This paper conducts in-depth research and practice on the construction of medi- cal data knowledge graphs, exploring their core value in the medical field. With the explosive growth of medical data from electronic health records, genomic sequenc- ing, and medical literature, data silos and heterogeneity have restricted medical efficiency and intelligent development. Medical data knowledge graphs address this by integrating multi-source heterogeneous data into a structured semantic network, enabling effective organization and application of medical knowledge. The study details the construction process, including data acquisition from clini- cal databases and literature, preprocessing (cleaning, annotation), core technologies (ontology design, entity recognition, relation extraction, knowledge fusion), and graph database-based storage and query methods. It compares technical routes (rule-based, machine learning, deep learning) and highlights innovations like com- bining BERT and graph neural networks to enhance entity extraction and relation prediction accuracy. Practical applications in clinical decision support (e.g., accelerating differen- tial diagnosis via disease-symptom-drug relationships), rare disease diagnosis, and public health management (e.g., tracking infectious disease spread) are explored through case studies. The paper also summarizes challenges such as data privacy, dynamic knowledge updates, and terminology standardization, proposing solutions and future directions. This research provides theoretical and practical support for advancing medical informatization and intelligence, contributing to improved service quality, reduced costs, and precision medicine.
Han et al. (Sun,) studied this question.