As the annual conference of the Radiological Society of North America (RSNA) opened this week at Chicago’s vast McCormick Place Convention Center, a palpable shift in attention was evident, as the demand for value from radiological healthcare services appeared to accelerate not only in the U.S., but across the advanced industrialized world.
A strong surge in emphasis on artificial intelligence and machine learning was evident everywhere at RSNA 2017, whose attendance, as measured in terms of total advance registration, was almost exactly identical to what it had been in 2016 (at 48,445 this year, compared to 48,888 last year; and with 23,097 professional registrants this year, compared to 23,656 last year). The emphasis was evident in keynote addresses and presentations, and absolutely evident in the exhibit halls (in which 667 exhibitors were present, compared to 691 last year), where large banners and bold signage proclaimed a new era of innovation around new forms of supportive intelligence.
Indeed, the topic of artificial intelligence/machine learning, in the context of transformation of the profession, was an explicit element in the president’s address given on Sunday, Nov. 26, by current RSNA president Richard L. Ehman, M.D. Under the headline, “Is It Time to Reinvent Radiology?” Dr. Ehman, according to the report published Monday morning in the Daily Bulletin, the association’s official publication, told his audience on Sunday that “No other major medical specialty is so profoundly based on a remarkable invention that changed the world. And it is your collective focus on the well-being of our patients,” he told the radiologists gathered at the convention center, “that drives all of our efforts and innovations.”
Dr. Ehman directed some of his remarks to the subject of the need for radiologists to invest in leading-edge radiological research. But he also urged audience members to join in leveraging information technologies and strategies to help demonstrate the value of what radiologists do to enhance overall patient care delivery. “These innovations—whether they are machine learning, highly-focused protocols or value-focused reengineering—may allow our most powerful diagnostic tools to be used even more effectively for the benefit of patients.” What’s more, Ehman said, “Imagine a world in which our most advanced imaging technologies are widely available, modestly priced, and considered high-value, first-line diagnostic tools. Knowing that disruption and reinvention of our practice is inevitable, we must guide the process ourselves instead of having it imposed on us by others outside our discipline.”
A Cresting of the Hype Cycle?
The comments that Dr. Ehman made in his president’s address could scarcely have been imaginable ten years ago, when radiologists were under far less pressure to produce so many studies so quickly, and with such accelerating policy, regulatory, and payment demands as nowadays. Yet the reality now for radiologists is such that there is no alternative but to innovate in order to survive, and to prove their worth in a healthcare system increasingly focused on value, Rasu Shrestha, M.D. told Healthcare Informatics. Dr. Shrestha, a practicing radiologist and the chief innovation officer at the 20-plus-hospital UPMC health system based in Pittsburgh, said in an exclusive interview that the die has long been cast in the direction of radiologists’ needing to very actively prove their value not only to the purchasers and payers of healthcare, but also to the referring physicians with whom they interact on a daily basis. And that, Shrestha said, will necessarily require the investment in, and implementation of, every useful type of artificial intelligence tool and strategy relevant to the optimization of diagnosis, speed and efficiency of diagnostic imaging study development, operational efficiency, and communications with referring physicians.
As a result, Shrestha said, “When I was chair” of RSNA’s Scientific Program Committee (he spent a year as chair, with his term as chairman ending a year ago, though he remains on the committee as a member), “one of the things that I pushed for was embracing machine learning.” Still, he conceded, “I think we’re in the eye of the storm now in terms of the hype cycle. You could see the storm brewing; and we in the imaging industry had a lot to do with that, because it’s a technology whose time has come,” at a moment when the volume of data and information involved in radiological practice is exploding, and the need for speed is accelerating as never before. Radiologists will perforce need to leverage AI tools and strategies simply to stay productively and effective in practice, he believes.
Even so, Shrestha made it clear that he sees a great deal of confusion around terms and concepts in this area. He said that it’s important to distinguish between artificial intelligence, machine learning, and deep learning. He has in fact developed formal definitions for all three phenomena. According to Shrestha, “artificial intelligence is an umbrella term,” referring to “any technique that tries to mimic human intelligence, logic, and inference, using ‘if-then’ rules. Machine learning,” he has written in a presentation, “is a subset of artificial intelligence that includes abstruse statistical techniques that enable machines to improve tasks with experience That category includes deep learning,” which is “a subset of machine learning that is composed of algorithms that permit software to train itself to perform tasks, like speech and image recognition, by exposing multilayered neural networks to vast amounts of data.”
