The American College of Radiology Data Science Institute (ACR DSI) recently released of a series of standardized artificial intelligence (AI) use cases to help advance imaging in AI. Down the road, they could help create an “AI ecosystem” for radiology, wrote Bibb Allen, MD, chief medical officer of the ACR DSI, in a recent Journal of the American College of Radiology editorial.
Allen explained these structured use cases are important because they will greatly assist in driving the adoption of AI tools in clinical practice.
“A structured AI use case includes a narrative description and flowcharts that define exactly how an AI algorithm takes in images and/or other information from the clinical workflow and provides specific output to end users,” wrote Allen, a radiologist at Grandview Medical Center in Birmingham, Alabama. “Structured AI use cases also include parameters for how algorithms are trained, tested, and validated for regulatory approval and clinical use; how they are deployed into clinical workflows; and how their effectiveness can be monitored in clinical practice.”
Structured AI use cases also allow radiologists to be leaders in assisting developers to create AI algorithms appropriate for clinical practice which can enhance clinical value for radiology patients and healthcare systems, Allen noted.
Currently, the ACR DSI has been helping radiology professionals develop clinically relevant AI use cases aimed to improve patient care and increase the value of radiology professionals to their healthcare systems.
Additionally, the ACR DSI developed Technology Oriented Use Cases in Healthcare AI (TOUCH-AI). This open-framework authoring system defines AI use cases for radiology that pose problems solvable by AI and presents high-value clinical needs.
“The TOUCH-AI framework provides detailed descriptions of the goals the algorithm should meet, the required clinical inputs, how the algorithm should integrate into the clinical workflow and how it should interface with both human end users and an array of electronic resources, such as reporting software, PACS and electronic health records,” Allen wrote.
The process helps ensure two things: patient safety and that selected AI use cases have data elements for effective clinical integration using existing workflow tools.
To facilitate the involvement of radiology professionals in the AI use case development process, the ACR DSI established 10 data science subspecialty panels composed of clinical experts to evaluate and choose the highest value use-case proposals for development, according to Allen.
“By collaborating through radiology specialty societies, radiologists can develop larger pools of thought regarding the highest priority for use cases for the radiologic sciences,” Allen wrote. “Additionally, individual developers and institutions can submit use cases they are currently working on to the ACR DSI and have them encoded with the data elements specifying standardized methods for annotation of training sets, validation, integration and monitoring in clinical practice in order to facilitate clinical deployment of their ideas and algorithms.”
Fifty AI use cases were released this month by the ACR DSI, including cases in radiation oncology, interventional radiology and all diagnostic radiology subspecialties. The use cases are open to developers. The FDA has also been working with the ACR DSI to define use cases as part of the FDA Medical Device Development program for facilitating FDA clearance of AI software.
“The ACR DSI structured use-case development program is a cornerstone in the creation of an AI ecosystem for radiology,” Allen wrote. “Structured AI use cases for the radiologic sciences can convene multiple stakeholders, ensure patient safety, promote diversity in algorithm development, and foster collaborations with federal regulatory agencies and even Congress to facilitate the introduction of AI algorithms into the market, which will enhance the care radiology professionals provide for their patients.”