ACR Launches First AI Practice Parameter and Global Quality Registry

TL;DR: The American College of Radiology (ACR) approved the inaugural ACR‑SIIM Practice Parameter for imaging AI and introduced Assess‑AI, the first global AI quality registry, to standardize AI use in radiology.

The ACR Council voted on the new practice parameter at its 2026 annual meeting in Washington, DC. The guideline, developed with the Society for Imaging Informatics in Medicine (SIIM), outlines step‑by‑step procedures for selecting, testing, deploying, and continuously monitoring AI tools in imaging departments.

Simultaneously, the ACR Data Science Institute published a technical paper describing Assess‑AI, a registry that aggregates de‑identified AI performance data from participating sites. The service links to ACR Connect, extracts surrogate outcome labels from radiology reports using large‑language‑model prompting, and provides interactive dashboards for longitudinal performance tracking.

Key components of the practice parameter include establishing an AI governance committee, maintaining an inventory of all AI applications and versions, conducting local acceptance testing before clinical rollout, and defining stop‑rules for model drift or safety concerns. Facilities that follow the protocol can apply for the ACR Recognized Center for Healthcare‑AI (ARCH‑AI) designation, joining a network that shares best practices and benchmark data.

Assess‑AI expands the ACR National Radiology Data Registry portfolio. It currently supports use cases such as intracranial hemorrhage detection, pulmonary embolism screening, bone‑age assessment, and breast density estimation. By collecting anonymized patient demographics, exam metadata, and report‑derived labels, the registry creates national benchmarks that sites can compare against their own results.

Christoph Wald, MD, PhD, MBA, highlighted that the platform lets facilities “manage performance, product selection and risk” through real‑time analytics. The closed‑loop workflow also enables local review of discordant cases via ACR Forensics, feeding insights back to AI developers for model refinement.

For clinicians, the combined rollout means a clear pathway to adopt AI responsibly while gaining visibility into how each algorithm performs in real‑world settings. Hospitals can now quantify AI impact on diagnostic accuracy, reduce variability, and demonstrate compliance with privacy regulations such as HIPAA.

What to watch: Adoption rates of the ACR‑SIIM parameter and enrollment in Assess‑AI will reveal how quickly radiology departments standardize AI governance and whether national benchmarks drive measurable improvements in patient outcomes.