One reason that it is so important to distinguish among these terms, Shrestha said, is the unfounded fear around them. “I think the term artificial intelligence is a misnomer,” he said. “Not only is it scary—as in, it’s going to put radiologists out of a job; it’s also confusing, as in, it will ‘cure cancer.’” In fact, he said, “’AI’ should stand for ‘augmented intelligence, in that it provides a level of support to the care we provide as humans, augmented by exponential capabilities that these machines can bring to the table, while our ability to bring empathy, trust, and communication to the table, are enhanced by our actually having the time to reconnect with our patients as human beings.”
Feeling the pressure everywhere
Jonathan Messinger, M.D., a practicing neuroradiologist, the chief of radiology at South Miami Hospital, and the medical director for imaging informatics for Baptist Health South Florida, the six-hospital integrated system of which South Miami Hospital is a member, agrees with Shrestha that radiologists need to demonstrate their value to referring physicians and patients. “We need to find a way to prove our value in the arena of care; otherwise, we’re going to get passed by,” Dr. Messinger told Healthcare Informatics. Messinger, whose health system is partnering with the Burlington, Mass.-based Nuance Healthcare, agrees with Shrestha that leveraging AI tools will improve radiologists’ efficiency, while not replacing practicing physicians. “You’ve seen everyone starting to freak out over the idea of artificial intelligence potentially replacing human beings, but that’s not going to happen at all. It will create augmented intelligence. And in fact, the radiologists who don’t use AI will be left out,” as the push towards value accelerates.
Tarik Alkasab, M.D., service chief for informatics/IT and operations at Massachusetts General Hospital in Boston, and radiological clinical lead for the enterprise electronic health record at Partners Healthcare, the integrated health system of which Mass General is a member hospital, agrees. Alkasab, whose organization is also partnering with Nuance around AI tools said, “The vision of this suite of tools is designed to help the radiologist to increase the value of what we’re doing; what’s more, this suite of tools will help to provide guidance to referring physicians, helping to direct the patients onto good clinical pathways that could at least partially be driven by findings or data from imaging exams.”
Dianna Bardo, M.D., a pediatric radiologist and neuroradiologist at Phoenix Children’s Hospital, adds that there are some very practical needs that AI tools could address. “Workflow is one of the biggest challenges,” Dr. Bardo told Healthcare Informatics. “I’m the director of body MR at Phoenix Children’s, and I can tell you that there is a whole series of workflow and pre-workflow processes” that could be improved through AI-facilitated data analytics. Indeed, Bardo and her colleagues are working with the Netherlands-based Philips on some of those areas right now. “MR is particularly difficult operationally,” Dr. Bardo said, explaining some details of the MR care delivery process. “It involves more setup, is more technically challenging, and takes longer. And training a technician to do inpatient pediatric work involves a special level of complexity.” In fact, she said, “If we could do things like initiating sedation and anesthesia outside the MR suite, we could optimize the magnet time”—meaning, the amount of time that pediatric patients are actually inside magnetic resonance imaging machines. Further efficiency could allow the hospital to perform more MR studies within each day, not only optimizing patient throughput, but also opening up great scheduling flexibility for parents, who often have to take off work to facilitate their children’s imaging procedures.
What’s more, Bardo said, “With artificial intelligence, we’ll see better connectivity” around the elements of clinical decision support, image-sharing, and other core processes involved in radiological practice. And, paired with the increasing use of the cloud—Phoenix Children’s has begun a full-scale migration of its image storage to the cloud—the potential is great for real breakthroughs in efficiency and effectiveness, she noted.
Meanwhile, with vendors rushing headlong to invest in and develop artificial intelligence tools, UPMC’s Shrestha believes that it will be important for radiologists to be very actively involved in helping to guide the industry forward. “The American way tends to be to let the market determine winners and losers,” he said. “But I think it’s really important to contextualize these innovations back into patient care. There should be some direction from clinician leaders, guiding us towards some specific things, enabling, for example, the crunching of vast troves of data on a scale previously impossible.”
