法获取收缩压和舒张压,进一步修正血压波形。此外,借助血压模拟平台,初步研究了基于BP神经网络预测模拟血管内压力波形的方法,为人体血压波形测量方法的研究提供了一种思路。
[Abstract]:According to the 2016> summary of the Chinese cardiovascular disease report, the mortality of cardiovascular disease accounts for the first of the death of the residents, and it is increasing year by year. The prevention and treatment of cardiovascular disease should be solved urgently. Pulse pulsation and blood pressure change from heart beat. Clinical research shows that some abnormal information has been contained in the pulse in the early stage of cardiovascular disease. In the signal and arterial blood pressure waveforms, it is of great significance for the clinical medicine and basic medicine to obtain more physiological, pathological information and help to the effective prevention and diagnosis and treatment of cardiovascular diseases if it can accurately detect the full pulse information and non-invasive continuous monitoring of blood pressure waveform. The methods of blood pressure measurement and the methods of blood pressure measurement are reviewed in this paper. On the basis of the domestic and foreign research status of the pulse signal detection and analysis methods, the advantages and disadvantages of the existing methods are analyzed. The conclusion is drawn that the pulse information of the current pulse signal detection method is not comprehensive, and the non invasive measurement method of the blood pressure waveform still has many defects, and the detection pulse based on the binocular vision measurement principle is proposed. First, a flexible probe is used to transform the pulse pulsation of the radial artery into the deformation of the probe film, and the change information of the film is captured by an industrial camera, and the pulse pulsation is visualized by the industrial camera, and a comprehensive pulse letter is detected in the image way. In view of the shortage of camera acquisition software, based on the application programming interface of the camera, the dual camera synchronous acquisition software is developed two times under the environment of Microsoft Visual C++2008. With the assistance of the dual port Gigabit Ethernet card and the MP425 data acquisition module, the camera's external trigger mode is used to achieve the stable synchronous acquisition of the dual cameras. And automatic continuous preservation. Select the appropriate high precision calibration board, and self-made backlight, use Zhang Zheng friend calibration method to calibrate the binocular vision system working in the small field of view and in the environment of micro distance. The precision of the calibration result is high and the requirement of pulse amplitude measurement is satisfied. Secondly, the dynamic pulse graph based on the binocular vision pulse detection system is used. According to the deformation of the probe film caused by the pulse pulsation, the image gray level is changed in the frame. The pulse wave is extracted by the common basic image statistics, and the method of obtaining the multipoint pulse wave based on the image entropy is proposed. The method is efficient, the wave stability is good, and the method can be accurately obtained. In order to obtain the actual amplitude of pulse pulsation, a multi point pulse wave extraction method based on the feature point longitudinal displacement is proposed to obtain the actual amplitude of pulse pulsation. The centroid of the feature pattern in the pulse image is extracted as the feature point, and the "bow" scan detection idea will be introduced. The feature points in the image are matched; the spatial position of the feature points is calculated based on the binocular vision theory, and then the longitudinal displacement is extracted. According to the variation of the longitudinal displacement of the multiple feature points, the multi point pulse wave can be obtained to accurately reflect the actual pulse amplitude and pulse pulsation. The waveform effect is superior to the change of the area of the characteristic pattern. Method and method based on image entropy. In addition, the accuracy and validity of the method are verified static and dynamic. On the basis of the longitudinal displacement of multi feature points, the three-dimensional shape of dynamic pulse is obtained. According to the spatial position of multiple feature points at the same time, the three-dimensional shape of thin film is reconstructed at this time, and the result accords with the actual situation; The force deformation of the head film under pulse pulsation, according to the pulse amplitude of multiple feature points at the same time, reconstructs the deformation of the film caused by pulse pulsation at this time. Based on the pulse image sequence, the dynamic pulse three-dimensional shape that changes with time is obtained and the pulse image is presented. The basic concept of pulse width and pulse length is based on the basis of the pulse image sequence. At the same time, the calculation method is defined and the pulse width and the pulse length are calculated on the basis of the dynamic pulse three-dimensional shape. Finally, the blood pressure waveform is obtained based on the pulse wave. The blood pressure simulation platform is constructed to simulate the blood circulation system of the human body, which is used to compare and analyze the relationship between the radial displacement waveform of the vessel wall and the change of the pressure change in the tube, and the binocular vision pulse is obtained. A conclusion is drawn that the detection system is in good agreement with the pressure transmitter, and a research method based on pulse wave is proposed. After analyzing the linear relationship between the pressure of the blood vessel and the radial displacement of the tube wall, the mathematical expression of the blood pressure from the pulse amplitude is deduced and the pulse center is determined as the center of the pulse. The best position of blood pressure waveform was obtained and the blood pressure waveform was estimated preliminarily. A simple linear correlation analysis method was used to determine the pulse characteristic parameters which had strong correlation with the blood pressure value. The systolic and diastolic pressure was obtained by multiple linear regression analysis, and the blood pressure waveform was further revised. In addition, the BP God based on the blood pressure simulation platform was used to preliminarily study the deity based on the blood pressure simulation platform. The method of predicting and simulating the intravascular pressure waveform by network provides a new idea for the research of measuring the blood pressure waveform.
【学位授予单位】：兰州理工大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：R443.5;TP391.41

【参考文献】