【作者】 陈健；

【导师】 王太宏；

【作者基本信息】 厦门大学 ， 精密仪器及机械， 2020， 硕士

【摘要】 糖尿病对世界各国人民的生活和工作造成了极大地影响,且临床上尚无根治的方法,目前的方法都是通过频繁监测患者的血糖结合药物来管理患者的血糖水平。现有临床上检测血糖都是通过大型仪器采血方式测量,市场上也出现了微创采血测量血糖的仪器,此类有创或微创方法虽然结果较为准确,但采血会给使用者带来创伤,容易感染,因此无创检测血糖的技术具有重要的意义。本文设计出一种无创血糖实时监控系统,并且设计出了样机,能实时测量人体血糖值。具体工作如下:(1)针对目前国内外研究无创血糖的成果,提出了结合能量守恒法和光谱法测血糖的方法。该方法增加了血糖参数特征,对血糖的计算更加精确严格,使结果更加准确。(2)本文设计出一种无创血糖实时监控系统,整个系统可分为硬件部分以及软件部分。硬件部分采用模块化设计思想,设计了一种集成多种传感器的数据采集仪器,采用STM32F103RCT6作为处理器,实时采集能量守恒法和光谱法测血糖的数据。采集的两种数据通过串口交替发送至上位机。(3)对光电容积脉搏波进行滤波去基线,同时提取特征。结合微创血糖值进行建模,使用多元线性回归、k近邻算法回归和支持向量回归三种机器学习算法模型进行训练,得出血糖无创测量函数,建立起无创血糖检测的方法流程。(4)对这三种不同机器学习算法设计具体实验进行对比,比较它们的准确度、相关系数、均方误差以及均方根误差。实验证明:基于支持向量回归算法的测量准确度最好,相关系数高达0.865,同时均方误差和均方根误差比多元线性回归、k近邻算法回归都更低。然后选取了基于高斯径向基核函数的支持向量回归模型作为最优模型,将该算法用于实时检测人体血糖。更多还原

【Abstract】 Diabetes has a great impact on the lives and work of people around the world,and there is no clinical cure.The current method is to frequently monitor patients’blood sugar combined with drugs to manage patients’ blood sugar levels.Existing clinical blood glucose testing is measured by blood collection by large instruments.There are also instruments for measuring blood glucose by minimally invasive blood collection.Although the results of such invasive or minimally invasive methods are more accurate,blood collection will bring trauma to users,Easy to be infected,so non-invasive blood glucose detection technology is of great significance.This paper has designed a non-invasive real-time blood glucose monitoring system,and designed a prototype that can accurately measure human blood glucose levels.The specific work is as follows:(1)According to the current research results of noninvasive blood glucose at home and abroad,a method for measuring blood glucose by combining energy conservation method and spectroscopic method is proposed.This method increases the characteristics of blood glucose parameters,makes the calculation of blood glucose more precise and strict,and makes the results more accurate.(2)This paper designs a non-invasive blood glucose real-time monitoring system,the entire system can be divided into hardware and software parts.The hardware part adopts a modular design idea,and designs a data acquisition instrument integrating multiple sensors.It uses STM32F103RCT6 as a processor to collect real-time energy conservation and spectrometry data for blood glucose measurement.The collected two kinds of data are sent to the host computer alternately through the serial port.(3)In the software part,the photoelectric volume pulse wave is filtered to remove the baseline,and features are extracted at the same time.Combined with minimally invasive blood glucose value for modeling,multiple linear regression,k-nearest neighbor algorithm regression and support vector regression are used to train three machine learning algorithm models to obtain blood glucose non-invasive measurement functions and establish a non-invasive blood glucose detection method.(4)Compare these three different machine learning algorithm design specific experiments,and compare their accuracy,correlation coefficient,mean square error and root mean square error.The experiment proves that the measurement accuracy based on the support vector regression algorithm is the best,the correlation coefficient is as high as 0.865,meanwhile the mean square error and root mean square error are lower than those of multiple linear regression and k-nearest neighbor algorithm.Then the support vector regression model based on Gaussian radial basis kernel function is selected as the optimal model,and the algorithm is used to detect human blood glucose in real time.更多还原