【作者】 周玉；

【导师】 汪仁煌；

【作者基本信息】 广东工业大学 ， 检测技术与自动化装置， 2007， 硕士

【摘要】 随着中医国际化以及全球数字化信息的发展，对中医数字化、标准化的研究走上日程，为了促进中医的继承和发展，对四症客观化的要求日见迫切。舌诊是中医中最常用和最具有临床应用价值的诊法之一。随着图像处理、模式识别、人工智能等计算机技术的日益完善，舌诊的自动化成为新兴的研究内容，它可以避免以前的中医舌象诊断的非量化、不确定性等缺陷，有很高的理论研究与应用价值。本文的研究正是围绕这个方向展开，本文的目的就是为中医数字化做一些有意义的探索。医学诊断是医生根据病人的病史、症状等信息得出诊断结果的过程。如果医学诊断实现自动化，可以解决医学专家缺少、节约资金、减少病人的痛苦等问题以及保留珍贵的医学经验和资料，给人类带来很多好处。但是由于医疗诊断是一个复杂系统，很难用形式化表示，也就是说用一般的逻辑推理方法进行设计的效果并不令人满意，所以也就很难实现数字化和自动化。在中医舌诊诊断中尤其如此。医生诊断过程实质是由收集到的各种症状，通过分析获得诊断结果的过程。因而可以把这个过程看作是一个非线性映射过程，用神经网络来实现。人工神经网络模拟人脑的认知特点，通过对大量实例样本的学习、训练，从而掌握难以用解析形式表达的知识。由于神经网络具有分类、学习、记忆、联想、计算、容错、实时性等特性和功能，应用于诊断也是较为合适的。本文以中医舌诊为背景，提出了神经网络的模糊控制技术和竞争神经网络相结合的诊断方法。在该方法中，用神经网络实现模糊推理，用竞争神经网络实现诊断。该方法不仅具有模糊逻辑和神经网络两者的优点，而且竞争网络的应用使得系统有了在线学习的能力。作者在本论文中主要完成了以下两项工作：1、自组织竞争神经网络的应用。自组织竞争神经网络采用无监督的学习过程，它比有监督学习更接近生物神经系统，无监督学习按照预订的规则和输入模式，重复修改系统中各连接权值，直至形成一种全局有序的结构形态。自组织竞争神经网络通过对网络权值的调整，使所有矢量都在输入矢量空间相互隔离，形成了各自代表输入空间的一类模式，从而实现聚类功能。利用自组织竞争网络，对舌诊数据的聚类、识别分析进行了理论研究，并给出一个具体的仿真实验，通过实验说明竞争神经网络在中医舌诊应用中能够达到较好的分类识别的功能。2、模糊逻辑与神经网络技术的融合应用。模糊逻辑是一种处理非线性和不确定问题的有力工具，它比较适合表达模糊或定性的知识，但缺乏自学习和自适应能力。把神经网络学习机制引入到逻辑推理中，使传统的逻辑推理不仅具有逻辑思维及语言表达能力，而且具有自学习和联想能力。通过神经网络和模糊逻辑的融合技术，研究了一种新的中医诊断的方法，实现了专家诊断推理过程的模拟。通过实验表明该方法是有效和可行的，同时，该方法的应用，为医疗诊断过程提供了一种新的途径。更多还原

【Abstract】 With the international development of Traditional Chinese Medicine (TCM) and the development of global digital information, the digital and standardized study of TCM get more and more important, we must carry out the objective research on four diagnostic methods of TCM: inspection,auscultation and olfaction, inquiry pulse-feeling and palpation, in order to promote the succession and development of TCM. Tongue Diagnosis is one of the most common and the most valuable diagnoses on clinical practice in TCM. With the being perfected computer techniques such as image processing, pattern recognition, artificial intelligence etc., the automation of the Tongue Diagnosis becomes the burgeoning research content. It can avoid some limitations such as non-quantities and uncertainty in the old Tongue Diagnosis, so it has high value of theoretical research and practice. This paper just studies in this direction and the objective of this paper is that points out a new path and direction for the further study of the objective tongue inspection of TCM.Medical diagnosis refers to the process of identifying a disease of a patient by considering his medical history, symptoms and other sign, Diagnosis automation has many advantages. It can meet the needs of the medical service on far way areas where medical experts are in great need of, furthermore it can save much money, relieve patients’ pains, keep precious medical experience and data, and bring many benefits to human beings as a result. But the medicine diagnostic system is an extremely complex system. It is very difficult to use the formalized expression. In other words, the using of the general logical inference method hard to attain satisfied effect,therefore it is also difficult to realize the digitization and the automation.Tongue Diagnosis of TCM also is so. Medical diagnosis can be regarded as a non-linear mapping process, so yet can be carried out through the Neural Network. Because of the characteristics, such as classification, studying, memory, associating, calculating, fault-tolerant, real-time quality of the neural network, etc, it is comparatively suitable to adopting the Neural Network and using it as diagnose systematic model.This paper take the Tongue Diagnosis as a background, through merging fuzzy logic inference, neural network control technology, we present an intelligent inference model based on competitive neural network, which has succeeded in imitating the TCM specialist’s diagnosis inference. In this model, the use of artificial neural network technology to realize fuzzy inference and realizes the diagnosis with the competition nerve network. This fuzzy control algorithm integrates advantages of both fuzzy logic control and artificial neural network, and the application of competitive neural network enabled the system to have the on-line study ability.The author has mainly completed following two tasks in the paper:1、The application of self-organizing competitive neural network. Self-organizing competitive neural network is an unsupervised learning process. According to beforehand rules and input model, the network modifies the weights continuously until the whole system is order. Self-organizing competitive neural network can make the weights separate mutually through the adjustment of the input weight, and the weights can represent different models in different space, so the input can be clustered by the self-organizing competitive neutral net. In the research process of tongue inspection of TCM, we conductedthe fundamental research about the inference and discrimination of data of Tongue Diagnosis. At the same time,we have produced a concrete simulation experiment. Through the experiment, we can find that the use of competitive neural network can arrive the better classified recognition in the Tongue Diagnosis application.2、The fusion application of Fuzzy logic and neural network technology. Fuzzy logic is suitable for dealingwith unascertained, online and ill-posed problem, but it is lack of the ability of self-learning and self-imaging. Since neural network is applied to fuzzy Inference system, the conventional fuzzy inference system has not only the ability of logic reason and language expression but also the ability of self-learning and self-imaging. Through the fusion technology of neural network and the fuzzy logic, we have studied one new method about TCM’s diagnose and inference, which has succeeded in imitating the TCM specialist’s diagnosis inference. Simulation and elementary experiment show that the result is testified to be feasible and effective and a kind of new approach for medicine diagnostic process is provided.更多还原