Construction of a Digestive System Tumor Knowledge Graph Based on Chinese Electronic Medical Records: Development and...

Area of Science:

• Medical Informatics
• Artificial Intelligence
• Oncology

Background:

• Increasing digestive system tumor incidence and mortality in China necessitate improved data analysis.
• Chinese electronic medical records (CEMRs) contain valuable clinical data for understanding tumor diagnosis and treatment.
• Knowledge graphs offer a powerful framework for processing and organizing complex medical information.

Purpose of the Study:

• To construct a semantic-driven digestive system tumor knowledge graph (DSTKG).
• To represent knowledge from CEMRs with fine granularity and rich semantics.
• To address challenges in constructing Chinese tumor knowledge graphs, focusing on schema and semantic relationships.

Main Methods:

• Developed a DSTKG construction framework utilizing CEMRs.
• Created a knowledge graph schema with 7 classes and 16 semantic relationship types.
• Employed knowledge extraction, named entity linking, and graph visualization for DSTKG creation.
• Evaluated DSTKG quality across data, schema, and application layers.

Main Results:

• The DSTKG received a high expert rating (4.20 overall).
• Schema structure rationality, scalability, and result readability scored exceptionally well (above 4.67).
• DSTKG represents more granular entities and relationships than existing Chinese tumor knowledge graphs.
• The knowledge graph demonstrated flexibility and potential for personalized customization.

Conclusions:

• A granular, semantic DSTKG was successfully constructed.
• The DSTKG provides a foundation for building tumor knowledge graphs and for intelligent applications using CEMRs.
• Further research incorporating diverse data sources and advanced assertion classification is recommended to enhance DSTKG capabilities.