AI-Powered Medical Data Standardization: Launch of FHIR-Starter Project
KI-basierte Datenstrukturierung für Labor und Gesundheitswesen - LABO.de
Summary
The Fraunhofer Institute, leading the 'FHIR-Starter' research project, has begun efforts to automatically structure unstructured medical text using LLMs and NLP. This aims to convert non-structured data, such as PDF clinical reports, into standardized formats, thereby improving the efficiency of utilizing healthcare data.
Details
In Germany's healthcare sector, medical information—such as consultation records—is primarily exchanged via physician's notes in PDF format. This necessitates that doctors manually read these full texts and transfer relevant details into Practice Management Systems (PVS) or Hospital Information Systems (KIS), a process that is time-consuming and prone to errors. Furthermore, unstructured data presents a major challenge for medical research. The 'FHIR-Starter' project addresses this by providing a software service that analyzes full text documents using Large Language Models (LLMs) and Natural Language Processing (NLP). It extracts necessary information from these PDF-like texts and converts them into standardized formats. Specifically, it utilizes FHIR (Fast Healthcare Interoperability Resources), the international standard for medical data exchange, along with LOINC and SNOMED-CT coding systems. This initiative aims to facilitate data exchange between different software systems, making collected health data more effectively usable. The project is funded as part of the 'Generative KI für den Mittelstand' innovation competition by the German Federal Ministry for Economic Affairs and Energy. Key development challenges include ensuring data reliability and adhering to high data protection standards.
Original content copyright by respective publishers