Much of the College’s work to support and guide the use of AI in radiology practices is conducted somewhat behind the scenes. But by being the center of a multipronged collaboration made up of regulatory agencies, industry, patients and radiologists, the ACR continues to play a vital role in how AI is used in practices.
“AI used appropriately in a safe environment, in a way that mitigates bias and is available to all of our patients, can be transformative,” says Bibb Allen Jr., MD, FACR, ACR Data Science Institute® (DSI) chief medical officer. “But at each turn, there are steps that will need to happen in the regulatory process and in the research phase (such as the ability to validate models and generate outcomes data). These are things the ACR has always been good at.”
In light of the prediction that AI would significantly impact the way radiologists do their jobs, the ACR DSI was established in 2017, with Keith J. Dreyer, DO, PhD, FACR, leading the initiative as its chief science officer. “We began the DSI without knowing when and exactly how AI would influence our work,” says then-BOC Chair James A. Brink, MD, FACR. “Six years later, I am struck by how rapidly AI has entered society, never mind medicine, in ways we never anticipated — the large language models such as ChatGPT and the option to process images through other models. I think we were prescient to some degree in 2017, and it’s a good thing because the world is evolving even more rapidly.”
Working Together
In 2014, the ACR formed the Commission on Informatics. Commission Chair Christoph Wald, MD, PhD, MBA, FACR, sees the DSI as the external-facing entity of the Commission. The DSI interacts with industry trade groups, such as the Medical Imaging and Technology Alliance, regulators, the Office of the National Coordinator for Health Information Technology and the FDA, among others.
“The ACR is in frequent, regular conversations with these groups,” Wald says. “That’s a unique aspect for the College — we have this position of being the honest broker because were present such a large number of radiologists. In addition, our robust ACR Government Relations arm works closely with these groups when they ask for input on data science, AI-related issues or issues such as image sharing. We’re an important conduit to and from the government.”
Accomplishing Big Things
In the next 20 years, AI is going to transform how we take care of patients, but it is not going to take care of them for us, according to Allen. “This is one more step in adding value, one more tool that lets us have a bigger leadership role in transforming how radiologists practice,” he says. “I think the DSI has played a key role in that. Keith’s vision was extraordinary, as it usually is, and to see us now six years later recognized as the organization people want to talk towhen they’re interested in AI is quite an accomplishment.”
Allen cites the College’s work in creating standards and aligning regulatory agencies during the early days of digital mammography as an example of the ACR’s strengths. “In the ’70s, mammograms were still being done with standard radiographs, so we created the Mammography Accreditation program,” Allen says. “We worked with Congress to make a law that said safe and effective mammography was dependent on following these standards. That went to the FDA, which said, ‘OK, we’re going to adopt the program.’ Then you look at the Medicare Improvements for Patients and Providers Act of 2008 and standards for imaging, and facilities have become accredited to provide safe and effective care in advanced imaging.”
The greatest accomplishment related to DSI and the Commission on Informatics, according to Wald, is the mere fact the ACR took a leadership stance early on in identifying AI as a game changer and putting pieces in place to manage related changes. “We established the DSI as the go-to entity to have discussions about this technology,” Wald says. “We convened stakeholders, whether they be practicing radiologists, radiology leaders, government agencies or industry, and we made a home for all of them in this space that also leverages support from other organizations like the RSNA and the Society for Imaging Informatics in Medicine, bringing together committees that deal with the same issues and coordinating with each other to avoid duplication of effort.”
AI used appropriately in a safe environment, in a way that mitigates bias and is available to all of our patients, can be transformative.
Guiding the Integration of AI
Although AI algorithms are already helping radiologists better manage worklists, including the remarkable ranking of critical emergency room studies based on degree of concern, some of the most impactful uses of AI in radiology are still to come.1
“I’m very excited to think about what ChatGPT-type technology can do to make clinical decision support something that is fully utilized in practice,” says former ACR BOC Chair Geraldine B. McGinty, MD, MBA, FACR. “We don’t want for a set of symptoms. I want them to be typing in a patient’s history and for ChatGPT or the like to say, ‘You should think about a CT for this patient’ or ‘This patient would benefit most from an MR study.’”
The predominant type of AI in current practice today solves individual diagnostic tasks and problems, Wald says. “I’m more excited about AI that will either solve multiple tasks at once, or quantitative AI, or AI that can handle things radiologists can’t easily do.” Examples, he says, include looking at the liver and quantifying the amount of fat, iron or inflammatory activity or facilitating opportunistic screening.
Through it all, the expectation is that the ACR will be the authority to guide radiologists as they implement practice changes related to AI. “I want the ACR to be a trusted place for patients to feel at ease if the ACR says an algorithm is safe and effective because we’ve built trust with patients around accreditation,” McGinty says. “And I want us to take the patient trust, the ACR infrastructure and the relationships with the regulatory agencies and apply what we’ve done in accreditation to AI.”
Former BOC Chair James H. Thrall MD, FACR, echoes that outlook for the future: “The ACR has been the organization in imaging that sets standards, and I see that happening again with AI.”
Using AI Tools to Improve Practice
Can AI replace radiologists? In 2015, Pedro Domingos wrote in his book The Master Algorithm, “It’s not man versus machine; it’s man with machine versus man without.”2
Allen says that concept is applicable to the emerging use of AI in radiology. “Radiologists who use AI-enabled tools will be able to integrate more information into their practices than ever before, which will allow them to take better care of their patients,” he says. “And their specialty society should play an important role in facilitating the transformation.”