MaxQ AI chairman & CEO Gene Saragnese talked to reporter Sarah Faulkner about the company’s software that uses AI to interpret medical images and patient data. Want to see these interviews live? Join us at DeviceTalks.
Sarah: Digital health, artificial intelligence and machine learning – these terms represent a lot of the buzz emerging from the medtech industry this year. And as is often the case with buzz-worthy technologies, there are plenty of companies touting the next big innovation. I think it’s safe to say that AI has not yet fundamentally disrupted the healthcare industry – at least in the ways that many predicted it would – but MaxQ AI CEO Gene Saragnese pointed out to me in our interview that in order for doctors and patients to reap the benefits of AI-enabled tech, it has to be applied to the right problems.
Gene: The technology at MaxQ is based upon machine learning, machine vision technology, and machine learning technologies have had a huge impact on virtually every industry, except for healthcare and applying that technology to healthcare has it’s unique problems – data is less available. Amazon has millions of customers and millions of pieces of data, so that part is easy. In healthcare, it’s not so easy. The data is sparse, it’s corrupt. You have to think differently when you apply this technology to healthcare but I am confident that applying it correctly to the right problems will have a tremendous effect.
Sarah: At MaxQ, Gene and his team have honed in on a very specific set of problems that bring patients to the emergency room. The first several software products emerging from MaxQ are focused on addressing intracranial hemorrhage and stroke detection.
Gene: Time-sensitive, life-threatening conditions are places where we often don’t either act quick enough or have the expertise at the bedside to make good crisp decisions and it turns out, stroke and trauma are two areas where you make the right decisions, patients get better. If you make the wrong decisions, they may not.
Sarah: The company’s software is designed to reduce misdiagnosis and healthcare costs by helping doctors to make faster and more accurate decisions when diagnosing things like stroke and brain trauma. When building its products, the company follows a few major design considerations, according to Gene.
Gene: Well we want to augment the physicians with AI, machine learning, and essentially, improve the expertise or project expertise and augment those physicians so they can make good decisions at the bedside. We must, must, must integrate in a very very seamless manner. Our goal is zero clicks, not reduce the number of clicks but fundamentally, we process all of our information in the background. When a physician sits down to read the images or diagnose the patients, that information is available, no buttons, no clicks, no nothing.
Gene: Our overall goal is to redefine the diagnostic quality if you will, and accuracy and speed and that’s the overall business that we’re in. Imaging is important to us but we consider ourselves a diagnostic company, not just an imaging company. Lastly, we must solve an economic problem. We must have an economic impact associated with the technology and outcomes. Why? Because again, we can’t afford to give care to the people that are out there today. We got to create capacity. We’re still making errors on the order of 30% in the diagnostic space, which creates a lot of waste, both social as well as economic waste. So as I said, the problems we’re focusing on must have an economic impact and we must impact those economics through quality and outcomes.
Sarah: One thing that really stood out to me in this interview with Gene is his use of the phrase “augment physicians’ – that his company’s tech was built to augment physicians, not replace them. This was so interesting to me because I think a lot of folks have the impression that eventually people who have jobs like radiologists will be replaced by AI-enabled tech. So I asked Gene: does he envision a future where that could be reality?
Gene: No, I don’t believe in the concept of replacement. I believe that the tools work together with physicians. The word, augmentation, is designed to really embrace this concept of helping and aiding a physician, allowing a community hospital physician to make decisions that perhaps a neuro-radiologist would. It is augmentation.
Gene: The other thing I really want to point out in particular to radiologists of the world, that this actually represents, I think, an opportunity. I do believe that the radiology community will evolve over time and these tools have become a very natural part of who they are and how they work. Radiologists have been incredibly forward-thinking in terms of technology. You think about CT machines as being high tech, yeah but informatics in hospitals really have their origins in radiology. They’re the first ones that bring, really bring informatics to the bedside. I look at this as an opportunity and I think the discipline is going to evolve over time from reading images to interpreting the full nature and diagnosing patients. So, I think it’s an evolution and it’s a shifting of skills and I think that this will allow us to reach more patients, the concept of augmenting physicians will allow us to reach more patients and I do think it will encourage and create this evolutionary path towards a higher level of diagnostics.
Sarah: And while some are wondering if AI will eventually replace doctors, others are wondering if it will have any impact at all. And that’s likely thanks to a number of products on the market that haven’t exactly lived up to all they promised to be. Here’s what Gene had to say when I asked him how he separates the hype from legitimate innovations.
Gene: Well I really love this question and it’s one of my deeper concerns. AI, machine learning is a tool. It’s a remarkable tool. It’s had a huge impact, as I said earlier, in other industries. That having been said, in healthcare, it still comes down to, what do you do with it? And I see people creating tools using AI that quite frankly, I shipped as product 15 years ago. To simply use AI to solve a problem that’s already been solved, makes really no sense. How do you apply it to new problems? And I also think about the language of AI and unfortunately, we’re painting a lot of different… I’ll use the term applications if you will and I’ll try and define that, with the same brush.
