MeKG: Building a Medical Knowledge Graph by Data Mining from MEDLINE

Abstract

Mining data on a knowledge level can help to achieve a higher performance of a decision support system. This study built a knowledge graph based on MEDLINE that has a large number of articles in the medical domain. MEDLINE uses Medical Subject Headings (MeSH) for document index. Based on MeSH, articles are extracted from the MEDLINE correspondent to medical subjects. Using the MeSH as the backbone of knowledge base, the MEDLINE articles were used to generate instances which helped to populate the knowledge base. This approach facilitated the creation of a knowledge graph that was capable of discovering the hidden knowledge among concepts of MeSH. The knowledge graph had a significant effect on improving the quality of healthcare. The contribution of the research is on a framework for building knowledge bases. Moreover, the approach provided an essential source at the knowledge level for researchers in healthcare.