温控近红外光谱用于葡萄糖的高灵敏检测

Temperature-Dependent Near-Infrared Spectroscopy for Sensitive Detection of Glucose

温控近红外光谱在混合体系定量分析以及分子间相互作用分析中已得到应用.本工作利用互因子分析方法(MFA)探索了温控近红外光谱技术测定血清样品中低浓度葡萄糖含量的可行性.测量了不同温度下葡萄糖含量分别为1.0~15.0 mmol/L和0.0~1.0 mmol/L的两组血清溶液的近红外光谱,并利用MFA进行了定量计算.结果表明,利用该方法可以实现样品中微量葡萄糖的准确定量,MFA提取的标准信号(SS)的相对强度和葡萄糖浓度之间的线性相关系数R分别为0.9923和0.9895,预测均方根误差(RMSEP)分别为0.35和0.07 mmol/L.

Temperature-dependent near-infrared(NIR) spectroscopy has been proposed and used in the quantitative analysis of multi-component mixtures and the understanding of the interactions in solutions. Mutual factor analysis(MFA) was developed, in our previous work, to detect glucose content in aqueous solutions and serum samples using the NIR spectra measured at different temperatures. The essence of the algorithm is to extract and compare the spectral component, named as standardized signal(SS), mutually contained in the spectral data of different samples. The relative quantity of SS can be used to build the calibration model for quantitative analysis. Furthermore, the spectral information of water can be used for the analysis, because the change of the water spectrum with temperature is a reflection of the change in glucose content. In this work, serum samples with low glucose concentration were prepared and measured at the temperature range of 30~60 ℃ with a step of 5 ℃. The feasibility of MFA in the quantitative determination of low concentration samples was further studied. Serum solutions with glucose content of 1.0~15.0 mmol/L and 0.0~1.0 mmol/L were prepared, respectively. Before calculation of MFA, continuous wavelet transform(CWT) was used to improve the resolution of the spectra. The results show that MFA can achieve an accurate quantification of the glucose content. The linear correlation coefficients(R) of the calibration models between the relative quantity of SS and the concentration of glucose are 0.9923 and 0.9895, respectively, and the root-mean-squared error of prediction(RMSEP) are 0.35 and 0.07 mmol/L, respectively. The relative error of predicted concentration of samples in the validation set obtained from the calibration model of samples with a concentration of 1.0~15.0 mmol/L are in the range of-12.00%~5.64%, which are in a reasonable level for clinical uses. Temperature-dependent NIR spectroscopy combined with MFA may be a potential way for detecting the micro-content components in complex aqueous systems.