近红外光谱无创血糖检测的模拟样品试验研究

提出了采用牛奶作为模拟样品，对血糖测量中的测量方法和波长选择等基础问题进行了研究。在1560～1750nm和2090-2190nm的波段：对模拟样品的近红外漫反射光谱，分别采用多元散射校正(MSC)、一阶微分和矢量归一化进行预处理，用偏最小二乘法(PLS)建立了乳糖的校正模型。在2090～2190nm波段内，采用一阶微分配合多元散射校正，模型相关系数达到0．99，预测集样本的标准偏差(RMSEP)达到0．045。以上研究结果为血糖测量的深入研究提供了理论和实践上的指导。

Research on Noninvasive Blood Glucose Measurement with Simulate Sample by NIR Spectroscopy

The article puts milk forward as simulate sample for blood glucose measurement. With the new simulate sample, the authors make deep researches on measuring method and wavelength choice in the field of noninvasive blood glucose measurement by near-infrared diffuse reflectance spectroscopy. Depending on the absorption coefficient of lactose, the authors choose the two wavebands from 1 560 to 1 750 nm and form 2 090 to 2 190 nm. The model was built by partial least-squares(PLS), several pretreatment methods (including multiplicative scatter correction, first derivative and vector normalization) and different spectral regions were compared. The best correlation values for lactose were 0.99 and the best RMSEP values for lactose were 0.045. The results provide us with valuable theory reference and practical experiences and lead us to next research on noninvasive blood glucose measurement.