振動光譜技術於黃耆、當歸之產地判別與農藥殘留分析之應用

在中藥材所種植產地，會因其生長環境，進而影響其有效比例的組成，而道地的藥材不論藥性或價格，都有所其差異性。另一方面，台灣的中藥材以進口為主，官方為維護中藥的食品安全，故衛生福利部設立邊境查驗制度，目前使用的農藥殘留檢驗設備精確度佳，但儀器昂貴及檢驗流程耗時等缺點，使其無法普及於大量樣本檢測，而振動光譜(Vibrational spectroscopy)可用於快速對於樣本進行定性或定量分析。本研究透過振動光譜技術，分別使用近紅外光譜(Near infrared spectroscopy, NIR)、傅立葉轉換紅外光譜(Fourier-transform infrared spectroscopy, FTIR)和表面增強拉曼(Surface enhanced Raman spectroscopy, SERS)結合多變量分析，對於中藥材進行產地判別，並利用SERS建立農藥特徵之拉曼光譜圖庫，對其進行農藥殘留量之半定量分析。產地判別產地判別選用黃耆樣本產地包含四川、內蒙古、甘肅，及山西亦共55件；當歸樣本產於陝西、甘肅與四川共55件。農藥殘留芬選擇試驗對象為常見農藥之殺菌劑貝芬替，分析中藥材樣本選用黃耆。產地判別結果中，以隨機森林(Random forest, RF)搭配FTIR，於黃耆和當歸產地中，皆得100%準確度，然測試樣本較少，因此持保留看法。而K-近鄰演算法(K-nearest neighbor algorithm, KNN)和支持向量機(Support vector machine, SVM)可得較客觀分析結果，於黃耆產地判別中，SERS模型分辨結果最佳，得79-82%之準確度，而在當歸產地判別中，由FTIR建立模型最佳，得73-76%之準確度。在農藥殘留檢驗上，可判別出貝芬替之七支特徵峰值位置，分別位於624, 771, 1003, 1222, 1269, 1459和1514 cm-1處。而於貝芬替殘留於黃耆之複合樣本中，以3.34, 4.29, 8.75及13.41 ppm等四種農藥殘留濃度之樣本進行測試，可於771, 1003, 1222及1269 cm-1等四峰值位置建立濃度檢量線。本研究成功透過振動光譜於黃耆和當歸進行產地判別，並使用表面增強拉曼光譜以黃耆中之貝芬替殘留量進行試驗，成功確認該殺菌劑之拉曼指紋圖譜、特徵峰值位置，及建立藥劑殘留之半定量濃度檢量線。此方法有機會為中藥材中之農藥殘留檢驗，提供一較快速且降低檢驗成本的替代方案。

關鍵字

振動光譜；農藥殘留；中藥材；產地判別

並列摘要

Chinese herb medicines from different origins may be priced differently due to the active ingredients, which affects the medicinal properties. On the other hand, most of the herbs in Taiwan are imported. To ensure the safety of imported herbs, the Ministry of Health and Welfare inspects pesticide residues of the herbs using high-end instrument. Although the inspection can detect pesticide accurately, it cannot be applied to large number of samples because of long processing time and high operation cost. Vibrational spectroscopy can be effectively used for qualitative or quantitative analysis. In this study, near-infrared spectroscopy (NIR), Fourier-transform infrared spectroscopy (FTIR), and surface enhanced Raman spectroscopy (SERS) were combined with multivariate analysis respectively to identify Chinese herb medicines origins. Besides, SERS was used to develop a preliminary semi-quantification methods for screening pesticide residuals of imported herbs. Two herbs, Astragalus and Angelica were selected as the testing objects. Astragalus samples were cultivated in four China provinces (Sichuan, Inner Mongoria, Gansu, and Shanxi) and Angelica samples were cultivated in in three China provinces (Shaanxi, Gansu, and Sichuan). In pesticide residual analysis, Astragalus was selected as the testing object and carbendazim was used as the examinable target of pesticide. In the section of origin identification, FTIR combined with random forest (RF) in the origins identification shown the best performance of Astragalus and Angelica, and both of the identification rates reached 100%. However, the number of samples in the testing sets were small so that the result was reserved. The objective analysis results were obtained by K-nearest neighbor algorithm (KNN) and support vector machine (SVM). SERS shown the best performance in the origins identification of Astragalus, and identification rates were in the range of 79-82%. FTIR is the best one in the origins identification of Angelica, and identification rates were in the range of 73-76%. In the section of pesticide residual analysis, seven characteristic peak positions of carbendazim were identified, which were located at 624, 771, 1003, 1222, 1269, 1459, and 1514 cm-1, respectively. Based on the test result from four concentrations of 3.34, 4.29, 8.75 and 13.41 ppm of carbendazim-astragalus mixed samples, there were four peaks – 771, 1003, 1222, and 1269 cm-1, could be selected to develop the calibration curves to semi-quantify the carbendazim residues in astragalus. This study successfully applied vibrational spectroscopies to identify astragalus, and angelica origins, and applied SERS to develop Raman fingerprint spectra for carbendazim, identify its characteristic peaks, and build the calibration curves for the residue semi-quantification in astragalus. It is a promising method to provide a faster and lower cost way to examine the pesticide residues in Chinese herb medicines.