Higher Than Technology: The Real Challenges of AI Adoption in Radiology

AI-assisted reading technology has already matured considerably. As we explored in PART 1, experts acknowledge its clinical value. Yet actual adoption in hospitals often feels slow. Why?
European experts point to issues beyond the technology itself — administrative complexity, economic models, and trust — as the larger barriers. This article examines these real-world challenges and emerging solutions.
 

Barrier 1: Administrative and Regulatory Complexity

Deploying new software in a hospital environment requires navigating IT approval processes, data privacy regulations, and compliance procedures. Prof. Marie-Pierre Revel identifies this process itself as a significant obstacle.

“The difficulty of convincing IT managers, explaining there are no GDPR issues — that whole administrative and regulatory part is a real challenge in the adoption process.”
— Marie-Pierre Revel, Professor of Radiology Hôpital Cochin, Université de Paris

 

Barrier 2: The Absence of an Economic Model

Using AI tools in clinical practice is not free. Clear economic models covering software licensing, infrastructure, and operational costs are needed. Prof. Revel highlights this often-underestimated challenge.

“Something we tend to underestimate is that we’ll need to build a proper economic model. Using AI won’t be free for radiologists, and they’ll need to be financially compensated.”
— Marie-Pierre Revel, Professor of Radiology Hôpital Cochin, Université de Paris

This economic challenge is finding a breakthrough in Germany. Prof. Vogel-Claussen explains how changes in the reimbursement landscape are poised to remove this barrier.

“Starting in April 2026, lung cancer screening with AI will be covered by public insurance in Germany. For the first time, AI used in lung cancer screening will be reimbursed. I believe that barrier is about to disappear.”
— Jens Vogel-Claussen, Director, Department of Radiology Charité – Universitätsmedizin Berlin

 

| The Korean Landscape: Deferred New Medical Technology System

In Korea, AI-assisted reading reimbursement is also progressing. Through the Deferred New Medical Technology evaluation system, temporary non-reimbursed early market entry is now possible, and chest CT AI analysis technology has already been designated. Germany’s public insurance coverage and Korea’s non-reimbursed early entry pathway both point to a shared global trend: the accelerating reimbursement of AI in radiology.

 

Barrier 3: Trust and Quality Assurance

It is difficult to adopt an AI system without confidence in whether it can be trusted. Experts from both sides highlight this fundamental challenge.

“We may have a sort of a priori reluctance in using these tools, because we first need to know if we trust the system or not.”
— Martine Remy-Jardin, Thoracic Imaging Specialist, France President-Elect, Fleischner Society (2027)

“There are different AI systems with different levels of quality. And there is still no quality assurance system in Europe, so we have homework to do.”
— Jens Vogel-Claussen, Director, Department of Radiology Charité – Universitätsmedizin Berlin

On the other hand, when organizational support is in place, adoption barriers are significantly reduced.

“We have seven biomedical and data engineers helping us implement AI tools within our department and workflow. I have to say, I haven’t encountered any major barriers.”
— Luis Gorospe, Radiologist Ramón y Cajal University Hospital, Madrid

[Watch the Full Interview] Click here to see the experts discuss the future of AI in radiology.


| PART 2 Summary
The real barriers to AI adoption are not technical but administrative, economic, and trust-related. Germany’s public insurance coverage and Korea’s deferred evaluation pathway demonstrate a global trend toward resolving the reimbursement challenge.
➤ In PART 3, we look at how AI is actually saving lives in lung cancer screening, with clinical evidence and firsthand experience.