Research On Medical Case-Based Reasoning Based On Ontology

In the medical filed, it is a hotspot to study the case-based reasoning technology, which makes up for the shortage of rule-based reasoning technology that has the difficulty of acquiring rules and weak ability to explain anomalies. However, each hospital establishes its own CBR system based on own practical applications and case representation, leading to inability to exchange, integrate and share the cases among CBR systems.This paper combines the ontology technology with case-based reasoning technology based on the application requirements and characteristics of disease diagnosis and treatment, achieving the ability of sharing case. The followings are details:1) This paper proposes the ontology building method which is a continuous abstract and depth progress from user model to implementation model, considering the Gruber's five principles of ontology building. Using the multi-layer architecture, we separate the ontology-based medical case-based reasoning framework into the application layer, function layer and the data source layer, with each layer having its own capability.2) According to the five requirements and the philosophy of case design, we propose an ontology-based medical case representation method. Case is divided into basic information, problem description and treatment, which takes full advantage of the ontology technology and supports the data sharing of the heterogeneous case database,organizing, maintaining and extending the case easily.3) Taking into account the characteristics of hybrid ontology structure and the need for practical applications, we propose an integrated ontology mapping method-IMNAS (the Integrated Method of Name, Attribute and Structure) which uses the name of concept, attribute and structure of ontology as the accounting objects. In the implementation of structure-based mapping, the way of traversing the ontology structure is from bottom to top, by setting the threshold to lower the algorithm's time complexity.4) According to CBR practical applications and data types'features, this paper proposes calculation methods of similarity on Boolean, Symbol and Set data type, and an asymmetric similarity calculation method adapted to the data types Integer and Float. In the fifth part, we use the diabetic hypertension cases as experimental subject to verify the effectiveness of the proposed method and the ontology-based case representation in practical application.