The American College of Radiology (ACR) recently sent comments to the White House Office of Management and Budget (OMB) in response to OMB’s release of a draft memorandum, “Guidance for Regulation of Artificial Intelligence.” 

In addition to registering its support for the U.S. Government’s efforts to promote useful, safe, and effective AI innovation, the ACR outlined 10 priorities focusing on both regulatory and nonregulatory approaches to AI-powered and enabled technologies and industries.

First, the ACR said, actions surrounding AI must ensure continued public trust in the emerging technologies, and recommended the U.S. government collaborate with third parties – such as professional organizations representing end-users of AI – to develop certification measures, validation services and real-world performance monitoring agencies.

Next, the ACR articulated its continuing support fo federal processes that would ensure regulatory transparency and accountability.

Third, ACR voiced support for “scientific integrity in the federal rulemaking and guidance processes,” while also pointing to the importance of  best practices for regulatory agencies including “transparently articulating the strengths, weaknesses, intended optimizations or outcomes, bias mitigations, and appropriate uses of the regulated AI applications.”

Next, the  ACR agreed with the OMB that oversight approaches should be based on the application of risk assessment and management, and it emphasized that certain sectors or agencies, including healthcare, can have oversight gaps, and in these cases, third-party validation or certification can be helpful.

Fifth, the ACR argued that a major component of the government’s regulatory considerations related to AI should include the utility of the AI applications for the intended end-user communities. “For example, the ACR DSI (Data Science Institute) has worked with the radiologist community and other stakeholders to define publicly accessible imaging AI use cases for algorithm developers to create marketable solutions of high value to radiologist end-users. The U.S. Government should consider this program and strive to support similar collaborations across healthcare and in other sectors.”

Other priorities the ACR proposed included encouraging performance-based and flexible approaches to ensuring health and patient safety, the need to validate and monitor datasets used to train AI algorithms,  while pivoting to support rapid changes and updates is critical for regulatory bodies, “full disclosure and transparency of any premarket AI review, including ensuring a significant level of data traceability and insight into the training data used in creating new models,” regulatory agency coordination that is transparent to public stakeholders, AI developers, and users.

“The top concern for any AI used in healthcare,” the ACR emphasized finally, “must be patient and public safety.”