Specialists in radiology, AI can change healthcare forever—if they do this one thing

The worlds of radiology and artificial intelligence (AI) are at a bit of a crossroads, according to a new analysis published in the Journal of the American College of Radiology. While experts from both fields remain focused on making technological breakthroughs, the actual relationship between radiology and AI is not getting the attention it deserves.

“Although new investigations on the potential application of AI in radiology are published regularly, little focus has been made toward navigating this new intersection between the tech world and AI and radiology,” wrote lead author Paul H. Yi, MD, of the Johns Hopkins University School of Medicine in Baltimore, and colleagues.

The authors wanted to gain a better understanding of where the relationship between radiology and AI stands at this particular moment in time, so they developed a “word cloud” by looking at the titles of the top 25 nonscientific news articles found with a simple Google search of “artificial intelligence radiology.” Four themes were identified: radiologists, AI and machines, optimism and ambition, skepticism and realism.

Reviewing each of these themes, Yi and colleagues determined that one thing is absolutely vital to the future of radiology’s relationship with AI: collaboration. If specialists from the worlds of radiology and AI can work together, instead of against one another, the sky might truly be the limit.

“There are strengths in both radiologists and in AI, as well as truths in both views of optimism and skepticism toward the application of AI in radiology,” the authors wrote. “We propose that these differences can be complementary. For example, in clinical practice, an AI algorithm with a high negative predictive value may be a cost-effective tool to detect and ‘remove’ normal or negative examinations from radiologists’ queues, enabling them to focus more of their time toward interpreting abnormal examinations. Such a strategy would prove particularly useful for examinations with a high rate of negative studies, such as head CT and MRI for evaluation of headache, of which 1.5 million are ordered annually in the United States with only 1 percent to 3 percent of studies yielding significant findings.”

Radiologists also have an opportunity to actively manage the integration of AI into current workflows, according to Yi et al. Will the specialty take advantage of that opportunity?

They also have a similar opportunity in regards to reimbursement models. Radiologists can work together with “the tech world, insurance companies and healthcare administration” to work out the finer details of such reimbursement—or they risk being left out of the discussion altogether.

“By combining the expertise of radiologists with the innovation of the tech industry, we can achieve more than we might alone,” the authors concluded. “To do so, however, will require mutual understanding and collaboration between the two groups, with increased realism among the tech sector about the difficulties of applying AI to radiology and increased optimism among radiologists to the technological possibilities of AI.”