Development of Intellectual Medical Information System Based on Decision Trees and Expert Knowledge

Method for Architecture of medical information system construction is described. The proposed architecture sup-ports two approaches for knowledge gathering and use. The first one is Data Mining for data about actions that are already performed and decisions that are already taken, and also for other online data from database. We use one of wide-spread classification methods - decision trees. The new algorithm of iterative decision tree building is introduced. It allows working with classes represented by finite sets (of particular decisions).

The second approach presented in the architecture is the expert knowledge processing. Experts has deeper understanding of processes in particular domain then the knowledge, which can be deduced from database data. Expert knowledge is necessary for non-standard circumstances and thus, hybrid system is more practical then one based on pure Data Mining or on pure expert knowledge.

Business-processes, expert knowledge and user actions in the medical information system are grouped into do-mains. Domains are highly tailored, and every user application that works inside some domain is supplied with in-tellectual component. This component supports a user in domain-based decision making in several levels: from hints about filling simple form data to regular users to decision support for medical specialists. Another task of domain-based components is to export generalized data about business processes to high-level intellectual component that supports complex management decision making.