Study On The Clinical Decision Support System Based On The Method Of CBR-RBR Integrated Reasoning

Along with the development of medicine and health, progress of science and technology, the medical science and technology and medical infomationization have been continuously improved. As a very active branch of medical knowledge engineering and artificial intelligence research, clinical decision support system (CDSS) is always the focus of research and application at home and abroad.The core of clinical diagnosis and treatment is the processes of disease diagnosis, disease control through medical intervention, and finally cure the disease. CDSS simulates the processes of clinical diagnosis and treatment based on the principle and method of artificial intelligence. The processes include the patient clinical data collection, the clinical medical knowledge acquisition, the patient information and the medical knowledge matching, the matching result interpretation, providing suggestion of diagnosis and treatment and so on. CDSS can improve the service quality of medical and health institutions, through reducing the rate of misdiagnosis and standardizing diagnosis and treatment behavior.But, what can be accepted by doctor and put in clinical use is few in the current CDSS. The main reason is that the current CDSS relies too much on clinical medicine knowledge inference rule, and pay too little attention to diversity, variability and uncertainty factors of disease. Such systems cannot help doctors in the case of perplexing patient and disease. Some problems of rule-based reasoning (RBR) system can be solved to a certain extent by the method of case-based reasoning (CBR). But the CBR research is just in method state. There are some shortcomings when CBR be used only, such as:it is difficult to express the deep domain knowledge; case indexing and matching criteria are difficult to unify; case retrieval and similarity matching algorithm has yet to be further improved. Fortunately, CBR is a method rather than a technology, so it can promote own development, through combining with new technologies and methods of other academic subjects. CBR and RBR can play their respective advantages by integrating, expanding the application scope of CDSS, and improving the application effect.Based on the basic idea of the control theory, this thesis explores a new thought of CDSS development in order to solve the problems of CDSS research and application, through the combination and integration of the theory, method and technology in artificial intelligence and computer technology. The main research contents are listed as follows:1. In order to establish unified standards of CDSS, based on the idea of ontology and analysis of clinical medical features, this thesis proposes the construction method of clinical medical ontology. The detailed methods of rule-based and case-based knowledge representation are described, which includes the standard of clinical medical knowledge visualization, the method of clinical medical knowledge SAGE modeling, the representation of clinical case knowledge, and the structure of clinical case knowledge base.2. Knowledge acquisition is a key problem of CDSS. The communication between clinicians and knowledge engineer is very important. This thesis analyses the way of clinical medical knowledge acquisition, introduces the detailed procedures of rule knowledge acquisition, which includes knowledge representation, SAGE modeling and the rules generation through the examples. Text information extraction is the first step of knowledge acquisition. It is very important to case knowledge acquisition. This thesis expounds the recognition method of medical named entity based on both finite state automata and conditional random field (CRF), analyses the results of experimentation, and describes clinical case knowledge acquisition method and procedures.3. The method of CBR-RBR integrated reasoning is the focal point of this thesis. On the basis of analyzing logical expressions of rule-based forward reasoning, backward reasoning and forward-backward inference, and the four-phase CBR cycle, this thesis proposes the CBR-RBR integrated reasoning model and main procedures of the construction method of rule sets based on SAGE model. Similarity calculation is the most critical problem of case-based reasoning, especially for the clinical case with flexible structure. This thesis designs the similarity calculation methods in connection with the distinguishing feature of clinical case and the variety of data types, such as text type, degree type, value type, date time type and so on, through analyzing the existing method on similarity calculation. In order to accomplish automatic calculation of similarity weight, the calculation method of item weights based on the idea of TF-IDF, and the comprehensive weights based on self-organizing competitive neural network, are put forward, which are tested and verified through experimentation. 4. In order to optimize the structure of clinical case knowledge base, and for the sake of realizing transformation from case knowledge to rule knowledge, this thesis proposes hierarchical clustering analysis method and condensation degree analysis method based on case similarity matrix norm. The significance and actual values of these analysis methods are proved by data analysis of diabetes mellitus type2and hypertension cases.5. In hand of CDSS practice, this thesis describes the system framework and main functions of CDSS based on the above theory and method. The clinical experiment of CDSS is completed in a general hospital, and the experimental effect is analyzed and estimated.Disease control is not only the means of treatment for diseases, but also is sometimes a therapeutic target. As a kind of support platform, the CDSS in this thesis can help clinicians in the whole process of disease diagnosis, disease control and finally cure the disease. It is also an important tool for study of intervention control theories of some senile diseases. Research of the theory and method of CDSS is based on clinical application as guide. In the research process, this thesis emphasizes on clinical medical treatment process characteristics, makes full use of advantages of case-based reasoning and rule-based reasoning, and blends process control thought of disease. It accelerates the pace of construction of clinical medical knowledge base, through enhancing flexibility and adaptability of the clinical case knowledge representation method, lets more clinicians willing to use CDSS, and can help them improve ability of clinical diagnosis and treatment. It is crucial on research and development of CDSS that combining theory and clinical practice of clinical decision support system. The research results of this thesis have been verified by experiment of clinical data, and have important theory and application values. The CBR-RBR integrated clinical decision support system model which is described in this thesis has an important role in promoting development of CDSS. It has a certain role in promoting disease intervention control theory and application which blending modern control theory and modern clinical medicine theory.