- 期刊
應用振動光譜技術於黃耆、當歸之產地判別
Discrimination of Geographical Origins for Astragalus and Angelica Using Vibrational Spectroscopies
摘要
黃耆與當歸為臺灣常用之中藥材，常見用於中醫藥方、藥膳養生與日常保健等方面。然不同產地所生產之藥材，因其內含有效成分含量差異，可能影響市場價格，而有魚目混珠的疑慮。目前常用中藥產地檢測之方法包括元素分析、高效能液相層析儀，與氣相層析質譜等方式，其檢測精度佳，然樣本前處理較為繁複、檢測步驟、流程較長與花費較高。以振動光譜進行農產品成分分析行之有年，其樣本處理及數據取得相對簡單快速，故本研究將比較三種不同振動光譜，應用於黃耆與當歸產地判別之效果。試驗中藥材選用黃耆與當歸二者，各55項樣本。黃耆樣本產地來自中國甘肅、山西、四川與內蒙古四省；當歸則源自中國四川、甘肅與陝西等三省。三種測試振動光譜分別為近紅外光譜、傅立葉轉換紅外光譜和表面增強型拉曼光譜，取得光譜數據經預處理後，以主成分分析、K─近鄰演算法、隨機森林和支持向量機四種演算法建立產地判別模型。於最終實驗結果，以傅立葉轉換紅外光譜結合隨機森林方法建立出之模型最佳，於黃耆可達到97%，當歸可達到94%的判別效果。由於樣本收集不易故數量有限，初步研究成果可提供相關研究，及設備開發者作為前置試驗參考。
並列摘要
In Taiwan, astragalus and angelica are two herbs commonly used in traditional Chinese medicine. Authentication of products from different geographical origins is important because they are priced differently due to differing medicinal properties. Analytical methods for origin determination of Chinese herbs include the application of elemental analyzer, high performance liquid chromatography, and gas chromatography/mass spectrometry. Although these instruments can provide better accuracy, complicated pretreatment processes and higher cost hinder their use. Vibrational spectroscopies have been applied to analyze the composition of agricultural products for many years as a simple and fast way to acquire sample data. This study applies three different vibrational spectroscopies to develop models to determine the origins of astragalus and angelica. 55 astragalus and angelica samples each were collected. The astragalus samples were from the provinces of Gansu, Shanxi, Sichuan, and Inner Mongolia; the angelica samples were from the provinces of Sichuan, Gansu, and Shaanxi in China. Three vibrational spectroscopies, including near-infrared spectroscopy, Fourier-transform infrared spectroscopy, and surface enhanced Raman spectroscopy were applied. After acquiring sample spectra and preprocessing with a pretreatment algorithm, four methods-principal component analysis, k-nearest neighbor, random forest, and support vector machine-were adapted to develop discrimination models for the places of origin. Experimental results show that applying random forest to the Fourier-transform infrared spectra provides the best performance. Accuracies of origin discrimination could reach 97% for astragalus and 94% for angelica. Due to the difficulty of the sample collection, the sample numbers in this study were limited. The preliminary results provide an important foundation for further discussion in applying vibration spectroscopies to Chinese medicine and as a reference for instrument developers.