近红外光谱快速测定糖化血红蛋白的模型优选

Model Optimization for Rapid Determination of Glycosylated Hemoglobin by NIR Spectroscopy

糖化血红蛋白(HbA1c)是评价长期血糖控制的基本指标,与糖尿病及其并发症密切相关,目前它的检测方法都较为繁琐。近红外光谱(NIR)技术具有检测快速且无污染的优势,但它不能直接分析HbA1c这一相对性指标。针对NIR技术不能直接分析HbA1c的问题,采用间接分析方法,即先定量分析总血红蛋白(Hb)和糖化血红蛋白的绝对含量(Hb·HbA1c),然后根据这两个指标的预测结果算出HbA1c的预测值。为了实现高精度定量分析,采用移动窗口偏最小二乘(MW-PLS)和等间隔多元线性回归(EC-MLR)两种方法进行特征波长筛选。结果表明:两种方法得到的HbA1c的检验预测均方根误差(V_SEP)和预测相关系数(V_RP)分别为0.44%、0.918和0.50%、0.908,两种方法均取得了比较好的预测效果。因此,采用NIR技术间接分析HbA1c的建模方法是可行的,这为糖尿病的快速筛查、诊断、预防和控制提供了有效的新途径。

Glycated hemoglobin(HbA1c)is a basic indicator for evaluating long-term blood sugar control.It is closely related to diabetes and its complications.At present,its detection methods are relatively complicated.Near-infrared spectroscopy(NIR)technology has the advantages of rapid detection and pollution-free,but it cannot directly analyze the relative index of HbA1c.In view of the problem that NIR technology cannot directly analyze HbA1c,an indirect analysis method is adopted,that is,first quantitatively analyze the absolute content of total hemoglobin(Hb)and glycosylated hemoglobin(HboHbA1c),and then calculate the predicted value of HbA1c based on the predicted results of these two indicators.In order to achieve high-precision quantitative analysis,two methods of moving window partial least squares(MW-PLS)and equal interval multiple linear regression(EC-MLR)are used for characteristic wavelength screening.The results show that the root mean square error(V_SEP)and prediction correlation coefficient(V_RP)of HbA1c obtained by the two methods are 0.44%,0.918 and 0.50%,0.908,respectively.Both methods have achieved relatively good prediction results.Therefore,it is feasible to use NIR technology to indirectly analyze HbA1c modeling method,which provides an effective new way for rapid screening,diagnosis,prevention and control of diabetes.