【作者】 刘雪松；

【导师】 程翼宇；

【作者基本信息】 浙江大学 ， 药物分析学， 2005， 博士

【摘要】 目前,中药产业中缺乏能够全面反映中药质量的快速、有效的分析方法,这导致了中药产品质量不稳定可控,严重制约我国中药制药工业的发展。因此,在当前药物分析学研究领域急待攻克的诸多关键科学问题及共性关键技术中,一项十分重要的研究任务是研发中药材和中成药产品快速分析技术,建立中药生产过程在线分析方法学,实现中药制药过程质量监测和优化控制,从而保证中药产品质量的稳定性和均一性。近红外光谱(Near Infrared Spectroscopy,NIRS)是一种先进的计算机辅助快速分析测试技术,有望解决中药快速检测的难题。由于中药化学物质体系的复杂性,在实际应用中,使用常规的NIRS校正建模或模式分类等方法往往不能取得理想的定量或定性分析结果,导致其成为阻碍中药NIRS分析技术应用发展的瓶颈。为此,有必要研究中药NIRS计算分析方法学,特别是发展NIRS非线性校正建模、光谱信息特征提取、化学模式信息处理以及模糊分类辨识等方法,发展形成中药材和中成药产品快速分析新技术,实现中药生产全过程质量控制,这对于推进中药现代化进程具有重大科学意义和显著的实际应用价值。 本文根据中药质量控制关键环节中的不同要求,提出系列相关计算分析方法用于解决NIRS分析技术在中药领域应用的技术难题,探讨利用NIRS分析技术提高中药分析和质量控制水平的途径,以发展形成中药质量快速分析新方法。本文主要研究内容包括: 1.针对复杂中药材质量辨识难题,提出基于模糊神经元分类器(FNN)的NIRS分析方法用于对天然中药材质量的快速鉴别。阿胶实验结果表明,所提出的方法对分类界面含糊的复杂中药材体系辨识的学习能力和外推能力强,精度高,大大优于参比经典的BP-ANN,是一种简便、无损、快速有效的方法。 2.针对目前中药注射剂中缺少快速质量鉴别方法的现状,利用NIR透射光谱分析方法基于自组织映射神经网络(SOM)和模糊神经网络(FNN)实现了对不同厂家参麦注射液质量的快速鉴别,所提方法是适合常规分类困难的中药注射剂质量鉴别的有效工具,可望发展成为中药质量类别快速测定方法。 3.通过中药活性成分含量测定中NIRS分析方法学的研究,提出利用神经更多还原

【Abstract】 The modernization and internationalization of Traditional Chinese Medicine (TCM) are blocked by its poor consistence and instability. One of the main reasons is the absences of fast and efficient analytical methods for quality control of TCM. Among the key technologies need to be investigated in pharmaceutical analysis field, one of main tasks is to develop fast analytical technologies and methodologies for TCM, establish on-line analytical methodologies for TCM process, realize the optimal control and evaluation of the quality of TCM in its pharmaceutical process and ensure consistence and equality of its quality. Near infrared spectroscopy (Near Infrared Spectroscopy, NIRS) is an advanced computer aided fast technique, and it could be a promising solution to solve the problem of fast measurement for TCM. Since the composition of TCM is quite complex, the traditional methods for modeling and pattern recognition sometimes could not get the ideal quantitative or qualitative results. So it is difficult to use NIR spectra practically in pharmaceutical industry of TCM. Therefore, it is necessary to carry on the studies on computational methodologies of NIRS, especially develop methods for non-linear modeling, feature extraction, chemical pattern information processing, fuzzy classification and etc., develop the new analytical technologies for pharmaceutical process of TCM and realize the quality control for the whole pharmaceutical process of TCM. Such studies have great scientific meaning and remarkable practical value in progress of TCM modernization.In this paper, some computational analytical methods have been developed according to the content and demand of quality control of TCM. These methods could be used to solve the application problems of using NIR technologies for TCM. The approaches improving the level of analysis and quality control of TCM are discussed, and could be used to develop new fast analytical methodologies of quality control for TCM. The main results obtained from the paper are listed as following:1. To distinguish quality of complex TCM, a fast analytical method using NIRS based on an adaptive fuzzy-neural classifier (FNN) is proposed. The experiment results on Colla CoriiA sini show the proposed method has strong learning and extrapolated abilities in distinguishing the quality of TCM with ambiguous boundary. Compared with method using traditional BP-ANN, its identification result is more accurate. It is showed that the method proposed is fast, convenient, non-destructive, and effective.2. In order to solve the problem of lacking fast method for evaluating quality of TCM injections, the methods using near infrared transmission spectroscopy based on a self-organizing mapping neural network (SOM) and FNN are developed respectively. Distinguishing different manufacturers of Shenmai injection is investigated as an example to test the performance of the methods. The methods developed could be used as new approaches to classify the complex TCM which has serious non-linear phenomena and could not be easily classified by traditional methods.3. Studies on the methodologies of determinating active components of TCM using NIR spectra analysis are systematically carried on. Methods using artificial neural network (ANN) and support vector machine (SVM) combined with different data preprocess methods are proposed to solve the non-linearity in the NIR reflectance spectra of Panax notoginseng root herb. These proposed methods effectively reduce the predictive error and are the suitable tools for non-linear modeling.4. Studies on applications of on-line NIRS technologies and computational analytical methodologies are carried on. Analytical models are successfully established to make a fast on-line measurement of Radix salviae miltiorrhizae’s extraction in pilot-scale and industrial process respectively. The BP-ANN models are established to solve the non-linearity in the measured NIR spectra and improved the quantaitive analytical accuracy. A method using the absolute distance of standard deviation of the on-line measured spectra is developed to make fast estimation of process ending point during the optimization of TCM process. The results well inosculates the real industrial process. These methods developed could be applied inthe whole production process of TCM.In conclusion, the studies on computational analytical methodologies of NIRS for TCM help to improve the level of theoretically research and application of NIR technologies. These developed analytical methods could solve the problem of absence of fast methods of quality control for TCM. The results indicate that all the developed methods could be extent to fast quality identification and on-line measurement of TCM, having applicability and innovabality. It provides new approaches in strengthening the quality control in production process and improving the analytical level of TCM.更多还原