利用近红外血流容积光谱相减方法无创伤定量分析人体血液生化成分时,获得的在体血液光谱对应的样品光程是不确定的。为研究样品间的光程差异对定标模型预测精度的影响,配制了30份模拟血清样品,利用可变光程样品池,采用傅里叶光谱仪分别测量了相同光程和不定光程两组样品的近红外光谱。以分析血清白蛋白成分为例,对两组样品的定标模型精度进行比较,结果显示不定光程组的模型精度与相同光程组的模型精度相比明显下降,交叉检验标准差(RMSECV)由110.0mg/dL增大至156.0mg/dL。采用多元散射校正算法对上述光谱数据校正后,RMSECV降低至98.1mg/dL。对比分析处理前后两组模型的精度,证实了采用适当的预处理方法能够有效校正不定光程引起的光谱误差,提高模型预测精度。 When using near-infrared flow volumetric spectroscopy subtraction method to quantitatively analyze human blood biochemical composition without invasiveness, the sample optical path length corresponding to in vivo blood spectrum obtained is indeterminate. In order to study the influence of the optical path difference between samples on the prediction accuracy of the calibration model, thirty simulated serum samples were prepared. Two groups of the same optical path and the indefinite optical path length Near Infrared Spectroscopy of Samples. Taking the analysis of serum albumin as an example, the calibration accuracy of two groups of samples was compared. The results showed that the precision of the model group was significantly lower than that of the same optical group. The standard deviation of cross test (RMSECV) From 110.0 mg / dL to 156.0 mg / dL. RMSECV was reduced to 98.1 mg / dL using the multivariate scatter correction algorithm to correct the above spectral data. The accuracy of the two models before and after the comparison was analyzed. It was confirmed that the appropriate preprocessing method could effectively correct the spectral error caused by the indefinite optical path and improve the prediction accuracy of the model.