The Design And Implementation Of Medical Process Analytical Platform

With the rapid development of biomedical informatics in China, clinical information systems are more and more widely and practically applied in medical institutes. These systems record information about various medical processes and the information can be used to generate clinical event log, which can be utilized to discover knowledge and patterns of medical processes for the improvement of clinical diagnosis and treatment. Medical quality is the core of hospital management and lifeline to a hospital. To improve medical quality through medical process analysis can create huge economic and social benefit.Although current medical process analysis apply data sources from different medical information systems or medical domain and develop specific analytical algorithms separately, they usually have certain procedures in common, such as data preprocessing, data analysis, analytical result interpretation and derivation, etc. While developing analytical methods during medical process analysis, the researchers usually have to repeat the implementation of certain basic algorithms to construct the integrated method. In the aforementioned separate mode of development of medical process analysis methods, it is too difficult for the researchers to reuse data service, data analysis methods or compare algorithms in specific step of a method. For these reasons, this thesis proposes a medical process analytical platform to support the medical process analysis researchers to apply heterogeneous data source, to reuse basic algorithms and to quickly start up the development of novel analytical method.The design of the medical process analytical platform mainly focuses on the satisfaction of the following requirements:To address the problem of transforming heterogeneous data sources into a unified form that is applicable in medical process analysis, this research designs a ontology-based data transformation to provide data service for the researchers. Medical process case base, unified data access interface, case data format are developed to offer basic support for the researchers to utilize data service.Medical process cases can be extremely complex and contain hundreds and thousands of steps that the researchers usually lost in static event logs with such a huge volume of steps. To develop an automatic display method can provide an intuitional and convenient way to observe medical process cases for the researchers.Therefore, this research designed the visualization service module in the platform to provide the dynamic display of medical process cases. Visualization service also allows the researchers to select temporal intervals of interest to probe into the case to analyze it dynamically.In order to support algorithm reuse and rapidly construct novel analytical method in a modulized way, this research also designed and implemented a plugin-based framework as the main structure of the analytical platform. The plugin based framework provides the dynamic load, unload, management and collaborative application of plugins which support the researchers to rapidly construct novel analytical method through invoke and connect existing plugins. This framework also makes it much easier to compare different algorithms that have aims in common in the same circumstance.Through the design of data transform method that compatible to heterogeneous data sources, the visualization of medical process cases based on dynamic display, the design of plugin based extension framework, the research work in this thesis provides the researchers with an integrated platform to support high reusability, rapid construction and convenient comparison. The platform improves the efficiency of the development of medical process analytical method and therefore also promotes the prosperity of medical process analysis research.