Text

Introduction: The “Open Medical Inference” (OMI) project aims to enable nationwide discovery and use of AI services across 16 German medical institutions, leveraging “Fast Healthcare Interoperability Resources” (FHIR), “Representational State Transfer” (REST) and “Digital Imaging and Communications in Medicine” (DICOMweb) protocols to enable healthcare data exchange and remote AI inference. The “Radiological Cooperative Network” (RACOON) serves as a multicentric research infrastructure connecting the radiology departments of all German university hospitals through standardized nodes, enabling collaborative radiological studies. Integrating OMI's AI service discovery capabilities with RACOON's infrastructure promises to enhance collaborative research accessibility. University Medicine Frankfurt (UMF) leads the development of a synchronization module to connect both systems and proposes a prototypical technical concept that ensures efficient connection while maintaining compliance with data safety regulations and requirements from both OMI and RACOON.

Methods: The integration concept leverages HL7 FHIR for high interoperability data exchange and DICOMweb for web-based medical imaging through RESTful services. The proposed synchronization module employs a research PACS for storing image data alongside “Firemetrics”, a FHIR-based interoperability solution from University Hospital Essen (UME), providing high-performance FHIR server capabilities with full relational SQL support.

Results:The proposed architecture developed with stakeholders from OMI, RACOON, UMF and UME integrates heterogeneous data sources through standardized transformation pipelines. Clinical HL7v2 data from ORBIS database mirrors undergo CSV-to-FHIR ETL processing, while radiological DICOM images from RIS/PACS systems are subjected to defacing, debranding and skull-stripping before being passed to DICOM-to-FHIR pipelines using the OMI-developed DICOMweb Adapter. All resources are harmonized and pseudonymized through a centralized FHIR Gateway before being stored in local FHIR servers. The OMI Client orchestrates controlled data transfer to remote AI inference servers registered in the OMI Service Registry. Key is the synchronization module (Test NODE), featuring a Research PACS for image storage alongside a full Firemetrics instance. The architecture enables bidirectional FHIR/DICOMweb communication between OMI Gateway and RACOON nodes, enabling researchers to conduct multimodal research.

Discussion: Our proposed technical concept addresses a need for a connecting component that allows OMI and RACOON to access each other´s infrastructure. Early stakeholder engagement ensured clinical-technical alignment. Current limitations include pending real-world implementation and performance testing under clinical loads. Active development of OMI components necessitates continuous collaboration between synchronization module developers and component teams. This collaboration ensures technical compatibility, unified functionality, and scalable architecture enabling seamless integration across both ecosystems for all partaking institutions.

Conclusion: This collaborative framework successfully bridges OMI's AI service discovery capabilities with RACOON's multicentric radiological infrastructure, establishing a scalable blueprint for nationwide integration. By leveraging standardized protocols and consortium-developed solutions, the architecture enables seamless data exchange while maintaining compliance to both OMI and RACOON requirements. Following the prototype validation at the UMF, systematic deployment across all participating institutions is planned.

Acknowledgements: OMI is funded by the German Federal Ministry of Education and Research (BMBF) under the funding reference number 01ZZ2315H. RACOON is funded by NUM 2.0 under the funding reference number 01KX2121.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.

Literatur

[1] Sigle S, Werner P, Schweizer S, Caldeira L, Hosch R, Dyrba M, et al. Bridging the Gap Between (AI-) Services and Their Application in Research and Clinical Settings Through Interoperability: the OMI-Protocol. 2024 Feb. DOI: 10.34657/13458
[2] Salg GA, Ganten MK, Bucher AM, Kenngott HG, Fink MA, Seibold C, et al. A reporting and analysis framework for structured evaluation of COVID-19 clinical and imaging data. npj Digit Med. 2021 Apr 12;4(1):1-9. DOI: 10.1038/s41746-021-00439-y
[3] HL7 International. HL7 FHIR [Internet]. 2025 [cited 2025 Apr 1]. Available from: https://www.hl7.org/fhir/
[4] Medical Imaging Technology Association (MITA). DICOM. DICOMweb. 2025 [cited 2025 Apr 14]. Available from: https://www.dicomstandard.org/using/dicomweb
[5] Firemetrics GmbH. Firemetrics - Interoperable Medical Intelligence & AI [Internet]. 2025 [cited 2025 Apr 16]. Available from: https://firemetrics.ai/index.html

Erratum

The title of the abstract has been corrected.