基于外部参数正交的无创血糖测量温度校正

Temperature Correction of NIR Reflectance Spectrum of Noninvasive Blood Glucose Measurement Based on EPO

近红外人体血糖浓度无创检测所面临的主要问题之一是人体温度变化干扰测量光谱。为此,提出使用基于外部参数正交化(EPO)的光谱预处理方法,对测量部位体温改变时的光谱进行温度校正。该方法仅需预先采集人体体温变化时的漫射光谱,即可获得消除温度干扰的滤波矩阵,利用该矩阵可以将不同体温下的光谱校正至基准温度水平。预先对外部干扰变量单独建立模型,与血糖浓度预测模型的建立分离。EPO原理提出组成光谱空间的干扰信号空间与有用信号空间正交,即温度光谱响应与葡萄糖浓度光谱响应之间彼此正交,而在实际测量中,仪器系统漂移,人体出汗等共模干扰常导致有用信号和干扰信号存在偶然相关,影响了消除温度干扰的效果。因此在进行温度校正之前,首先对原始光谱进行位置差分处理,经验证位置差分方法能够消除仪器系统漂移带来的共模干扰,获得的吸光度光谱中温度响应部分和浓度响应部分彼此正交。使用蒙特卡洛模拟人体三层皮肤模型获得血糖光谱数据,模拟样品参数均根据实际人体实验中的参数水平设置。对受温度干扰的光谱进行位置差分处理后使用EPO进行温度校正,然后利用校正后的光谱数据建立偏最小二乘回归(PLSR)模型。与校正前光谱建模结果比较,校正后的模型均方根误差(RMSEC和RMSECV)明显降低,相关系数得到一定的提高,同时主成分数减少,验证了该温度校正方法的有效性。

One of the main challenges in the noninvasive sensing of blood glucose by near-infrared(NIR)spectroscopy is that human temperature changes interfere with the measurement spectrum.In this paper,a spectral pretreatment method based on external parameter orthogonalization(EPO)is proposed to eliminate the spectral variations from temperature interferences caused by the temperature changes at the measured position.This method only needs to collect the diffuse spectrum when the body temperature changes in advance,using whichwe can obtain the filtering matrix to eliminate the temperature interference.This matrix could be used to calibrate the spectrums at different body temperatures to the reference level.This method establishes a model for external disturbance variables separately in advance and separates it fromthe model between glucose concentrations and the diffuse reflectance.The principle of EPO indicated that spectral space is composed of interference signal space and useful signal spacethat are orthogonal to each other.In other words,the temperature response and the glucose concentration response in the spectral is orthogonal to each other.However,in the actual situation,the instrument system drift,common-mode disturbance such as human body sweating often leads to useful signal and interference signal accidentally correlational that does harmto the effectiveness of EPO.Therefore,we first use the differential correction method based on the spectra from the reference position and measuring position on the original spectrum.It has been proved that the differential correction method can eliminate the common-mode interference brought by the instrument system.In addition,the temperature response part and the concentration response part of the obtained absorbance spectrum are orthogonal to each other.In this paper,the spectral data were obtained by Monte Carlo simulation of human three-layer skin model,and the parameters of the simulated samples were set according to the actual human experiment.The EPO method was used tocalibrate temperature interference in spectra after the differential correction method,and then a partial least squares regression(PLSR)model was established based on the corrected spectral data.Compared with the spectral modeling results before calibration,the root means square error of calibration(RMSEC)was reduced;the correlation coefficient was improved,and the number of principal components was reduced,which indicated the effectiveness of the temperature correction method-EPO.