Study On Medical Knowledge Acquirement And Discovery

Automation of medical knowledge acquirement and discovery is significant to medical research. It is essential to solve the conflict of "data rich but knowledge poor" which is important to improve the medical researchers' knowledge. In this paper, a systematic architecture of medical knowledge acquirement and discovery is provided including medical knowledge acquirement system based on extended knowledge editor and medical knowledge discovery system based on data mining technique. A solution based on this architecture is designed to diabetes syndrome and complication study, which includes all function modules of the medical knowledge discovery architecture and the medical knowledge acquirement architecture. "Knowledge editor" is used to summarize the medical knowledge to systematic model, which is used as the prototype of data mining. The task of this study is how to promote, fuse and complement the techniques of "data mining" and "knowledge editor" in medical research.In this subject study, we focus on the technique details of quantitative mining from qualitative data, cluster analysis of quantitative data, visualization and modeling expression of knowledge and interactive data mining under the user instruction. Besides establishment of the architecture of knowledge acquirement and discovery, we discuss the association rule model and rough information decision model which are the typical techniques of quantitative mining from qualitative data and artificial neural network and fuzzy C-eans cluster which are typical to quantitative data mining. The classic association rule model is modified in two ways below: the Apriori attribute of data set is used to solve the rules' redundancy which means that the redundant rules would be stepwise eliminated from the rule set with k-1 length while a new rule is added to the rule set with k length. Like the mutliscale analysis, an inferior support threshold index is added to balance the quality and quantity of rules output by the system, which could reduce the rules thrown away in the long knowledge rules mining process. Scheme based on the span tree in graph theory is designed to implement knowledge induction and visualization and Unified Modeling Language is used to express the knowledge to computer. Refinement and fusion model of text knowledge is also studied. An interactive data mining system based on knowledge editor is designed to analyze in the diabetes research. All of these data mining techniques are tested in the epidemiological analysis of diabetes syndrome and complication and the statistic distribution analysis of some physiological parameters relevant to diabetes mellitus.