近红外光谱技术结合竞争自适应重加权采样算法用于中药定量分析

发布时间：2019-05-10 06:31

【摘要】：目的:基质复杂、谱带重叠严重,影响了中药近红外定量模型的准确性。为解决以上问题,探讨竞争自适应重加权采样(Competitive adaptive reweighted sampling,CARS)变量筛选方法在中药材、中药提取物和中成药的定量分析中的应用。方法:采集葛根药材、葛根提取物和愈风宁心滴丸的近红外漫反射光谱,测定葛根素含量。分别优化光谱前处理方式,剔除奇异样本后,运用CARS法筛选出的相关变量,建立偏最小二乘法(PLS)校正模型。结果:原料、中间体和制剂的定量模型交互验证均方差(RMSECV)分别为0.35%,1.76%,0.54%,与基于全光谱建立的模型比较,原料、中间体和制剂的CARS-PLS模型的预测准确度均有提高。结论:竞争自适应重加权采样变量筛选方法可以提高模型的预测能力,并有效简化运算过程,为中药的快速、无损检测提供了新的思路。
[Abstract]:Objective: complex matrix and overlapping spectral bands affect the accuracy of NIR quantitative model of traditional Chinese medicine (TCM). In order to solve the above problems, the application of competitive adaptive weighted sampling (Competitive adaptive reweighted sampling,CARS) variable selection method in quantitative analysis of Chinese medicinal materials, extracts of traditional Chinese medicine and proprietary Chinese medicines was discussed. Methods: the content of puerarin was determined by near infrared diffuse reflectance spectroscopy (NIR) of Radix Puerariae, Pueraria extract and Yufeng Ningxin dropping pills. After optimizing the spectral pre-processing mode and eliminating the singular samples, the partial least square (PLS) correction model was established by using the correlation variables selected by the CARS method. Results: the mean variance (RMSECV) of raw materials, intermediates and preparations were 0.35%, 1.76% and 0.54%, respectively. compared with the model based on full spectrum, the mean variance of raw materials was 0.35%, 1.76% and 0.54%, respectively. The prediction accuracy of CARS-PLS model for intermediates and preparations has been improved. Conclusion: the competitive adaptive re-weighted sampling variable screening method can improve the prediction ability of the model and simplify the operation process effectively, which provides a new way of thinking for the rapid and nondestructive testing of traditional Chinese medicine.
【作者单位】：中国食品药品检定研究院;北京同仁堂股份有限公司;
【分类号】：R284.1

【相似文献】