MLEE: A method for extracting object-level medical knowledge graph entities from Chinese clinical records

Area of Science:

• Medical Informatics
• Natural Language Processing
• Knowledge Representation

Background:

• Medical knowledge graphs are crucial for causal inference in healthcare.
• Electronic clinical records offer vast data for machine learning-based medical knowledge discovery.
• Accurate entity extraction aligned with clinical text logic is vital for high-quality medical knowledge graphs.

Purpose of the Study:

• To develop and validate a method for extracting medical entities from real Chinese clinical electronic records.
• To address challenges in building robust medical knowledge graphs from unstructured clinical data.

Main Methods:

• Defined a computational architecture named MLEE (Machine Learning Entity Extraction).
• MLEE is designed to extract object-level entities with "object-attribute" dependencies.
• Utilized real Chinese clinical electronic records from 1,000 patients for experimentation.

Main Results:

• The proposed MLEE method demonstrated effectiveness in extracting medical entities.
• The approach showed promise in enhancing the consistency of entity extraction with clinical text logic.
• Experimental validation was conducted using data from Shengjing Hospital of China Medical University.

Conclusions:

• The MLEE architecture provides a viable solution for extracting structured medical entities from electronic clinical records.
• This method contributes to building higher-quality medical knowledge graphs.
• Further research can leverage this approach for advanced clinical data analysis and knowledge discovery.