If there’s a single medical specialty that is in particular flux right now because of changes in information and medical technology, it is radiology, which has always been the most technology-sensitive of all the specialties. Not only is artificial intelligence (AI) beginning to be applied both to diagnostics and to study prioritization; new developments in technology are also playing a key role in reducing disparities in health outcomes by making care more accessible in rural and remote areas, such as in the implementation of handheld ultrasound devices bringing specialized care to remote regions in India and sub-Saharan Africa.
Per all that, there are leaders in radiology who are looking at the international landscape and applying their analysis to potentialities on a global level. One of those is Mathias Goyen, Prof. Dr.med. Dr. Goyen is a professor of radiology at Hamburg University in Hamburg, Germany, and a practicing radiologist there. He is also chief medical officer for Europe, the Middle East, and Africa, for GE Healthcare, where he is focusing his analysis and research on future trends, including precision oncology and AI trends. He spoke recently with Healthcare Innovation Editor-in-Chief Mark Hagland about the trends shaping radiology globally. Below are excerpts from that interview.
What are the trends you’re looking at right now, at the broadest level?
There are several; one is the trend of using AI to transform data into actionable insights. There is an incredible amount of data in healthcare, but it’s spread across too many places, and it’s not actionable. Our customers want us to provide data that provides longitudinal insights, enabling quicker and more agile action. There’s also the trend of AI being embedded into devices, and that continues to be important. And customers [of solutions companies including GE] are saying, ‘You guys are good at Pixel AI. But please relieve us from all the distractions around us. Up to 90 minutes a day, a radiologist is doing everything but radiology, looking for cases, looking for files. So streamlining workflows becomes more and more important. How can we relieve the radiologist from the repetitive stuff that no one wants to do? That is a huge challenge.
And in that regard, we’re seeing AI advancing strongly in the area of study and workflow prioritization, correct?
Absolutely. Just to give you one example of where applying prioritization can really making a difference is around diagnosing pneumothorax, collapsed lung. That is a life-threatening condition and can truly prove deadly if not identified early. And it can take two to four hours before a radiologist is given access to an x-ray. Clinicians are looking for easy-to-read chest x-rays faster and in a more prioritized way. With an algorithm built in, the x-ray machine can alert the radiologic tech on a level of 1-10 what the chances are that a patient has a pneumothorax. The way this can work is with a so-called “traffic-light system”—green, yellow, red—and a percentage system, giving the tech, for example, a 97-percent likelihood that a particular patient as a pneumothorax. And that condition is easy to diagnose; it takes like five seconds.
As we all know, there was great fear among some radiologists several years ago that AI would be brought in to totally replace them. Instead, they’re finding that AI really is their friend.
Yes, that’s right, AI is your invisible friend. AI per se will not at all replace radiologists; but what is true is that the radiologists leveraging the power of AI will replace the radiologists who won’t use it. The job description of the radiologist of the future will change dramatically. And think of the growing field of interventional radiology; I cannot see AI taking over the work of interventional radiologists, but they use AI to guide the wire to the region of interest. I always say, would you like to sit in a plane without a pilot? No. And autopilot has not replaced the human pilot; it’s augmented the pilot’s work. Take the example of Sully Sullenberger [the airplane pilot who famously landed US Airways Flight 1549 in the Hudson River in New York City in 2009 after both of the plane’s engines were disabled by a bird strike]. Autopilot could never have landed that plane on the surface of the Hudson. The situation still required a pilot. But what AI can do is to relieve the radiologist from mundane, boring, mechanical tasks.
The unthinking tasks, in other words? So that the technology allows the radiologist to focus on the tasks that require thinking?
Yes, and the technology will give the radiologist more time to talk to the patient. It’s a great opportunity to get out of our darkrooms and get into the daylight and to take care of more sophisticate cases. And to summarize it well, AI is like an invisible friend helping the radiologist in their day-to-day work.
Technology might really come to the rescue of health systems in the United States, Germany, and all over the world, as the world copes with intensifying shortages of radiologists, correct?
Yes, staff shortages have been a top priority for leaders across healthcare, on a global level, and burnout is more widespread than ever. And rad tech shortages. In Germany, there are very few rad techs who know how to do MRI. At RSNA, I participated in a session around this topic. We know AI can reduce radiologists’ diagnostic uncertainty it also assists them in richer reports that can produce higher-confidence reports, and that in turn can reduce radiologist burnout.
And there is an ongoing shortage now of radiologists in Germany and across Europe, correct?
Yes; in fact, the only bigger shortage in Germany right now is in pathologists.
How will these trends in technology and staffing intersect with other trends in the industry?
There will be a consolidation in the market. If you look at companies like ours, developing together with our customers applications—frankly, we get approached every week by startup companies, and quite frankly, very often, and this is the irony, while every startup wants to make the life of the radiologist easier, often, they introduce more complexity. Colleagues tell me, if something involves one click more, on an app, I won’t do it, I won’t use it. So the best AI is one that is invisible, doing the magic in the background, and making the radiologist’s workflow seamless. And that is where we are headed. There will be a consolidation in the market.
Not every app will turn out to be a great app that really helps the radiologist. And it’s not like I buy three apps and I’m a smart hospital. It’s really a strategy: clinicians, IT staff, technical staff, c-suite, must be committed. You need change agents ready to move towards a new environment. So five years from now, it will be the AI-powered radiologist who is able to digest the incredible amount of data they’re dealing with. The best human brain cannot process this incredible avalanche of data. In fact, 97 percent of all medical data goes directly to the archive and is not looked at once generated. We need the AI to look at patterns in the image. It’s disease prediction: we want to predict the course of disease. Radiomics is another area that will evolve forward [with, according to the American College of Radiology, radiomics referring to the extraction of mineable data from medical imaging… (something that) has been applied within oncology to improve diagnosis, prognostication, and clinical decision support, with the goal of delivering precision medicine]. Five years from now, radiologists will work a little differently, will be empowered, and will be able to generate a report faster and better. And we’ll be able to detect cancer earlier.
What should senior healthcare IT leaders be thinking about all of this right now?
Every radiologist who is particularly smart knows that the health IT guy is my person, and I have to work closely together with him. I’m a radiologist. So the health IT guys have to help me to build a strategy, to decide on platforms. Health IT is an integral part, and every radiologist understands that. We have to leverage them, they’re our friends.
