Q&A: GE Healthcare’s Chief Medical Officer for Europe on Which Tech Is Helping Most in COVID-19 Response

Mathias Goyen describes how the pandemic is altering the medical-tech landscape

8 min read

Illustration of Mathias Goyen
Mathias Goyen
Illustration: Jacob Thomas

When the last global pandemic broke out, in 1918, it ravaged a population with essentially no technological countermeasures. There were no diagnostic tests, no mechanical ventilators, and no antiviral or widely available anti-inflammatory medications other than aspirin. The first inactivated-virus vaccines would not become available until 1936. An estimated 50 million people died.

For the current outbreak, a best-case scenario could limit fatalities to 1.3 million [PDF], according to projections by Imperial College London. That in a world with 7.8 billion people—more than four times as many as in 1918. Many factors will lessen mortality this time, chief among them better, more consistent implementation of social-distancing measures. But technology will also be a primary bulwark. Enormous sums are being spent to ramp up testing, diagnosis, modeling, treatment, vaccination, and other tech-based responses.

Mathias Goyen is in the thick of these efforts. As GE Healthcare’s top medical officer in Europe, he is a key figure in a US $20-billion-a-year medical-technology colossus. Based in Germany, he has been in the vanguard of GE’s efforts to confront the disease on the continent with the most cases and fatalities caused by COVID-19.

Goyen, a former radiologist and academic, joined GE in 2012 as a research executive and was promoted to Chief Medical Officer, Europe, three years ago. He is known for evangelizing the use of artificial intelligence and other advanced technologies to help manage the vast data streams and challenging computational chores on which so much of health care now depends. He’ll imagine the capabilities of an AI-based system that considers both MRI and genotype data to deliver a personalized diagnosis, while also extolling the virtues of AI-based software for helping hospitals deal with such routine chores as scheduling, staffing, and allocation of resources.

IEEE Spectrum spoke to Goyen about how the pandemic is reshaping medical-tech priorities and research. He explained how the pandemic is accelerating the application of artificial intelligence to a broad variety of challenges in health care, including a very current and remarkable one: A sophisticated neural network, using computed tomography (CT) data, could soon be capable of distinguishing between the pneumonia caused by the novel coronavirus, SARS-CoV-2, and that precipitated by some other cause, such as bacteria or an intense allergic reaction.

This interview has been edited and condensed for clarity.

Spectrum: Which technologies are you seeing particularly high demand for?

Goyen: We see an explosion in the demand for medical equipment. Starting with normal X-rays to CT, of course, as the imaging modality to assess the lung. But also ultrasound. Probably ultrasound is not a natural candidate, you might think, to assess the lung. But here, dealing with the pandemic, all of a sudden ultrasound really has a major role. You can use it right in the ICU; you don’t have to bring the patient to the radiology department. It’s inexpensive; it’s available; it’s a bedside test. There’s a huge, huge demand. [Editor’s note: A recent Spectrum article described surging demand for ultrasound systems.] It’s safe to say that COVID-19 has helped accelerate the adoption of digital health technologies.

Spectrum: It seems like there are some particular technologies that are really rising to the top and will probably benefit substantially when all this is over.

Goyen: Exactly. I would say the overriding theme is AI, artificial intelligence. Because if you look at it, AI is playing a part in each stage of the COVID pandemic. From predicting the spread to also powering tools that can really help humans in the hospital to limit human interaction. I mean, robots can help hospital personnel or can disinfect rooms. They can provide telehealth. And electronic health records—it’s very important in the ICU to have access to those.

It’s not like nothing was there before. Of course it was there. But now you see that hospitals that had prepared well—pre-COVID—now really have advantages because they have an electronic patient record that people now need. With ICUs full of patients, you need access to images and records like that. And so I would say that AI plays a significant role in all different aspects, from prediction to diagnosing COVID. Just think of implementing AI in CT: That helps the radiologist in diagnosing those cases, all the way to therapy.

Spectrum: What are some areas where artificial intelligence is coming to the fore?

Goyen: You can implement AI at different levels in the health system. At one level, AI is built right into the machine, into our equipment: the X-ray machine, the ultrasound device or a CT. I call this the individual level. Then there is the departmental level; I call it non-pixel AI. To streamline workflows, for example in radiology departments. To deal with no-shows. To predict if people will really come to an exam or not. And then you can extend it to a network level. I’m sure you’ve heard about our [hospital] command centers, where we use predictive analytics to really manage patient flow in entire hospitals…where you have full transparency. Where are the empty, clean beds that you can use to really admit a new patient?

Spectrum: I would imagine that these command center and scheduling programs are being severely tested in hospitals that are in, or close to, hot zones with enormous caseloads. Are these systems able to keep up with that? Because I imagine they weren’t designed for some of the caseloads that are being seen.

Goyen: These command centers were also built to serve as incident command centers in the event of, let’s say, a mass-casualty disaster. Now with the COVID-19 pandemic, I would say this is the right solution. To help with the pandemic, we introduced new tiles [for the command centers], different tiles that fill different needs the hospital has. One tile is the Infectious Disease Tile, and the other tile is the Critical Resources Tile….The Infectious Disease Tile helps to ensure that COVID-19 patients are matched with critical, and usually scarce, resources. The Critical Resources Tile helps manage ICUs—for example, how to find ventilators across the hospitals in a region. These two tiles can be set up remotely within about two weeks.

