System architecture for temporal information extraction, representation and reasoning in clinical narrative reports

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Electronic Medical Records (EMRs) contain crucial temporal information vital for healthcare.
• Extracting and interpreting temporal data from clinical narratives remains a significant challenge.
• Accurate temporal understanding is key for clinical decision support and research.

Purpose of the Study:

• To propose an integrated architecture for processing temporal information in clinical narrative reports.
• To establish a foundation for applications that determine the time of clinical events.
• To enhance healthcare practice and research through improved temporal data analysis.

Main Methods:

• Development of an annotation schema and tagger for temporal expressions.
• Implementation of a Natural Language Processing (NLP) system for event extraction and temporal data association.
• Creation of a post-processor with a knowledge-based subsystem to resolve temporal expressions, addressing granularity and vagueness.
• Utilizing a reasoning mechanism that models clinical reports as Simple Temporal Problems (STPs).

Main Results:

• The proposed architecture effectively processes temporal information in clinical narratives.
• The system can encode, extract, and associate medical events with formalized temporal data.
• The post-processor successfully resolves temporal expressions, handling complexities like vagueness.
• The reasoning mechanism provides a framework for temporal problem-solving in clinical reports.

Conclusions:

• The integrated architecture offers a robust solution for temporal information extraction from clinical narratives.
• This approach lays the groundwork for advanced healthcare applications leveraging precise temporal event data.
• Improved temporal analysis in EMRs can significantly benefit clinical practice and medical research.