Application and Research on Data Mining Technology in the Hospital Information System

Abstract

Most of the current medical database in hospital information system is poorly applied, the low level data integration and superficial analysis can not satisfy the automatic access to medical knowledge, not to mention the requirement of hospital management. This paper introduces the key technologies in medical data mining such as data pretreatment, anonymization, identity transformation, etc. It elaborates the basic process of data mining in hospital information system, including data extraction and pretreatment, mining algorithm execute, model discovering, knowledge expression and evaluation. It also concretely discusses the application of data mining technology in hospital information system in two major aspects: modal construction and implementation process. KeywordsHospital information system(HIS); data mining; model; process I. KEY TECHNOLOGIES IN HOSPITAL DATA MINING A. Data pretreatment Data pretreatment is an important step in the process of data mining (DM), especially when the database contains noisy, incomplete and in-consistent data. It usually takes about 60% time to treat the original data in the whole procedure of data mining, and only 10% the real time for key DM process. The main steps of data pretreatment include data cleaning, data integration, data transformation and data elimination. B. Data anonymity and identity transformation Data pretreatment is an important step in the process of data mining, especially when the database contains noisy, incomplete and in-consistent data. It takes about 60% time to treat the original data in the whole process data mining, and the time for data mining is only 10%. Data pretreatment includes mainly data cleaning, data integration, data transformation and data elimination. Because medical information involves the problem of patients’ privacy, the medical information needs special treatment, such as the anonymity of the patient and the identity transformation. Anonymity means to remove the real identity of the patient, or to replace the true identity with wrong identity. After the process of anonymity, researchers can not query any private information of patient through clinic records. Identity transformation differs lightly from anonymity. The identity after the transformation may still contains some real information of patient, and can be viewed only by the authorized researcher. C. Data mining on medical text data For the medical text information, it’s hard for people to interpret the data, such as image, signal and other clinic data with non-standardized method, and difficulty to implement the data mining with these non-standardized data. At present, medical text data standard transformation works well with computer technology. The main three steps are: analysis of source statement, transformation and target statement creation. The main difficulty of transformation is that the source statement is not unique, so it needs to collect all types of source statements. The current machine transformation can treat sentence of less than 10 words. XML (Extensible Markup language), a structured language can be another way of text data standardization. XML can not only create structured text data, but also a good tool to treat and share data. XML is a key technology for data mining and knowledge finding. D. Data mining on image data The current medical images come from imagery machines, such Computed Tomography (CT) and type B ultrasonic, which have been proved a reliable assistant diagnose mean for the doctor. Different from data mining using pure digital data, it’s more difficult to accomplish the data mining through medical images, so the development of effective image data mining tool becomes a key technology. Data mining through medical image includes the following aspects:  Removal or reduction the influence of the image noise to enhance the target image quality and to fetch the border of the target organization.  Description and characterization of the target organization on concept level, to acquire and validate the dynamic scope of some correspondent parameters;  Management and index of the medical image data. Nowadays, breakthrough of data mining to SPECT image has been achieved. Furthermore, research on rapid and high quality mining algorithm, to guarantee the accuracy and reliability of the knowledge through data mining are also the key factors of medical data mining. Proceedings of the 2nd International Conference on Computer Science and Electronics Engineering (ICCSEE 2013) Published by Atlantis Press, Paris, France.