AUSTIN, TX - In a video interview, Dr. Eliot Siegel of the University of Maryland shared his highlights from the Society for Imaging Informatics in Medicine's Conference on Machine Intelligence in Medical Imaging (C-MIMI). He also offered his perspective on what's needed to drive the clinical adoption of artificial intelligence (AI) in radiology.
C-MIMI 2019 featured a number of high points, including a keynote presentation by Dr. Shez Partovi of Amazon Web Services; an update on regulatory policy for quantitative imaging and machine-learning tools by Nicholas Petrick, PhD, of the U.S. Food and Drug Administration (FDA); and a vendor panel on driving AI adoption in clinical practice, according to Siegel, who served as a conference co-chair.
It was exciting to hear that the FDA is looking more closely at continuously learning AI algorithms for medical images and related information, Siegel said.
"And so what that means is that rather than being frozen with a particular version of software that had been trained on a dataset, there's the potential for me, for example ... to be able to have the system watch the cases that I have at the University of Maryland, get feedback from me and my colleagues, and then be able to learn and potentially have predictions that are more specific to my population maybe than what were made initially on the dataset that the algorithm was trained on," he said. "So the possibility to finally realize this whole idea of machine learning where the machine continues to learn and benefit from the feedback from our own patient population and the radiologist's feedback I think is really exciting."
The vendor panel on driving the clinical adoption of AI also featured interesting discussions on a range of topics, including predictions and what's needed to overcome current barriers to implementation, according to Siegel.
"One of the things they had suggested that I found interesting was that although AI hasn't really arrived yet at this point, they seemed really confident that within one to three years, we'd be reaching a point where we might have major adoption -- say 50% or more of facilities significantly using AI," he said.
Siegel also shared his thoughts on what the radiology community can do to spur the process of bringing AI into routine clinical practice, including allaying the concerns of radiologists and figuring out the providers and standards that are going to be used for incorporating algorithms into clinical practice. Platforms and workflow orchestration will also be critical, he said.
