Medicine has been awash in data since doctors started recording patients’ blood pressure, temperature and other vitals, but the advent of digital imaging and monitoring in the past several decades has hit the industry with a tsunami of information. Muuranto, engineering director for mobile digital health and wearable monitoring at GE Healthcare, has been working with his colleagues to develop hardware and software to gather more health data from patients and write sophisticated software to make better sense of it.
“Very often now when you go to a doctor, they spend half of the visit entering data into the computer,” Muuranto says. “Similarly, in hospital care, a lot of the time of the nurses and doctors goes to entering and interpreting data. We are building systems that are more intelligent with the goal of starting to reverse that trend, help them get focused on what’s important or unusual, spend more time with the patient and improve treatment.”
Muuranto is working with his team on the front part of the system: wearable wireless monitors that patients stick to their bodies like Band-Aids and that collect medical-grade data around the clock at the hospital — and possibly, in the future, at home. On Friday, he spoke at a panel discussion addressing the medical applications and implications of AI at Slush, a huge technology conference that occurs in Helsinki every fall.
Since its humble beginnings as a mixer for tech-minded students from local universities keen to compare startup ideas, the gathering has grown into Europe’s answer to America’s South by Southwest, drawing the likes of GE, Amazon, Microsoft and Facebook, as well as luminaries such as former U.S. Vice President Al Gore, and venture capital gurus Michael Moritz and Vinod Khosla, among others.
Medical technology startups had a big presence this year at Slush, and Muuranto’s panel focused on AI and how it can make healthcare more efficient and effective.
“For the AI algorithms to learn, you need a lot of data,” Muuranto told GE Reports before his panel. “That’s because today, AI is not ‘intelligent’ as we understand it. It’s just very persistent and doesn’t get tired, even if you show it millions of different examples. This is good for finding interesting correlations, but AI doesn’t yet understand the causality.”
Muuranto said he believes that the oodles of data gathered by his wireless digital monitors and analyzed by AI and machine-learning software being developed by other teams at GE Healthcare could “act as a kind of early warning system. It’s not going to take the doctor’s role to make conclusions and be final decision-maker, but it could certainly get them focused on more immediate issues. The hope is that the doctor wouldn’t have to go through so much routine data, and could instead focus on findings or patterns that are out of the ordinary.”
For example, the flood of information could help doctors build “digital twins” of patients: virtual representations built over time from data gathered by Muuranto’s sensors and other data sources. Such digital twins are already common in the industry — GE service teams have built virtual doppelgangers of thousands of jet engines, gas turbines and other machines. But Muuranto cautions against taking the analogy too far.
“We can engineer a jet engine from scratch and as a result assess from data how it’s performing or how its function might deteriorate,” he says. “But we cannot engineer a human being. We are nowhere near fully knowing how the body works. As a result, AI systems in healthcare are often a black box. They can find interesting insights but they don’t know why.”
Hanna Viertio-Oja, Muuranto’s colleague at GE Healthcare in Helsinki, is working on opening the black box and helping doctors understand how AI comes to its conclusions.
She and her Helsinki-based team, together with colleagues in Israel and the United States, are working to develop a tool that will help clinicians have a better understanding of what is happening to their patients. The Helsinki team is responsible for developing predictive AI that can help doctors determine which patients might take a turn for the worse.
Viertio-Oja, a 26-year GE veteran with a doctorate in technology, says she and her team have learned from clinicians that “if we only provide them an algorithm which works like a black box and says that, ‘Hey, now this patient has a high risk of deterioration,’ but if it doesn't explain why it has come to this conclusion, it is not so easy for them to trust the algorithm. So we are working very hard on the explainability aspect, and we are developing a method to provide them information regarding why a certain conclusion has been made.”
Viertio-Oja says her team is using big data to pop the black box open and shed light on its internal reasoning. They are frequently talking to physicians and clinicians and comparing notes. “As result, our algorithms not only say, ‘Hey, there’s a 75% chance a patient will deteriorate,’ but it would also say that this is because his or her blood oxygenation has been increasing and because respiration is at this higher value.”
Going back to Muuranto’s weather metaphor, Viertio-Oja’s team’s technology could not only tell physicians to bring an umbrella, but explain why.
Muuranto and Viertio-Oja are also taking advantage of the relationships GE Healthcare in Helsinki has developed with European startups. In 2014, for example, the company leased a portion of a floor to Health Innovation Village, a startup incubator focusing on ideas that could transform the medical industry.
One company, the AI firm Top Data Science, now has 20 employees, and a Japanese investor. It has been working closely with Muuranto and Viertio-Oja’s teams to solve problems like spotting which patients in the intensive care unit are likely to deteriorate.
“Co-creation is the most important aspect of this collaboration,” says Timo Heikkinen, co-founder of Top Data Science, who appeared with Muuranto on the AI panel. “We are small and nimble and can move fast. We also bring skills and insights that big companies don’t always have, like looking at problems from a broad perspective, rather than focusing on narrow issues. But GE has deep engineering knowledge and partnerships with hospitals and other customers that can help us validate a solution quickly, get a result and learn from it. We know fast whether we are on the right path or whether we need to pivot.”
All three believe medical AI and data will transform healthcare and help doctors do their work. “Fifty years ago we had accountants and their work was largely manual,” says Heikkinen. “Today, we still have accountants, but many of the tedious parts of the job have been automated so they can focus on the most important tasks where they add the most value. I believe that AI will do the same for healthcare.”
But will medical AI migrate to the home like, say, GPS — originally used by the military — and make everyone obsess about their blood oxygenation and respiration levels? “In Finnish we have a saying that knowledge increases the pain,” says Muuranto. “But if you are going to be able to channel those insights and change your habits to decrease your chance of developing a certain condition, that’s an upside.”
This report discusses technology in development that represents ongoing research and development efforts. These technologies are not products and may never become products. Not for sale.