The American College of Radiology (ACR) Data Science Institute (DSI) is on a mission to implement artificial intelligence (AI) into clinical practice, according to a new analysis published in the Journal of the American College of Radiology.

At the ACR’s annual meeting in 2017, the creation of the DSI was officially announced, with a primary goal of working with radiologists and other industry leaders to “guide and facilitate the appropriate development and implementation of AI tools to help radiologists improve medical imaging care.”

“The institute and the commission stand ready to leverage AI and machine learning to advance the profession and improve patient care by using radiology professionals’ expertise to define pathways where AI will benefit our patients the most, and we can rest assured that the work of the college on AI and machine learning will confound those who see no role for radiologists in the future,” wrote co-authors Bibb Allen Jr., MD, chief medical officer of the ACR DCI, and Geraldine B. McGinty, MD, vice chair of the ACR Board of Chancellors.

The ACR DSI has four methods for achieving its goal of implementing AI in clinical practice.

1. Develop appropriate use cases and workflow integrations that improve patient care. Currently, the ACR DSI is looking to define use cases that will inform the development of algorithms to be used in clinical practice. For each subspecialty within radiology, the authors mention the ACR DSI will look at areas where imaging can be improved while noting the scenarios positively impacted by AI. The use cases contain data necessary to train algorithms and data that will be integrated into clinical workflow.
2. Protect patients through leadership roles in the regulatory process. To achieve algorithm verification, the ACR DSI will partner with other institutions to form verification data around the use cases developed. The data sets will contain images that aren’t part of the algorithm training. The ACR DSI will then partner with institutions willing to undertake algorithm verification.
3. Establish industry relationships by providing credible use cases, verification processes, and pathways for clinical integration. “The ACR has a long history of developing relationships with government regulators including the FDA and has created a number of informatics tools and frameworks for integrating clinical recommendations into our radiology reports,” the authors wrote. “The ACR can leverage these relationships and tools into initiatives that will facilitate the implementation of AI into clinical practice that will augment the value radiologists provide to our patients.”
4. Educate radiologists and all other stakeholders both the ACR’s role and the overall value of AI. Case development is important to the mission, but the authors explained that the ACR must also work with other leaders to both implement AI in clinical practice convey how important AI is to providing value-based care.

“The ACR is the voice of the profession of radiology with policymakers, and it is leveraging the credibility and reputation it has built to inform and educate principal stakeholders, such as the FDA, about the importance of rigorous algorithm validation and the need to put our patients at the forefront of decision making,” the authors concluded.