Photo courtesy of Philips
Editor's note: This article has been updated with context from Philips regarding the number of helium-free scanners installed in the U.S.
Shez Partovi, chief innovation and strategy officer at Philips, spoke with MobiHealthNews during the JPM Healthcare Conference about the company’s advancements in pathology and radiology and its perspective on responsible AI development.
MobiHealthNews: What is Philips doing here at JPM?
Shez Partovi: We did an announcement late last year, which is around our BlueSeal magnet.
So, the problem with MRI scanners is that generally they need a lot of helium. The helium is used to cool it down, and this helium, over the lifetime of the scanner, you have to keep replenishing the helium, and that means that the scanner is actually very expensive. Helium is a rare compound. It is not sustainable to the environment and because you have helium, it needs, by regulation, this little pipe that in case there's a problem with the helium, you press a magic button and all the helium goes out of the top.
So what we did is we invented a way in which you can have an MRI scanner in which you do not need to refill the helium. It's called a BlueSeal MRI, and it is helium free.
You can now even put it in a truck, and you can take it to communities where there is no MRI. Cost of construction goes down. You can put it in ambulatory sites anywhere, because building an MRI scanner in a particular location is hard if you need all that equipment for piping, and so on, versus just a sealed magnet. It needs no refills.
Then we did a few other things. So, not just the sustainability, helium-free angle, but we also embedded a significant amount of AI, and it does a few things. One, there is a technology we put in there called SmartSpeed. That means the scans occur three times faster.
We also embedded a whole lot of AI for diagnostic purposes and, in fact, even partnered with other companies to have AI directly on the scanner.
We invented the helium-free scanner. There are other competitors that have announced it. They haven't released it. We're at scale now. We have 2,000 installs in the U.S. We mentioned it about four years ago, five years ago, and it's been one that we've been continuing to improve and now with all the embedded AI. So that's one area that's really exciting for us, and that we are focusing on at JPM.
MHN: What does Philips have planned for 2025?
Partovi: So, let's switch gears, go from MRI scanners to digital pathology. What pathologists do is take a biopsy, put in the glass slide, then a pathologist looks at that under a microscope. It is the last frontier of medicine that is not digitized. So, the reason Philips led the way in pioneering digital pathology is actually because we invented the compact disc.
So, the compact disc has a laser beam that reads the grooves of the compact disc. The grooves are micron size. Pathology slides were a problem because cells are micron in size, and, for years, no one knew how to scan using just visible light.
At one point, somebody was sitting in a lab at Philips, who was on the compact disc team scanning with the laser, and heard about the problem of scanning cells, and said, "Well, the laser beam that we use for the compact disc, we could use for scanning slides." So, a number of years ago, that team moved the laser beam from scanning compact discs, moving it to a different department, started a new lab, a computational pathology lab, and started to scan slides.
So, we actually pioneered creating digital images from pathology, using the same laser beam that was in the compact disc. That was Philips technology. Our heritage of working in light goes a long way.
Now, about 20% of pathology labs globally are now digital, and we lead the way in doing that. And so, with the digitization of pathology slides, you have the perfect syzygy of textual data from electronic medical records, medical imaging from MRI, CT, ultrasound, angio, and now pathology. And so, the opportunity for AI-based understanding of disease conditions for cancer prediction, cancer detection and cancer follow-up now becomes a lot.
So, with the digitization of pathology, what we're now doing in 2025 is we are bringing our pathology solutions and PACS, which is what the radiologists' use to look at the images. We are integrating our digital pathology systems and our radiology PACs systems, and so, now you have this cockpit where you have both the medical imaging and the pathology imaging integrated for the same patient. Also, we and our customers, like NYU Langone, who is one of our collaborators for pathology, are building AI models for disease prediction and disease direction.
In 2025, you will see more and more announcements from us, from what we would call integrated diagnostics, between radiology and pathology, because at the end, the heart of medicine for diagnostic purposes is radiology and pathology.
So, that's in the area of automation, because we use genAI to summarize the prior findings, and use automation to point to all the prior pathologies. It's because today we believe that you still have the clinician in the loop. Yeah, you're really empowering the clinician to take away all the trivia that they need done so that they can just focus on reading the films.
MHN: Do you see that in the future maybe the physician won't be necessary?
Partovi: No. And not to be tongue in cheek about it, but there are a couple of reasons for that. So, first, practical. Given that baby boomers are entering the age of multiple chronic conditions, given that Millennials are entering the age where they are having children and increasing healthcare needs in a different way. So, wellness care, if you will, and sickness care, increasing the demand on the health system. Then you have shortage of staff, and burnout.
So, we're a long way away from saying that we don't need physicians. In fact, if AI comes in, it might actually fill the gap of the shortage in communities that don't have it. So, we're a long way away from saying and we will not need physicians. I don't know, actually, if we'll ever be there.
But point one is, practically speaking, for a decade to come, there's such shortage that, if the tools give time to physicians to move from administrative trigger work to patient care, we will just barely catch up with the demand of the pyramid of society that's moving to baby boomers and the pyramid that's coming. So, we might just make it if we use AI to give time to physicians.
So, the statement that we won't need them seems even contextually to where society is today, it does not ring true to me. If we were in a different time where the population pyramid looked different, I might say maybe. But where the population pyramid is, it is decades before we are at a point where we feel like, "Oh, well, we might not need them." So, that's a practical point one.
Practical point two is our point of view, at least to Philips, is that AI is a tool set for physicians and nurses so they can do their job better. It becomes another tool in the tool belt. So that is a point of view that we have, and a point of view that we take in how we build algorithms and how we build AI tools is, how do we give a physician's time back? How do we give them better analytics and insights so they can make better decisions just in time? How to improve access to care, so that in outlying communities there's actually access to those technologies. These are three things, if you will, our guiding principles for AI usage.
MHN: As someone who is very embedded in the innovation space, what makes you nervous?
Partovi: So, AI went through two winters. So, if you look at the past decade, there are two eras where we went through winters where it was like, "Oh, we can do everything," and then "Oh ..." Then it is like, "We can do everything," and then "Oh ..."
What I worry is that we overpromise in delivery again and actually end up retreating, because people say, "Oh ..." So, if I had a magic wand, I'd make it slow and steady. It would be the way to go, slow and steady, as opposed to this hype of it will not need doctors, it will solve all problems, it will cure all cancer, and then the third winter of, "Well, there we go again. It didn't do it."
So, to me, I really do think that there's an opportunity for a tectonic shift in the industry, and where you can really improve access to care with reduced cost of care. It appears as though this iteration of AI and generative AI has the catalytic properties needed to make that happen. And so, what I would say worries me is wild and crazy claims.
My greatest fear, categorically, is those that will come out. And, you know, you might even see them at JPM here and in other places – categorical claims that just seem outlandish. I wouldn't want that to be the metric by which AI progress is measured. I'd rather see it slow and steady, scientifically proven, validated claims, outcomes and impact, and then go next.
Also, I am asked often about regulatory. I actually don't think there's a problem. You know, you need a regulatory framework to have slow and steady. In the absence of a regulatory framework is when you can make wild and crazy claims, and do things that might, God forbid, hurt people and then take a pullback. So, I actually think the regulatory-framework developing is critical, so that we can do slow and steady. Everybody just wants a magic wand to solve it, but that's just not reality.