中药砂仁“品与质”电子鼻判别模型的建立

摘要

Objective Fructus Amomi(Sharen) is derived from the dry ripe fruit of Amomum villosum Lour., A.villosum Lour. var. xanthioides T.L. Wu et Senjen and A.longiligulate T.L.Wu, which is widely utilized for its clinic effects on digestive system. However, Fructus Amomi from different species and habitats, possessing different quality, is difficult to identify. In this study, we aim to develop a simple, rapid and reliable method for authentication of Fructus Amomi. Methods Twenty-five batches of samples of Fructus Amomi were collected and electronic nose was introduced into analyzing their odor with multiple mathematical statistics methods. Naïve bayes network (NBN), radical basis function (RBF) and random forest (RF) were applied to establish different classifiers while BestFirst+CfsSubsetEval (BC) was used to screen the attributes for searching sensor array with higher contributions. Results Firstly, after attribute-screening via BC, the established discriminative models via NBN, RBF and RF could successfully identify genuine and non-genuine samples, presenting correct judging ratios of 78% and 84% through ten-fold cross validation and external test set validation, respectively. Besides, quantity predictive models were constructed as well. In case of content of bornyl acetate, one of the effective components in Fructus Amomi, values were higher than 3.5 mg/g and lower than 1.8 mg/g with sensor response of 0.04 and 0.03, respectively. Conclusion In this paper, quality discriminative model and quantity predictive model of Fructus Amomi were established via electronic nose and multiple mathematical statistics methods. It indicates that electronic nose could be a promising method for quality evaluation of Chinese material medica.%目的：建立一种快速、有效识别中药砂仁品质的科学评价方法。方法引入电子鼻建立中药砂仁的气味指纹图谱，以不同品种、不同产地的25批砂仁样品为研究对象，采用不同分类器建立其“品与质”的电子鼻判别模型，并基于盲样回判进行模型的验证。结果经十折交叉验证法、外部测试集验证法验证，所建立的气味判别模型正判率分别为78%和84%，表明电子鼻可以完成对阳春砂道地与非道地样品的定性分析；此外，传感器响应值为0.04和0.03的砂仁，预测其主要挥发性成分（乙酸龙脑酯）含量情况分别为＞3.5 mg/g和＜1.8 mg/g，表明电子鼻还可进行定量分析。结论电子鼻技术可作为一种快速、有效鉴别中药“品与质”的潜在分析方法。