PITTSBURGH – Quality management is needed for successful use of AI in medical imaging, according to a presentation given June 9 at the Society for Imaging Informatics in Medicine (SIIM)-American College of Radiology (ACR) Data Science Summit.
Tessa Cook, MD, PhD, in her keynote shared resources from the ACR Data Science Institute (DSI) and how they help establish practice parameters for radiology departments using AI in their workflows.
“AI in imaging has reached the stage where you need quality management,” said Cook, who is the incoming chair of the ACR’s Commission on Informatics. She is also the vice chair of practice transformation at the University of Pennsylvania in Philadelphia. “Imaging AI accreditation is going to require thoughtful development. The process has started, but there is plenty of work still to be done.”
The ACR in recent years has released AI-focused resources for radiologists to use. One is the ACR Recognized Center for Healthcare-AI (ARCH-AI) quality assurance program, which sets guidelines for AI use in imaging interpretation that promote safety and efficacy. Requirements include establishing a governance team, inventory, testing, and integration into workflows.
Tessa Cook, MD, PhD, discusses recent work toward AI accreditation from efforts by the ACR and SIIM.
A 2025 paper that Cook cited stated that insights from ARCH-AI will inform the development of the formal accreditation program, which will lead to ACR Council approval, “currently anticipated in spring 2027.”
SIIM and the ACR in May issued a joint Practice Parameter for Imaging Artificial Intelligence (AI). This covers AI tool selection, pre-deployment evaluation, ongoing performance monitoring, and patient privacy protection, with facilities able to earn ARCH-AI designation through use of the practice parameter.
Cook also spoke about Assess-AI, a quality registry that monitors clinical AI in practice. She said 26 healthcare systems representing 171 facilities are currently enrolled in Assess-AI. The registry supports multiple imaging AI use cases, with participating facilities being able to visualize data completeness, monitor longitudinal concordance trends, compare performance with registry benchmarks, and explore discordance by demographic or technical factors.
Why the need for quality management?
Cook said accreditation programs such as the ACR’s aim to improve patient care and acknowledge practices that meet quality standards. This is because premarket testing of AI products does not always translate to real-world performance due to a variety of reasons. This is where quality planning, assurance, control, and improvement come into play for continuous review toward high-quality care.
Cook shares her wish list for what attendees of the SIIM-ACR Data Summit, as well as SIIM 2026, should take back to their practices.
Cook added that the ACR’s quality standards go through consensus development, are approved by the ACR Council, and undergo objective determination based on feedback from users. These standards can apply to not just radiologists and radiologic technologists, but also data scientists, manufacturers and developers, administrators, qualified end-users, IT professionals, and experts in legal, compliance, and privacy.
“You don’t just build accreditation overnight,” Cook said. “The technology that you are building accreditation for must be used in practice.”
Future plans
Cook said the DSI is further developing the accreditation process and will seek ACR Council approval.
Along with the SIIM-ACR Practice Parameter on Imaging AI, the ACR Council in May adopted the development of its accreditation program. Cook said work toward successful implementation of these will be ongoing.
“The DSI team is continuing to work on new use cases for Assess-AI…and we are working on how we can provide better reference standards for AI outputs in the form of [electronic health record] integration,” she said.
Check out AuntMinnie's full coverage of SIIM 2026 on our conference page.
