| 中医证候的复杂系统建模及其与疾病的相关性研究 |
| 其他题名 | Study on complex system modeling of syndrome in Traditional Chinese Medicine and its association with disease
|
| 陈建新
|
| 2008-05-25
|
| 学位类型 | 工学博士
|
| 中文摘要 | 我的研究课题是基于国家重点基础研究发展计划（973计划）资助项目“证侯规范及其与疾病、方剂相关的基础研究”，课题研究的主要任务包括两部分内容. 一部分是利用复杂系统的熵方法和抽象神经自动机理论为中医的证候建模并规范疾病中的证候，为证候的客观存在提供数理基础.另外一部分是研究中医证候的微观子集, 在给定证候诊断的前提下, 寻找出诊断证候的相应的理化指标或分子层面的指标,为证候的现代化和客观诊断奠定基础. 针对课题的主要研究任务，我的研究课题主要完成以下四方面的工作，包括： 1 基于改进互信息的复杂系统熵聚类方法 基于信息熵的互信息是刻画相关的一种非线性度量，目前已经大量应用于理论和实践中。但是,传统的互信息不能区分开正相关和负相关.本文改进了传统的互信息,使之能区分正相关和负相关,在此基础上提出了熵聚类方法,它不但能对自变量聚类,同时也可以实现一些变量在不同的类中出现,把此方法运用到冠心病和肾衰临床数据中,提取了疾病中的中医证候,并利用因变量数据验证了复杂系统熵聚类方法,得到了很高的敏感性,此方法为证候的规范化奠定了数理基础, 2 随机神经网络系统 随机神经网络有很好的学习性能,但是用随机神经网络聚类目前还没有相关的工作,同时,随机神经网络中的相位转移的发现目前也没有相关的工作.本文提出了一类基于多重类随机神经网络的聚类方法并运用到冠心病临床数据中,提取了冠心病中的中医证候,并验证这些证候,也达到很高的敏感性.同时,本文发现了齐次随机神经中存在相位转移,临界点与神经元的神经机制—阈值有关.最后,本文用随机神经网络系统阐释了同证异病和异病同证. 3 有监督分类方法建立宏观子集与证候的关系 四诊信息和中医证候存在着非线性关系.诊断是治疗的前提,如何把这种非线性建立起来关系到证候诊断的准确性.本文利用五种非线性的有监督分类方法，即贝叶斯方法、神经网络方法、决策树方法、支持向量机方法和logistic回归方法建立四诊信息与证候之间的关系，以冠心病数据为例,利用80个四诊信息对痰瘀互阻证候分类的准确性进行比较学习,寻找出解决这类问题最优的分类器,为提高诊断准确率奠定数理基础. 4 有监督分类方法研究证候的微观子集 目前中医临床上还没有准确的中医证候与微观子集的关系，这成为证候的客观诊断的障碍。本文通过从三个不同的微观子集层次来建立客观指标与证候的关系.在给出证候诊断信息的前提下,利用有监督分类方法和回归方法建立炎症因子理化指标、蛋白组学和代谢组学与血瘀证证候的关系，为客观诊断证候提供了新方法。 最后，对取得的研究成果进行了总结，并展望了需要进一步开展的研究工作。 |
| 英文摘要 | I have finished following work. It includes: 1.Complex system entropy cluster method based on revised mutual information Entropy-based mutual information is a nonlinear measure of association, which is extensively applied in theory and practices. However, traditional mutual information can not discriminate positive association and negative association. In this paper, we revised the version of mutual information to discriminate positive association and negative one. Based on this, we present the entropy cluster method, which not only can cluster, but also realize that some variables appear in different patterns. We apply this method to Coronary Heart Disease(CHD) and Chronic Renal Failure (CRF) data, retrieve syndromes in Traditional Chinese Medine (TCM) and use diagnosis data to validate the cluster method, reaches a high sensitivity. The method eastablishes mathematical and physical basis for syndrome standardization. 2.Random neural network system Random neural network(RNN) has a good ability to learn, but there is no research effort on using RNN to cluster. Meanwhile, there is also no phase transitions detected in RNN. In this paper, we propose a multiple class random neural network-based cluster method and apply it Coronary Heart Disease data, retrieve syndromes in TCM in CHD and validate them. The method reaches a high sensitivity. Meanwhile, phase transitions is detected in homogeneous random neural network, the critical point is associated with neural mechanism-threshold of neuron. Finally, we use RNN to interpret different diseases with same syndrome and different syndromes with same disease. 3.Supervised classification method to build relation between syndrome and macroscopical subset Macroscopical subset means four diagnosis information. It has a nonlinear relation with syndrome in TCM. Diagnosis is the precondition of treatment. How to build the nonlinear relation plays a key role in diagnosing syndrome accurately. In this paper, we use five supervised classification methods. i.e. Bayes methods, Neural Networks, Decision Trees, Support vector machine and logistic Regression to build nonlinear relation between syndrome and four diagnosis information. By CHD data, we use 80 variables to classify an important syndrome- TanYuHuZu to find the classifier with highest accuracy, which eastablishes the mathematical and physical basis for diagnosing syndrome accurately. 4.Supervised classification method to study microcosmic subset and syndrome There is no accurate relation between microcosmic subset and syndrome, which blocks the objective diagnosis of syndrome. In this paper,we build relation between objective specifications and syndrome from three levels. Physical and chemical specifications, proteome and metabonomics. The work provides new method for diagnosing syndrome objectively. At last, summarize the research result and prospect the future research work. |
| 关键词 | 复杂系统
信息熵
相关
中医
证候
有监督方法
随机神经网络
宏观子集
微观子集
Complex System
Information Entropy
Association
Traditional Chinese Medicine
Syndrome
Supervised Method
Random Neural Network
Macroscopical Subset
Microcosmic Subset
|
| 语种 | 中文
|
| 文献类型 | 学位论文
|
| 条目标识符 | http://ir.ia.ac.cn/handle/173211/6073
|
| 专题 | 毕业生_博士学位论文
|
推荐引用方式
GB/T 7714 |
陈建新. 中医证候的复杂系统建模及其与疾病的相关性研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2008.
|
修改评论