DEKGB: An Extensible Framework for Health Knowledge Graph

Abstract

With the progress of medical informatization and the substantial growth of clinical data, knowledge graph is playing an increasingly important role in medical domain. Medical domain is highly specialized with abundant high-quality ontologies, and has many professional sub-fields such as cardiovascular diseases, diabetes mellitus and so on. It is very difficult to build a health knowledge graph for all of the diseases because of data availability and deep involvement of doctors. In this paper, we propose an efficient and extensible framework, DEKGB, to construct knowledge graphs for specific diseases based on prior medical knowledge and EMRs with doctor-involved. After that, we present the detailed process how DEKGB is applied to extend an existing health knowledge graph to include a new disease. It is confirmed that using this framework, doctors can get highly specialized health knowledge graphs conveniently and efficiently.