HealthRecordCommunity
FHIR🌏 Germany🏛 Fraunhofer IESE (in collaboration with BIH Charité)Enriched

AI-Powered Project to Automatically Convert Unstructured Medical Records into FHIR Resources

FHIR-Starter - Berlin Institute of Health in der Charité (BIH)

August 24, 2026Fraunhofer IESE (in collaboration with BIH Charité)

Summary

A German research project has launched, utilizing Large Language Models (LLMs) and natural language processing to automatically structure and convert unstructured medical records, such as doctor's letters, into standardized FHIR resources. This aims to reduce manual data entry and enable secure, reliable use of anonymized patient data for research.

Details

This project, named 'FHIR-Starter,' was launched in February 2025 by Fraunhofer IESE, in collaboration with the Berlin Institute of Health @ Charité and Insiders Technologies. The service uses AI to analyze unstructured medical records—such as doctor's letters—and automatically structure them into standardized FHIR resources. Its primary goals are improving healthcare by reducing manual data transcription, supporting clinicians with automatically generated structured information (e.g., lab curves, medication lists), and making anonymized patient data securely available for research. The technical design emphasizes minimizing AI hallucinations using mechanisms like the Uncertainty Wrapper. Furthermore, it strictly complies with GDPR by running exclusively on the users’ own servers using open-source LLMs. Funded by the German Federal Ministry for Economic Affairs and Energy, this initiative supports digitization in the health sector and enables advanced secondary use of medical data for research and innovation.

📰
Read Original Article
bihealth.org
→

Original content copyright by respective publishers