A Framework for Extracting, and Validating Named-Entities to Integrate Openehr Using the Example of Free Text...

Area of Science:

• Medical Informatics
• Clinical Data Management
• Natural Language Processing

Background:

• Extracting information from unstructured physician notes in hospitals is a significant challenge.
• Existing solutions lack efficiency and reliability, often requiring human intervention.
• Factors like limited training data and sensitive patient information complicate automated processing.

Purpose of the Study:

• To develop a complete process for efficient data extraction, validation, and integration from unstructured clinical texts into openEHR.
• To address the open problem of reliable information extraction from physician-written hospital documents.
• To demonstrate the utility of the developed tools using molecular genetic findings.

Main Methods:

• Development of a comprehensive toolset for unstructured text processing.
• Implementation of a pipeline for data extraction, validation, and integration into openEHR.
• Application of the pipeline to free-text molecular genetic findings for use-case demonstration.

Main Results:

• The developed pipeline successfully extracts, validates, and integrates data from unstructured clinical texts.
• A specific use case involving molecular genetic findings demonstrated the system's effectiveness.
• The validation of the pipeline achieved a high F-score of 0.98.

Conclusions:

• The developed tools provide an efficient and reliable solution for integrating unstructured clinical data into openEHR.
• The pipeline effectively overcomes challenges associated with processing complex medical texts.
• The high F-score validates the robustness and accuracy of the proposed data integration method.