Design and Implementation of An Intelligent Clinical Laboratory Data Mining System

• Main Authors: LIN, NIEN-MAO, 林年茂
• Other Authors: 游坤明
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/18420213922747162482

碩士 === 中華大學 === 資訊工程學系(所) === 94 === AbstractAs medical personnel use the Hospital Information System to keep track of treatments patients received, the data was stored in the database of the hospital information system. The immense amount of these data is not only records of the entire medical treatment process including patients’ behaviors when they enter the hospital, their responses to therapies, the handling of doctors and nurses, and the final results; it is also the precious accumulation of experiences of many medical personnel and the fundamental information for medical researches. The database continues to build up over time. How to turn these implicit data into norms of explicit knowledge to upgrade medical care standards is therefore a very important issue. However, these medical data are still not easy to handle and study and are therefore difficult to use. To make these implicit data become explicit knowledge, one must rely on mature data mining technology and professional judgment or clinical evidence before these data can be gathered into knowledge database. Such knowledge can then be reference material to aid medical personnel in their clinical judgment and thus become the norms of explicit knowledge to upgrade medical care standards. At the same time, an ideal data mining system has to be able to manage data automatically and perform data mining rapidly and efficiently. It must also allow the user to adjust mining parameters at will and endure repeated use over a long period of time. In the end, the results are integrated into a usable, manageable and expandable database to meet the user’s needs. This paper is an answer to the abovementioned requirements for building an intelligent and adaptive data mining system. To verify the performance of our proposed structure, we have implemented an intelligent adaptive data mining system according to this thesis. In the experiments, our proposed system has excellent performance in valid support ratio, mining pattern results and execution efficiency. In valid support ratio, 9 categories were chosen to examine the mining results of 293 biochemical tests. When the support rate was 0.02, the valid support ratio stood at 0.686. Even when the support rate was 0.06, the valid support ratio still reached 0.463. This verifies the system can indeed effectively mine out useful data. In clinical testing, the mining results also conformed to the correlation between clinical diseases and test categories. In execution efficiency, when the support reduced, despite the types and the number of conditional samples increased, the mining only took a slightly more time. This shows the system is reasonably efficient. Consequently, so long as this system is continued to perform mining in other categories, there is no question that a more complete knowledge database can be established effectively. This also proves the knowledge extraction scheme of the intelligent adaptive mining system is feasible. Key words: Laboratory Information System, data mining, adaptive, knowledge database