Fraunhofer Project to Automate Conversion of Full Text into Structured Data Using LLMs
Fraunhofer-Projekt arbeitet an automatisierter Umwandlung von Volltext in strukturierte Daten
Summary
The Fraunhofer Institute (IESE) is developing a project that uses Large Language Models (LLMs) and Natural Language Processing (NLP) to automatically convert unstructured full-text medical documents, such as doctor's notes, into standardized data formats. This aims to solve issues of manual data entry and difficulty in research utilization within the German healthcare system.
Details
A three-year research project called "FHIR-Starter" is underway at the Fraunhofer Institute for Experimental Software Engineering (IESE). The project addresses the challenge of handling unstructured information—such as doctor's notes provided as full-text PDFs—in Germany’s currently lagging digital healthcare system. The goal is to develop a software service that analyzes these full-text documents using LLMs and NLP, converting them into standardized data formats. Key standards utilized include FHIR (Fast Healthcare Interoperability Resources), LOINC, and SNOMED-CT. The consortium, which includes the Berlin Institute for Health Research at Charité, is developing open interfaces to allow all stakeholders to automatically ingest the structured data into their respective systems. The team places a strong emphasis on ensuring data reliability and comprehensive data protection. Furthermore, the service is expected to significantly elevate the use of the electronic patient record (ePA). In the long term, it could enable doctors to view laboratory values over time or automatically generate medication lists using structured data, thereby achieving a fully functional and meaningful digitalization of the ePA.
Original content copyright by respective publishers