基于高光谱分析的淫羊藿药用成分快速检测研究  

Rapid Measurement of the Pharmacological Active Constituents in Herba Epimedii Using Hyperspectral Analysis Technology

作　　者：姜庆虎[1] 刘峰[1] 于东悦 罗惠 梁琼[3] 张燕君[3] JIANG Qing-hu;LIU Feng;YU Dong-yue;LUO Hui;LIANG Qiong;ZHANG Yan-jun(Key Laboratory of Aquatic Botany and Watershed Ecology,Wuhan Botanic al Garden,Chinese Academy of Sciences,Wuhan 430074,China;University of Chinese Academy of Sciences,Beijing 100049,China;Key Laboratory of Plant Germplasm Enhancement and Specialty Agriculture,Wuhan Botanical Garden,Chinese Academy of Sciences,Wuhan 430074,China)

机构地区：[1]中国科学院武汉植物园,中国科学院水生植物与流域生态重点实验室,湖北武汉430074 [2]中国科学院大学,北京100049 [3]中国科学院武汉植物园,中国科学院植物种质创新与特色农业重点实验室,湖北武汉430074

出　　处：《光谱学与光谱分析》2022年第5期1445-1450,共6页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金面上项目(32071675,31670346);湖北省研发创新平台建设项目(2018AFA164)资助。

摘　　要：中药材淫羊藿富含朝霍定和淫羊藿苷等黄酮类化合物,具有滋阴补肾、提高免疫力等功效,有较大的药用价值。当前,面对生产及育种过程中批量样品快速、无损检测需求的增加,传统的化学分析方法难以满足需要,而高效、廉价的现代高光谱分析技术备受青睐。但受制于光谱数据谱峰重叠及噪声的干扰,全波段光谱分析建模存在模型精度不高和运行效率低的问题。利用便携式地物光谱仪器获取淫羊藿可见-近红外光谱数据,借助遗传算法(GA)特征波段选择方法剔除无关波段,并与偏最小二乘回归(PLSR)分析建模技术结合,构建淫羊藿药用组分(朝霍定A、朝霍定B、朝霍定C和淫羊藿苷)高光谱GA-PLSR校正模型,探讨淫羊藿药用组分含量高效分析预测的可行性,并挖掘获取淫羊藿品质鉴定的重要光谱响应波段。结果表明:高光谱分析结合化学计量学在淫羊藿有效药用组分的快速无损检测方面具有相当大的潜力。与全波段PLS R校正模型相比,通过GA迭代优化,参与建模的有效光谱数据得到简化,GA-PLSR模型的测量精度和稳定性得到明显提升。主要表现在交叉验证的决定系数(R^(2)CV)得到明显提高,交叉验证的均方根误差(RMSE_(CV))普遍降低。其中,四种药用组分校正模型的R^(2)CV分别从0.645,0.720,0.718和0.642提升为0.671,0.835,0.782和0.796;同时,其对应的RMSE_(CV)值分别由2.102,2.896,21.069和1.221降为2.071,2.230,18.656和0.912。此外,明确了红边波段690~740 nm以及420 nm附近波段为淫羊藿药用组分朝霍定A、朝霍定B、朝霍定C和淫羊藿苷光谱鉴别分析的重要响应波段。该研究为高光谱技术淫羊藿品质准确高效鉴定和光谱传感器的波段设计提供一定的理论依据。Herba Epimedii contains high amounts of flavonoi ds,such as epimedin and icariin,which are efficient in tonifying kidney and im proving immunity.Nowadays,various chemical analysis methods have been applied to measure the flavonoid content of Herba Epimedii.However,these traditional m ethods are destructive,time-consuming,and costly and cannot meet the requirem e nts of massive samples analysis in pharmaceutical production and plant breeding.As a rapid and effective tool for quantitative determination and process monito ring,modern hyperspectral analysis technology has earned more and more concerns.However,for the full-range spectra,the existence of insignificant and irrel evant spectral variables can weaken the calibration models’accuracy and effici ency.Therefore,the spectral variables selection is essential to improve the pe rformance of the final models by eliminating the uninformative bands.In this st udy,the partial least squares regression(PLSR)coupled with the genetic algori thm(GA)variables selection procedure,namely GA-PLSR,was used to estimate ep imedin A,epimedin B,epimedin C,and icariin content in Herba Epimedii.This pa per aims to explore the feasibility of hyperspectral analysis technology in the measurement of the pharmacologically active constituents in Herba Epimedii and f urther explore their important spectral response bands.The results show thatthe hyperspectral analysis technology combined with chemometrics exhibited consider able potential for rapid and nondestructive assessment of Herba Epimedii.When c ompared with full-spectrum PLSR models,GA-PLSR models could improve the accur acies and robustness of epimedin A,epimedin B,epimedin C,and icariin content measurements(with R^(2)CV values increased from 0.645,0.720,0.718,and 0.642 to 0.671,0.835,0.782,and 0.796,and with RMSE_(CV) values declined from 2.102,2.896,21.069,and 1.221 to 2.071,2.230,18.656,a nd 0.912,respectively).Besides,we found some feature wavelengths,mainly aro und 690~740 and 420 nm,which play important roles in detecting p

关 键 词：淫羊藿 药用组分 高光谱技术 遗传算法 偏最小二乘回归 重要波段 

分 类 号：O657.3[理学—分析化学]