Lan Zhen-Wei, Ji De, Wang Shu-Mei, Lu Tu-Lin, Meng Jiang
Engineering Technology Research Center for Chinese Materia Medica Quality of Universities in Guangdong Province,Key Laboratory of Digital Quality Evaluation of Chinese Materia Medica, State Administration of Traditional Chinese Medicine, School of Traditional Chinese Medicine, Guangdong Pharmaceutical University Guangzhou 510006, China.
School of Pharmacy, Nanjing University of Chinese Medicine Nanjing 210023, China.
Zhongguo Zhong Yao Za Zhi. 2020 Aug;45(16):3863-3870. doi: 10.19540/j.cnki.cjcmm.20200523.304.
This study aimed to establish a rapid and accurate method for identification of raw and vinegar-processed rhizomes of Curcuma kwangsiensis, in order to predict the content of curcumin compounds for scientific evaluation. A complete set of bionics recognition mode was adopted. The digital odor signal of raw and vinegar-processed rhizomes of Curcuma kwangsiensis were obtained by e-nose, and analyzed by back propagation(BP) neural network algorithm, with the accuracy, the sensitivity and specificity in discriminant model, correlation coefficient as well as the mean square error in regression model as the evaluation indexes. The experimental results showed that the three indexes of the e-nose signal discrimination model established by the neural network algorithm were 100% in training set, correction set and prediction set, which were obviously better than the traditional decision tree, naive bayes, support vector machine, K nearest neighbor and boost classification, and could accurately differentiate the raw and vinegar products. Correlation coefficient and mean square error of the regression model in prediction set were 0.974 8 and 0.117 5 respectively, and could well predict curcumin compounds content in Curcuma kwangsiensis, and demonstrate the superiority of the simulation biometrics model in the analysis of traditional Chinese medicine. By BP neural network algorithm, e-nose odor fingerprint could quickly, conveniently and accurately realize the discrimination and regression, which suggested that more bionics information acquisition and identification patterns could be combined in the field of traditional Chinese medicine, so as to provide ideas and methods for the rapid evaluation and stan-dardization of the quality of traditional Chinese medicine.
本研究旨在建立一种快速、准确鉴别广西莪术生品与醋制品的方法，以预测姜黄素类成分含量，进行科学评价。采用一套完整的仿生识别模式。利用电子鼻获取广西莪术生品与醋制品的数字气味信号，采用反向传播（BP）神经网络算法进行分析，以判别模型中的准确率、灵敏度和特异性、回归模型中的相关系数以及均方误差为评价指标。实验结果表明，神经网络算法建立的电子鼻信号判别模型在训练集、校正集和预测集的三项指标均为100%，明显优于传统的决策树、朴素贝叶斯、支持向量机、K近邻和提升分类法，能够准确区分生品和醋制品。预测集回归模型的相关系数和均方误差分别为0.974 8和0.117 5，能够较好地预测广西莪术姜黄素类成分含量，体现了模拟生物特征模型在中药分析中的优越性。通过BP神经网络算法，电子鼻气味指纹图谱能够快速、便捷、准确地实现判别与回归，提示在中药领域可结合更多的仿生信息获取与识别模式，为中药质量的快速评价与标准化提供思路和方法。