Gene: The fact that it’s AI, you paint it with the same color and think it’s really the same thing, it’s not. What do I mean by that? First of all, you can apply AI to a number of different tools. Let me try and define a hierarchy. It’s my hierarchy, people can argue with it however they want and that’s fine but let me give you a sense of the way I think about this. At the lowest level of value and I emphasize value because there still is value, it’s just at the bottom of it, is workflow. Can I make people’s workflow better? I think that’s really important but as I define the rest of it, I hope you’ll sort of recognize that that’s probably at the bottom. The second piece of it would be to automate sort of existing procedures that AI can make faster, better.
Gene: It’s a little bit more workflow but it’s more about efficiency and productivity. But as you move up, then there’s more important things to do to help triage and prioritize cases, based upon actual understanding of the content, is important and that requires more sophisticated algorithms that can get patients treated sooner, faster, and hopefully more accurately because we’re triaging and getting the right eyes on the case and that person. The next higher level is to start pointing at things very specifically that a physician should consider, as in look here, do you see what I see? And providing second reads and assistance. When you go above that, then you start getting into a diagnostic realm, such as diagnostic rule-outs that say, hey, this particular indication is not here or a diagnostic quality interpretation. As you go up from there, there’s this concept of prognosis.
Gene: For example, let’s say is something going to get worse or better as in a symptom or a particular finding going to get worse or better. Above that is the prognosis of the patient and the full diagnosis of the patient. Above that is actually trying to match treatment pathway with that patient very specifically. We’re not anywhere near the top of that pyramid or the top of that stack in terms of what we’re doing today and my intent is to work up that vertical in creating value. What worries me deeply is that everybody thinks about AI as AI. It’s not and for whoever is listening or the audience who knows what C++ is, it’s just a language, it’s a software language. Who cares?
Gene: Nobody sells or presents for value based upon C++. Nobody cares. In time, as important as the technology is, people shouldn’t care. What you do with it is really what matters and because it’s so new in healthcare, I worry a little bit that we get caught up in the buzz of, it’s AI, when we really should be thinking about, what does it do? How reliably does it do it? And is it really solving a really hard problem that needs to be solved?
Gene: I sat across the table for literally 20 years from a major payer-provider and I would brief them on technology that was coming for 20 years.
Gene: And in the beginning, it was all about exciting new technologies that were coming and about, I would say, four or five years ago, I had the same conversation and it had been changing and changing and changing and then finally, the chair of that committee, who I’ve known for again, over 20 years, said to me, what I’m being asked to do is to treat 10% more patients with 10% less cost. It wasn’t about technology anymore. It wasn’t about what was new and exciting. It was about how do you solve the problem and the fundamental problem was, she was challenged to treat 10% more patients with 10% less resources. That’s a very, very simple and straightforward statement and that’s what healthcare is about today. You have to look at this holistically, you must, and that’s at the heart of what we’re doing here at MaxQ.
Sarah: Getting back to the company’s product portfolio, it’s been a busy couple of years for MaxQ. The team won breakthrough status last year for its intracranial hemorrhage detection software and has since secured a number of CE Mark approvals and FDA clearances for various products.
Gene: For any single algorithm and in our case, let’s just use intracranial hemorrhage detection as an algorithm, we’ll have three applications. The first application is, prioritization, which operates in the background, looks, listens, finds something, raises it’s hand and alerts the workflow that, hey, something is happening here. The second application is an application which really takes a look at the images and tells you where you should go look for findings, that’s the second level.
Gene: I’m going up this hierarchy I described earlier and then the next level up in the hierarchy is really looking at a true diagnostic test and the breakthrough product is a diagnostic rule out for the presence of intracranial hemorrhage and the claim that we’re making is that you can move a patient down a non-hemorrhagic pathway if we say there’s nothing present and it’s designed to help physicians, again, to make those crisp difficult decisions bedside in the area of stroke.
Sarah: Rather than modeling medtech’s traditional plan to get to market – which generally involves building a global sales force, for instance – Gene and the team at MaxQ are relying on their industry relationships.
Gene: Our strategy for commercial reach has really been focused on leveraging large OEM partners. Healthcare requires a lot of feet on the street, a lot of salespeople, and it’s difficult to build those sales forces. We have been very successful in developing relationships with GE and IBM, Samsung, TeraRecon, and those folks are taking our products to market. Our goal is to make those sales forces successful and provide the support there but we get this opportunity to leverage hundreds of salespeople with those partners and that’s the approach that we’ve taken. Now, what’s wonderful about that, as soon as we get regulatory approval in another location such as China or Japan, there’s immediately a sales force. It’s not about ramping up. They’re already there. I like that model very much.
Sarah: When asked about the coming months, Gene told me that the market can expect to see the release of additional applications in the intracranial hemorrhage space, but he stopped short of giving me any specifics.
Gene: Know that we’re focused on the area of time-sensitive, life-threatening situations that have an economic impact and an opportunity to help physicians make better decisions. So that’s what we’ll continue to do and so that’s sort of the near-term road map. The longer-term road map, hey, we have to have some secrets.