采用多光程长建模方法检测血液成分含量

为了提高近红外光谱血液成分含量分析模型的预测精度,利用多个光程长(optical path length,OPL)共同参与建模的方法进行血糖等6种血液成分的定量分析。通过微米位移机构实现不同光程长血液光谱的测量,由全自动生化分析仪给出生化成分分析结果,并出具化验单。采用偏最小二乘法(PLS2)进行血液的近红外光谱建模及预测。由于血液光谱存在显著的非线性特征,不同光程长的血液样本的等效吸收系数不同,同一波长不同光程长(0.20~1.25 mm)测得的血液光谱互不相关。主动把非线性特性作为一种测量手段引入,不再利用单个的最佳光程长建模,而是用各个血液组分对应的多个最佳光程长的近红外光谱同时参与建立校正模型,进行血液成分的分析预测。研究结果表明,多光程长建模方法用于血液成分含量分析,可提高血液成分校正模型的预测精度。

Application of Multi-Optical Path Length Modeling in the Quantitative Analysis of Human Whole Blood

In order to raise the predicting resolution of the calibration model,the quantitative analysis of six blood compositions by method of multi-optical path length is conducted.Blood NIRS is measured by the optical path length tunable system and the component s information is tested by the automatic biochemical analyzer.The partial least square method(PLS2) is used to model and predict the blood compositions.Due to the nonlinear characteristic relationship between the blood spectra and the concentration of the blood composition,the equivalent absorption coefficient of the blood sample of multi-optical path length is different.The NIR spectra of the human blood of multi-optical path length(from 0.20 mm to 1.25 mm) at same wavelength are irrelevancy.Here,taking nonlinear characteristic as a method actively,no modeling using one optical path length,but multi-optical path length,the concentrations of blood composition are predicted.It is revealed by the research that the quantitative analysis of whole blood by the method of multi-optical path length modeling method can raise the resolution of the blood calibration model.