Governance model for artificial intelligence applications in diagnostic imaging
The goal of this project, funded by HUS and Sitra, is to create a nationally unified, cost-effective, and secure governance model for the use of artificial intelligence in diagnostic imaging.
The project is based on a systematic AI governance model developed at the HUS Diagnostic Center for the evaluation, testing, and deployment of artificial intelligence applications in diagnostic imaging. As part of the project, a fracture detection pilot will be conducted in collaboration with the Wellbeing Services County of Päijät-Häme.
Project name
National scaling of the governance model for artificial intelligence applications in diagnostic imaging
Project duration
June 1, 2026 – December 31, 2027
Project funding
Sitra and HUS
Project team
The project team consists of the Diagnostics Center’s artificial intelligence team and digital radiologists, who have multidisciplinary expertise in the fields of technology, clinical work, data management, quality control, and regulatory affairs.
Significance of the project
Commercial artificial intelligence applications are already widely used in radiology, and there are already more than 300 CE-marked AI-based medical devices on the market. Artificial intelligence has significant potential to streamline diagnostics, free up staff time, and improve the efficiency of care processes amid growing demand for services. Despite this, the use of AI applications in clinical practice remains fairly limited.
Key barriers and obstacles to deployment include:
- the challenges of integrating artificial intelligence into existing workflows, adapting workflows to take advantage of artificial intelligence, and the permanent implementation of artificial intelligence in workflows
- established work practices and resistance to change
- limited resources and funding
- difficulties in translating the characteristics of products on the market into concrete clinical benefits
- inadequate training and skills in the use of new technologies
- a lack of standardized approaches to the evaluation, deployment, ongoing monitoring, and impact assessment of AI applications, which can lead to duplication of work, inconsistent quality, and the risk of implementing clinically or economically ineffective solutions, which can lead to suboptimal outcomes and jeopardize patient safety.
Identifying and eliminating high-risk or ineffective products is a key component of a high-quality governance model.
At its best, artificial intelligence can significantly improve the efficiency of diagnostic imaging services, enhance staff wellbeing, reduce manual work, improve the quality of diagnostics, and speed up patients’ care pathways. With the help of artificial intelligence, it is possible, for example, to:
- reduce imaging times and speed up workflows and turnaround times
- increase imaging capacity using existing resources, which will result in cost savings
- improve image quality and patient safety and enable the use of lower radiation doses
- systematically identify deviations and generate data for the continuous improvement of workflows, guidelines, and operating procedures, while also enabling entirely new types of streamlined workflows.
The goal of a unified governance model is to ensure that artificial intelligence is used in a controlled manner that delivers real benefits to patients, healthcare professionals, and healthcare organizations.
Expected project results
The project will develop a nationally scalable governance model for the evaluation, deployment, and continuous performance monitoring of diagnostic AI applications.
The national deployment and training package will be published for open and free use. The deployment and training package includes process descriptions, metrics, guidelines, report templates, training materials, and other instructions to support the evaluation, deployment, and monitoring of AI solutions.
The Päijät-Häme pilot will evaluate the effectiveness of the governance model in a variety of operating environments and generate evidence on the impact of artificial intelligence on the performance, usability, and productivity of fracture diagnostics.
The project will establish a nationwide collaboration network and conduct a study on the suitability of artificial intelligence platforms and various procurement procedures for nationwide joint procurement, which could reduce redundant administrative work, ICT costs, and also enable smaller wellbeing services counties to participate with fewer resources.
Based on the project’s results, the next steps will be to identify needs and draft recommendations for implementation of nationwide changes.
The governance model for artificial intelligence applications in diagnostic imaging to be developed in this project brings together the testing and evaluation, deployment, and quality control of artificial intelligence applications, as well as their monitoring during use. This model will reduce duplicate work, clarify responsibilities, lower the indirect costs associated with administration, and support the verification of the benefits of artificial intelligence through automated metrics.
The governance model is based on a multi-tiered approach that takes into account supplier reports, comparisons of AI-generated findings with radiology reports, and clinical impact analytics.
The governance model is based on an AI testing and deployment process developed at HUS and validated through scientific peer review, which has been used to evaluate approximately 60 commercial AI applications to date. Some of these solutions have been implemented in production because they have demonstrated objective and tangible productivity and quality benefits through the process. At the same time, the process has prevented or halted the implementation of dozens of solutions that were unsuitable for clinical use or uneconomical.
Additional information:
Testing process for artificial intelligence applications in radiology practice
Deployment process for artificial intelligence applications in radiology practice
As part of the project, a fracture detection pilot will be conducted in collaboration with the Wellbeing Services County of Päijät-Häme. The goal of the pilot is to assess the scalability of the governance model in various operating environments and to gather evidence on the impact of artificial intelligence on the quality, usability, and productivity of fracture diagnostics.
If you would like to receive updates on the project’s progress or invitations to collaborative events in the future, please send your contact information to the Project Manager, Mirva Savonen.
Learn more about the project and the use of artificial intelligence applications:
HUS Grand Rounds in Hand Surgery, November 4, 2026: Artificial intelligence at HUS Radiology, Suvi Syväranta, Chief Radiologist, Digital Process Owner for radiology tools, Diagnostic Center
Public Procurement Forum, November 12, 2026: How to combine procurement and subject matter expertise into an innovative competitive bidding process, Suvi Syväranta, Chief Radiologist, Digital Process Owner for radiology tools, Diagnostic Center
Trauma Imaging Training Day (SORF), November 20, 2026: Artificial Intelligence in Trauma Imaging – Where Are We Now? Suvi Syväranta, Chief Radiologist, Digital Process Owner for radiology tools & Mirva Savonen, Project Manager, Diagnostic Center
SRY AI Section Webinar, December 1, 2026: Developing a National Governance Model for Artificial Intelligence in Diagnostic Imaging – Survey Results. Mirva Savonen, Project Manager, Diagnostic Center
Duodecim training event at the Head and Neck Center, December 10, 2026: The deployment process of artificial intelligence applications in radiology practice, Suvi Syväranta, Chief Radiologist, Digital Process Owner for radiology tools, Diagnostic Center
Updated: 02.11.2026