基于外部参数正交的葡萄糖近红外漫射光谱温度校正

External Parameter Orthogonalization Based Temperature Calibration on Near Infrared Diffuse Spectra for Glucose Measurement

不同温度下葡萄糖的近红外光谱存在较大差异,故温度干扰将影响葡萄糖浓度的预测精度.本文提出采用基于外部参数正交化的光谱预处理方法,对不同温度下葡萄糖在浑浊介质中的漫射光谱进行温度校正.该方法仅需预先对代表性样品进行不同温度下的漫射光谱采集,即可获得消除温度变量的滤波矩阵,利用该矩阵可以将所有温度下的光谱校正至基准温度水平.该校正方法预先对外部干扰变量单独建立模型,此模型与葡萄糖浓度预测模型的建立是分离的.该方法与温度、葡萄糖浓度混合建模的传统方法相比,复杂度大大降低.针对葡萄糖在浑浊介质中的测量场合,本文提出两种对于温度校正中代表性样品的选取方法,并实验验证了两种方案对葡萄糖在脂肪乳溶液中漫射光谱温度校正的效果.该方法为人体无创血糖测量中的温度校正提供了借鉴.

Near infrared spectra of glucose at different temperatures usually show obvious deviation, thus, temperature influence would decrease the prediction accuracy of glucose concentration. A pretreatment method based on external parameter orthogonalization （ EPO） is proposed to correct the diffuse transmis-sion spectra of glucose in turbid media at different temperatures. A typical sample is needed to collect its spectra at various temperatures, afterwards, a filter matrix of external parameter （temperature） is estab-lished to correct spectra at other temperatures to the reference temperature. The establishment of filter matrix is separated from the establishment of glucose concentration prediction model. And the modeling complexity of the proposed method is much lower than that in hybrid modeling of temperature and glucose concentration. Specifically, two effective methods of choosing the typical sample in the measurement of glucose in turbid media （Intralipid） have been discussed, and the temperature calibration results of the two methods in the diffuse transmission spectra have been verified. This study would also provide refer-ence for the temperature calibration in non-invasive blood glucose concentration measurement.