A Study Of Linked Data Based Clinical Decision Support System

Clinical decision making is a complex process. Doctors make decisions according to the characters of different patients with medical knowledge and clinical experience in the process of clinical decision making. However, lacking of sufficient medical information causes improper decision, which sometimes makes medical malpractice. Meanwhile, the asymmetry of medical information sometimes causes misunderstanding of patient even medical dispute. In order to reduce human factor caused mistakes in clinical decision making and enhance the patient’s comprehension to medical advices, hospital wants to build a clinical decision support system which can retrieve similar history cases and present relevant medical knowledge.Aiming to the requirement of hospital, we proposed a linked data based clinical decision support system.This system uses linked data technology to reorganize data in HIS, and integrates external medical knowledge bases. It can select semantic similar cases from history records, prompt or warn potential risks, extract and present relevant medical information. And it will support the clinical decision making process for doctors, help patients understand their condition and improve the motivation and cooperation of patients. The research includes:1. We proposed a framework of linked data based clinical decision support system.This framework comprises 4 layers, namely data layer, integration layer, application layer and UI layer. Data layer comprises HIS database and linked medical knowledge base. Integration layer integrates these two sources of data to form a linked data net. Application layer is responsible for data query, analyze and manage on linked data net. UI layer presents the results of application layer to the users.2. We designed a clinical linked data model.Since a treatment is formed by a series of medical advices, this model uses medical advices of patients as its core. We use mapping file and regular expression to convert relational data into this model.3. We proposed an algorithm to integrate external linked medical knowledge base.Drugs and diseases were used as the entities in the integration process. For disease entities, we use ICD-10 code for integration. For drug entities, we used our integration algorithm based on the similarity of drug names and chemical formula.4. We proposed a similar case selection algorithm.We firstly propose a similarity algorithm based on classification tree. This algorithm uses classification information in linked medical knowledge base and calculates the semantic similarity between concepts. Secondly, we use this algorithm to calculate the concept similarity of drugs and diseases, and analyze the advantage of it. Finally, we proposed a case similartity based on patient’s drug history and disease history, and analyze the results.5. We implemented a prototype of our framework to validate the feasibility.We chose colorectal cancer treatment as application scenario, and implemented a prototype of our decision support system. We compared our method with existing methods in the final.It is proofed that the linked data based clinical decision support system framework we proposed has certain feasibility and usability.