ACR Council approves imaging AI practice parameter

The American College of Radiology Council has approved a new practice parameter for imaging AI, creating formal guidance for how radiology practices can deploy and manage AI tools in clinical care.

Approved at ACR 2026 in Washington, DC, the ACR-SIIM Practice Parameter for Imaging Artificial Intelligence was developed with the Society for Imaging Informatics in Medicine. The document applies to physicians, technologists, medical physicists, informatics teams, IT teams, data scientists, administrators, and others who deploy AI or use AI results in imaging workflows.

The guidance covers how practices can select AI tools, evaluate them before deployment, monitor performance over time, and protect patient privacy. It also outlines operational steps for creating an AI governance group, maintaining an inventory of AI tools, running local acceptance testing, tracking real-world performance, defining stop rules, and following HIPAA privacy and security requirements.

“This first-of-its-kind ACR-SIIM Practice Parameter outlines steps that imaging facilities can follow to help implement, use, and continually update AI,” said Tessa Cook, MD, PhD, chair of the practice parameter writing committee.

Facilities that implement AI responsibly may also seek the ACR Recognized Center for Healthcare-AI, or ARCH-AI, designation. ACR described ARCH-AI as an international AI facility quality assurance program intended to help sites adopt clinical AI with structured governance and quality processes.

The ACR Data Science Institute also published a Journal of the American College of Radiology article describing the technical framework for Assess-AI, which ACR calls the world’s first AI quality registry and data service for medical imaging.

Assess-AI supports postdeployment AI governance by measuring concordance between clinical AI outputs and radiology report-derived surrogate labels. The service uses deidentified data through ACR Connect, centralized analytics, and benchmarking so facilities can compare AI performance against national and peer-site data.

The registry is designed to monitor model inputs, versioning, concordance with radiology reports, site-level performance, subgroup performance, and variation over time. ACR said the system also provides dashboards for AI oversight teams to support governance and due diligence.

Current Assess-AI use cases include intracranial hemorrhage, pulmonary embolism, pneumothorax, large vessel occlusion, bone age, cervical spine fracture, breast density, pneumoperitoneum, tube malposition, pleural effusion, brain mass effect, and obstructive hydrocephalus.

Nabile Safdar, MD, SIIM board chair, said responsible use of AI in medical imaging is “an ongoing process rather than a single event.”

ACR said it will continue working with SIIM, the U.S. Food and Drug Administration, Congress, and other stakeholders on radiology AI implementation and oversight.