摘要
针对流行病学研究的特点,我们提出计算机辅助医学数据挖掘系统构架,以糖尿病并发症为研究实例,探讨医学数据的冗余性消除、规范化储存、知识归纳及可视化表达等问题。以天津总医院3 022例普查数据为研究对象,尝试解决用计算机实现糖尿病并发症这类定性数据的定量化数据挖掘和知识发现。通过对于43种并发症的定性数据挖掘,可以发现诸如高血脂、冠心病、高血压和脑血管病等具有明显并发倾向的知识规则18条。同时,采用知识树方式和决策树等方法实现知识规则的可视化表达。基于数据挖掘和知识发现计算机辅助医学数据挖掘系统能够对现有病历数据库中数据进行自动分析并且提供有价值医学知识,特别适合流行病学分析和全民健康评估,因此与社区医疗和医院HIS系统结合是未来一个非常现实的发展方向。
In this paper, a systematic architecture of medical data mining based on computer was provided for epidemiological analysis. Complications in diabetes mellitus were used as the cases under discussions on redundancy elimination, normalized storage, knowledge induction and visual expression of medical data. 3022 pieces of census records from Tianjin General Hospital were researched to find the solution of quantitative mining from qualitative data and knowledge discovery. From the qualitative data mining of 43 kinds of complications in diabetes mellitus, we found 18 knowledge rules with significant statistical meaning on concurrency relation, e.g. hyperlipoidemia, coronary disease, hypertension and cerebrovascular disease. And knowledge tree was noted to be an effective visual expression method for showing the rules generated from the above system. Medical analysis system based on data mining and knowledge discovery could generate effective knowledge rules from medical record database,which was found to be especially useful for epidemiological analysis and national health survey. So how to cooperate with community medical care and hospital information system in the near future is practically significant.
出处
《生物医学工程学杂志》
EI
CAS
CSCD
北大核心
2008年第2期295-299,共5页
Journal of Biomedical Engineering
作者简介
通讯作者。E-mail：sunset@eyou．com
