为考察不同类型光谱信息用于银杏叶质量快速分析的适应性, 收集了58个银杏叶样品, 采用高效液相色谱方法(HPLC)测定其黄酮及内酯类活性成分的含量作为定标和检验样本的因变量(y)值, 测定各样品的紫外、 近红外光谱及包含紫外、 可见及近红外信号的多源复合光谱信息作为样本的自变量(x)值; 分别采用偏最小二乘回归(PLSR), 以及根据待测样本在自变量空间最近邻K个样本与待测样本间的相互关系去预测其因变量值的KNN保形映射(KNN-KSR)方法, 建立银杏叶活性成分的光谱定量分析模型, 比较各光谱模型下检验集样本实测值与模型值的相关系数(R)、 均方根偏差(RMSEP)、 平均相对误差(MRE)。 结果表明PLSR方法所建立的三类光谱模型的各项指标均不及KNN-KSR方法、 且其紫外光谱模型的结果极差; 而采用KNN-KSR方法根据三类光谱信息预测银杏叶中黄酮、 内酯类成分时, R基本能达到0.8、 RMSEP分别小于0.05与0.025且其平均相对误差均在8%以下。 采用KNN-KSR方法根据紫外、 近红外及多源光谱信息均可实现对银杏叶中四类黄酮醇苷成分及三类内酯成分含量的快速分析, 突破了现有工作只是基于PLSR方法、 根据近红外光谱信息对银杏叶总黄酮醇苷进行定量分析的局限; 利用紫外和多源复合光谱信息及KNN-KSR方法进行银杏叶中黄酮醇苷及内酯类成分的快速检测, 为银杏叶质量分析提供了更多的方法和选择。 多源复合光谱仪具有体积小、 成本低, 便携的优点, 非常适合银杏叶药材现场采购的快速检测及后续产品的质量分析与监控。Infrared and Multi-Source Composite Spectral Information

In order to study the adaptability of using different kinds of spectra to analyze the quality of Ginkgo biloba leaves quickly, 58 samples of Ginkgo biloba leaves were collected. The contents of the active components of flavonoid glycosides and terpene lactones were determined as dependent variables (y) by high performance liquid chromatography (HPLC), and the independent variables (x) included ultraviolet (UV), visible and near infrared spectra signals. Quantitative analysis models of flavonoids and lactones in Ginkgo biloba leaves were established by partial least square regression (PLSR) and an innovative method of keeping a same relationship between X and Y space (KNN-KSR method for short). The method predicted dependent variables based on the object’s independent variables and the relationship between the object and its K nearest neighbors in independent variable space. Correlation coefficient R between the measured values and the model values, root mean square error of prediction (RMSEP), and the average relative error of the prediction (MRE) were applied to evaluate the models. All evaluated indicators of PLSR models based on three kinds of spectral information were inferior to those of KNN-KSR method, and the results of PLSR models based on UV spectra were very poor; However, when KNN-KSR method was used to predict the flavonoids and lactones in Ginkgo biloba leaves based on three kinds of spectral information, R was higher than 0.8; RMSEP of flavonoids and lactones were less than 0.05 and 0.025, respectively; MRE of flavonoids and lactones content were below 8%. UV, NIR and multi-source composite spectral information combing KNN-KSR method could achieve rapid analysis of four kinds of flavonoid glycosides and three kinds of terpene lactones in Ginkgo biloba leaves. The present work broke through the limitation of existing work that only analyzed total flavonoids in Ginkgo biloba leaves by PLSR method based on NIR; The proposed new ideas to rapidly determine flavonoids and lactones in Ginkgo biloba leaves using UV and multi-spectral information by KNN-KSR method provided more available methods and choices for the quality analysis of ginkgo biloba leaves. The multi-source composite spectrometer, which can provide spectral information of various types, is portable, of small volume and low cost. It is very suitable for the rapid detection of on-the-spot Ginkgo biloba leaves acquisition and follow-up product quality analysis and monitoring.