COMMAND CENTER VIEW: A new \u201ctile,\u201d or application, by GE Healthcare uses artificial intelligence algorithms to help hospitals keep track of the big caseloads associated with COVID-19 outbreaks.COMMAND CENTER VIEW: A new “tile,” or application, by GE Healthcare uses artificial intelligence algorithms to help hospitals keep track of the big caseloads associated with COVID-19 outbreaks.Image: GE Healthcare

Spectrum: One of the things that’s happened that’s very encouraging is that some large companies have chosen to work together. GE is working with Ford on ventilators in the United States. Can you tell me, are things like that happening in Europe as well? Do you feel that this is a model for progress?

Goyen: As you can imagine we’re exploring different options working with companies, governments, and other stakeholders to see how we can increase the production of, for example, ventilators. At this point I cannot say more, but we are exploring different opportunities there.

Spectrum: Are there any other technologies that you feel have promise? Ones that are perhaps unusual or overlooked?

Goyen: One technology I really like, we call it CT in a Box. Computed tomography in a box. We first delivered it in China. It can be used in pop-up hospitals. They’ve built hospitals in a week in China. And you see in the U.S. they also quickly built some extra space, and created some extra beds. CT in a Box is a very easy-to-install standard CT that we use. But it’s a special setup, I would say, that we can deliver in a short amount of time and set up quickly and that can be used in the ICU or directly attached to this newly built ICU, be it in a big building or whatever, like a field hospital…. We just deployed it in Paris, in France last weekend. [Editor’s note: the weekend of 11-12 April]. This is something that I personally like. We came up with a new product to respond to the increased demand of the customers in these unusual situations.

To help with COVID-19 diagnoses, a CT scanner specially designed for rapid installation was delivered to the Henri-Mondor AP-HP University Hospital in Cr\u00e9teil (Val-de-Marne), southeast of Paris. The system, built by GE Healthcare, was installed in under two weeks in early April\u2014a fraction of the time normally needed to put in such a complex unit.COMPUTED TOMOGRAPHY IN A BOX: To help with COVID-19 diagnoses, a CT scanner specially designed for rapid installation was delivered to the Henri-Mondor AP-HP University Hospital in Créteil (Val-de-Marne), southeast of Paris. The system, built by GE Healthcare, was installed in under two weeks in early April—a fraction of the time normally needed to put in such a complex unit.Photo: GE Healthcare

Spectrum: Can you say briefly what it is about CT that makes it invaluable and necessary? In other words, if you could do everything with ultrasound then you wouldn’t need your CT in a Box. But clearly CT does have a role to play.

Goyen: It’s because of its high sensitivity. You can say that CT is currently the imaging method of choice really to initially diagnose and also to monitor, meaning doing a follow-up to monitor patients with COVID-19…. Ideally, it would be nice to have an initial CT in all COVID patients, like an initial CT to see the lung involvement. To see the so-called ground-glass opacities. To see how badly the lung is damaged. And then, again ideally, it would be nice to have a follow-up CT every third day, but it’s not very practical. So now ultrasound comes in and you could do the follow-up with ultrasound.

Spectrum: You mentioned ground-glass opacities, a sign of fluid or other dangerous abnormalities in the lungs. These are things that are only visible with CT, correct?

Goyen: Yes. But ground-glass opacities can mean anything, right? One of the differential diagnoses is COVID-19. But it could also be just a normal bacterial pneumonia. It could be an allergic reaction. It could be many things. So a CT is very good: It has a very high sensitivity to detect ground-glass opacities. What everyone is working on now is applications—an app that also shows a high specificity and can really differentiate the ground-glass opacity in a non-COVID patient versus the ground-glass opacity in a COVID patient.

Spectrum: That’s something that might be possible soon?

Goyen: If you look at the literature, a lot of people claim they have the killer app. But if you look closer, they all say it’s high sensitivity only. I wouldn’t say it’s rocket science; it’s doable. To my knowledge no researcher, or no research group, has shown a high specificity with regard to COVID based on lung CTs. I wouldn’t say it’s impossible, and this is what people are working on right now.

Spectrum: And again, you’re talking about differentiating, say, between the ground-glass opacity of a COVID-19 viral pneumonia patient versus the ground-glass opacity of a pneumonia patient where the pneumonia wasn’t caused by COVID?

Goyen: Yes, exactly. And they can look pretty much exactly the same, at least to the human eye. You can only report as a radiologist that these are ground-glass opacities in the basal segment, let’s say for example, of the right lung. And as a radiologist only looking at the image, you cannot tell, is this caused by a virus? Is this caused by a bacteria? You don’t know.

Spectrum: But people are thinking it might be possible?

Goyen: People are thinking it might be possible with an algorithm using AI to see patterns in the image that are not visible to the human eye, not even to the best human eye.

Spectrum: That’s super exciting.

Goyen: And then if you have a cohort of a thousand patients for whom you have a follow-up, and you have different time points, so you know what is really COVID, and what the outcome was: Have they recovered? Have they died? And if you have this cohort, you can probably develop an algorithm that looks at the initial image and finds some patterns in the image not visible to the human eye. And this of course would be awesome, right? Being able to differentiate: This is a ground-glass opacity caused by coronavirus versus this is a ground-glass opacity caused by a hyper-allergic reaction or by a bacteria.

The Conversation (0)