The construction of a TCM knowledge graph and application of potential knowledge discovery in diabetic kidney disease by integrating diagnosis and treatment guidelines and real-world clinical data.

Abstract

Background: The complexity and rapid progression of lesions in diabetic kidney disease pose significant challenges for clinical diagnosis and treatment. The advantages of Traditional Chinese Medicine (TCM) in diagnosing and treating this condition have gradually become evident. However, due to the disease's complexity and the individualized approach to diagnosis and treatment in Traditional Chinese Medicine, Traditional Chinese Medicine guidelines have limitations in guiding the treatment of diabetic kidney disease. Most medical knowledge is currently stored in the process of recording medical records, which hinders the understanding of diseases and the acquisition of diagnostic and treatment knowledge among young doctors. Consequently, there is a lack of sufficient clinical knowledge to support the diagnosis and treatment of diabetic kidney disease in Traditional Chinese Medicine. Objective: To build a comprehensive knowledge graph for the diagnosis and treatment of diabetic kidney disease in Traditional Chinese Medicine, utilizing clinical guidelines, consensus, and real-world clinical data. On this basis, the knowledge of Traditional Chinese Medicine diagnosis and treatment of diabetic kidney disease was systematically combed and mined. Methods: Normative guideline data and actual medical records were used to construct a knowledge graph of Traditional Chinese Medicine diagnosis and treatment for diabetic kidney disease and the results obtained by data mining techniques enrich the relational attributes. Neo4j graph database was used for knowledge storage, visual knowledge display, and semantic query. Utilizing multi-dimensional relations with hierarchical weights as the core, a reverse retrieval verification process is conducted to address the critical problems of diagnosis and treatment put forward by experts. Results: 903 nodes and 1670 relationships were constructed under nine concepts and 20 relationships. Preliminarily a knowledge graph for Traditional Chinese Medicine diagnosis and treatment of diabetic kidney disease was constructed. Based on the multi-dimensional relationships, the diagnosis and treatment questions proposed by experts were validated through multi-hop queries of the graphs. The results were confirmed by experts and showed good outcomes. Conclusion: This study systematically combed the Traditional Chinese Medicine diagnosis and treatment knowledge of diabetic kidney disease by constructing the knowledge graph. Furthermore, it effectively solved the problem of "knowledge island". Through visual display and semantic retrieval, the discovery and sharing of diagnosis and treatment knowledge of diabetic kidney disease were realized.