Utilization of AI Technologies in Radiology Across the Netherlands (2020–2022)
Overview
From 2020 to 2022, the adoption of AI products in Dutch radiology departments increased steadily, with usage rising from 20% to 33% of hospital organizations. The diversity of AI products implemented grew fivefold, and AI was primarily applied in chest CT, neuro CT, and musculoskeletal radiograph analyses. Financial and IT integration challenges remain the main obstacles to broader AI implementation.
Background
Radiology is a leading specialty in adopting commercially available AI-based products, with over 200 CE-marked AI tools currently on the market. While vendors claim these products improve efficiency and quality of care, evidence of clinical impact remains limited. Understanding the adoption patterns and barriers at the hospital level is essential to realize AI's potential benefits in radiology practice.
Data Highlights
Year
Respondents
Departments Using AI (%)
Unique AI Products
Cumulative AI Implementations
Average AI Products per Department
2020
43
20%
7
19
1
2021
36
28%
Not specified
38
2
2022
33
33%
34
68
3
Key Findings
The proportion of radiology departments using AI increased from 20% in 2020 to 33% in 2022.
The number of unique AI products implemented rose from 7 in 2020 to 34 in 2022, indicating growing diversity.
AI applications focused mainly on chest CT (nodule and embolism detection), neuro CT (stroke and hemorrhage detection), and musculoskeletal radiograph analysis.
In 2022, 28% of users reported health improvements from AI, and 32% reported both health improvements and cost savings; no respondents reported cost savings alone.
Financial constraints and IT integration issues were the most frequently cited obstacles to AI adoption throughout the study period.
Legal issues as a barrier decreased over time, while concerns about validation increased.
Clinical Implications
Clinicians and hospital administrators should anticipate ongoing financial and IT integration challenges when implementing AI solutions in radiology. Prioritizing AI products with demonstrated clinical value, particularly in chest and neuro imaging, may enhance adoption and patient outcomes. Continued evaluation of AI effectiveness and cost impact is necessary to justify investment and guide policy.
Conclusion
AI adoption in radiology departments across the Netherlands has steadily increased with expanding product diversity and applications. Addressing financial and integration barriers will be critical to realizing the full clinical benefits of AI technologies.
References
Dutch Society of Radiology & AIforRadiology.com -- Utilization of AI in Radiology Netherlands 2020-2022
At ARVO, Cecilia S. Lee, MD, MS, and Aaron Y. Lee, MD, MSCI, discuss barriers to AI deployment in ophthalmology, including interoperability and model development.
