Subscribe to WHX Insights newsletter
Stay at the forefront of healthcare trends & innovations
Guiding imaging AI from innovation to impact
The American College of Radiology is setting global standards for quality, safety and governance
4 Min Read
Supplied
Radiology became the first truly digital medical speciality with the transition from film to picture archiving and communications systems (PACS) nearly a quarter century ago. This continued journey through digital image generation, communication, and digital storage of imaging information enabled quantum leaps in advanced image processing and reporting efficiency, setting the stage for the eventual emergence of artificial intelligence (AI) in radiology.
When deep-learning technology converged with large-scale graphics processing unit (GPU) training in 2012, hyperscalers and startups pivoted towards augmented intelligence. These technologies quickly entered digital medicine, particularly radiology. The US Food and Drug Administration’s Software as a Medical Device framework, introduced in 2016, reflected expectations of widespread adoption.
Early radiology AI products, however, were largely cleared for workflow optimisation and case triage rather than for enhancing diagnostic accuracy or radiologist efficiency. Still, proper triage demonstrated value in select cases by reducing time to treatment for positive cases, while radiologists began managing growing portfolios of AI tools.
A further breakthrough followed in 2017 when the arrival of transformer technology. This was incorporated into products such as ChatGPT, which launched publicly in late 2022, generating unprecedented interest in the consumer space, and also highlighting the potential of large-scale medical text applications. Subsequent evolution has led to what may prove to be the next true game-changers: radiology foundation models.
These scalable models, trained on vast datasets, can be adapted to a wide range of radiology-specific tasks, including complex cognitive functions such as image-to-text translation. Looking ahead, clinical practice is likely to see a convergence of diagnostic AI and foundation model technology, tailored to the problems radiologists need to solve.
ACR AI Initiatives
The American College of Radiology (ACR) Data Science Institute (DSI) has closely tracked these developments from the outset, supporting ACR members and radiology practices through numerous programmatic approaches, educational offerings, technology development, as well as economic and regulatory advocacy, while engaging with payers, industry and government agencies.
One of the earliest initiatives, the Define-AI Directory, involved practicing radiologists in creating structured clinical use cases, providing valuable insights for imaging AI product development. ACR AI-LAB was subsequently launched as an educational and simulated environment hosted on ACR infrastructure, enabling members to experience the entire AI product lifecycle, from model training to validation to inference, performance measurement and federated learning.
ACR also established AI Central, a searchable online repository of AI products designed to support more informed product comparison and purchasing decisions. ACR also convened expert groups to create guiding papers on the ethics of AI, data sharing, and the reliable and transparent use of imaging AI, while participating in international efforts to promote these principles across multiple continents.
In 2024, ACR launched ARCH-AI, the first national programme recognising AI quality assurance in radiology facilities. The initiatives brings together practices actively sharing implementation experiences, challenges, and solutions. To address the difficult task of post-deployment monitoring and performance validation, ACR also introduced Assess-AI, the first national imaging AI registry operated by the college. The registry enables automated, large-scale analysis of AI performance, providing practices detailed analytics and root-cause insights into performance changes over time.
ACR is also a founding member of the Healthcare AI Challege Consortium, whose computational platform provides a secure environment for radiologists to evaluate and grade foundation-model performance on real-world radiology tasks. As foundation models enter clinical workflows, robust local physician engagement and oversight will be essential to ensure their safe and effective use.
Taking ACR imaging AI initiatives to the world
New for 2025, ACR now offers a low-cost International Informatics Membership, opening the door for radiologists worldwide to engage directly with the college. International members can participate in committees and join the dialogue shaping the future of AI in radiology and healthcare IT.
ACR is also extending ARCH-AI recognition internationally, enabling practices around the world to collaborate with ACR staff on quality assurance processes while sharing regional perspectives. This growing ecosystem supports a deeper understanding of the global imaging AI workforce and helps establish the governance frameworks needed to protect patients while accelerating innovation.
Bottom line
AI is a global phenomenon, and ACR is stepping forward with its first-ever international engagement offering. Now is the time for radiologists worldwide to benefit from shared dialogue and collective insights. We are bringing ACR's decade-long investment in imaging AI to the global radiology community — and we want you to be a part of it.
