The EU AI Act in Hospitals: How to Integrate AI into Your Radiology Department

Regulatory requirements for the use of artificial intelligence require hospitals to strike a balance between innovation and safety. With JiveX, you can avoid data silos, maintain control, and store data in compliance with law.

The provisions of the EU AI Act, which are coming into effect in phases, are fundamentally changing the use of artificial intelligence (AI) in healthcare. Hospitals are required to ensure that the use of AI solutions is transparent, traceable, and compliant with the law. This makes it clear: It’s not just the performance of the algorithms that matters, but also governance, interoperability, and comprehensive documentation. You decide whether AI can be integrated into clinical processes safely, in compliance with regulations, and cost-efficiency.

Hospitals Have a Duty as AI Operators

The EU AI Act takes a risk-based approach and sets high barriers, particularly for high-risk applications—which include most AI-supported medical devices used in diagnostics. Hospitals typically assume the role of operator (deployer) and are therefore obligated to systematically monitor operations, ensure the quality of input data, and retain the automatically generated logs.

Added to this is the obligation to provide evidence: Hospitals must disclose how an AI results were incorporated into medical decisions and how human oversight was exercised in the process. Access to unverified results may be granted exclusively to authorized personel. Beyond the technology itself, a well-thought-out authorization framework, a structured AI governance model, and the development of AI literacy among staff are becoming increasingly important. The real challenge, therefore, lies not in the algorithm itself, but in its seamless and audit-proof integration into everyday clinical practice.

Compliance: An Infrastructure for All Requirements

As a component of the “European Strategy for Data,” the EU AI Act is closely intertwined with other regulatory frameworks such as the Medical Device Regulation (MDR), the General Data Protection Regulation (GDPR), and the IT security requirements of the NIS 2 Directive. Hospitals therefore need a platform that inherently provides all of this as an infrastructure service, rather than having to retrofit it individually for each application. At the same time, how data is processed determines its usability—and thus the quality of diagnostics. Even in the age of AI, the adage “garbage in, garbage out” still holds true. Only clean, uniformly structured data streams lead to reliable results that stand up to legal scrutiny.

Interoperability: How Standards Break Down AI Silos

In hardly any other medical field is the use of AI as widespread as in radiology: from prioritizing critical findings to the automated detection of lesions. However, many solutions operate as isolated silos, use their own user interfaces, or output proprietary data formats. This makes consistent, cross-departmental documentation nearly impossible and increases the effort required for integration.

Established standards such as DICOM, HL7 FHIR, and, above all, the new IHE profile AIRA (AI Result Assessment for Imaging) bridge this gap. AIRA uniformly defines how AI results can be evaluated in a clinical context and logged across systems.

Deploying AI Where Diagnoses Are Made

In our view, AI should ideally be deployed where radiologists work: directly in the viewer. That is why JiveX supports the IHE AIRA profile. The JiveX Viewer serves as the central integration point:

• Unified Visualization: Results from various AI providers are displayed together in a familiar visual context—without media breaks or additional user interfaces.
• Human-in-the-Loop: Directly at the diagnostic workstation, physicians document—in accordance with the AIRA profile—whether they accept, adjust, or reject an AI suggestion.
• Auditability: This structured interaction and evaluation data is recorded in an audit-proof manner, thereby providing the required evidence of human oversight.

There’s no doubt about it: The EU AI Act forces hospitals to strike a balance between the momentum of innovation and legal certainty. But with the help of an infrastructure where progress and compliance go hand in hand, you can master this balancing act.