[A customized method for information extraction from unstructured text data in the electronic medical records]

Area of Science:

• Medical Informatics
• Natural Language Processing
• Data Mining

Background:

• Electronic medical records (EMRs) contain valuable clinical information but are often unstructured.
• Extracting structured data from Chinese narrative text in EMRs is challenging for medical research.

Purpose of the Study:

• To develop and validate a rule-based information extraction method for unstructured Chinese EMR data.
• To enable practical utilization of clinical data for research and hospital surveillance.

Main Methods:

• A rule-based information extraction method was developed using annotated Chinese EMR corpora.
• Extraction rules were created using Perl regular expressions.
• Data extraction and organization into tabular format were performed.
• Validation involved artificial and automated verification on a clinical data integration platform.

Main Results:

• The method achieved high recall (87.6-89.2%) and accuracy (99.5-99.2%) in extracting diabetes history.
• F-Scores of 0.93 and 0.94 indicate robust performance.
• The system effectively organized extracted data for clinical research.

Conclusions:

• A novel algorithm combining NLP and rule-based extraction was designed for Chinese EMRs.
• The proposed method demonstrates superior performance compared to existing approaches.