AI in Diagnostics: Where are the Real Opportunities for Transformation? | LSI Europe '24

Andrew Pieprzyk 00:04
Good afternoon. My name is Andrew Pieprzyk. I'm the Vice President of Strategic Development for Hologic. I look after our diagnostics business internationally from a geographic perspective. I'm super excited to be here today with this group that I've gotten to know over the last week and a half. Very passionate about AI and the opportunities for transformation.

Pau Rodriguez 00:26
One of the things we've discussed is we are all aligned and fully believe in the opportunity for transformation, but we're going to take a step down from a pragmatic standpoint into some areas that are of interest on how to get from where we are today to really enable that transformation, because AI and diagnostics is a huge topic, and across the panel, we range from colorectal cancer to stroke to mental health,

Dror Zur 00:56
all different types of AI and all different sorts of challenges. I was speaking with Joe Mullins the other day, and navigating AI in the marketplace is like navigating in the Bermuda Triangle. For those of you that know what it is, you know that it's there. You don't really know what's inside, but you have to go through, and on the other side is that transformation. So that's really what we're going to focus on today. And there's a few themes that we have teased out that we thought you guys would appreciate. One is around credibility and transparency, and this is both to the healthcare professional and the patient, especially in the world of data privacy.

Andrew Pieprzyk 01:31
The trade-offs in terms of your product scope and profile. You might have a discrete application of AI, but that’s required or only works within a much larger care delivery ecosystem. So how do you work through all of those nuances, and that feeds into really what's your revenue model as an innovator, as a business leader? How do you take your discrete operation or your AI model, if it's part of a bigger ecosystem, and think about revenue generation short term versus value creation long term, and navigating a lot of those aspects.

Andrew Pieprzyk 02:07
And I'll have some, from a strategic standpoint, some lenses there of what we look at. And then we'll invite Dr. Gokce Gun, from a venture capital perspective, to do that. And ultimately, as I've gotten to know this group, this is going to be just a tremendous dialogue up here, because we've all agreed that we can probably have an eight-hour session on the Bermuda Triangle, depending on how you define those corners. So super excited for our discussion today. Team,

Andrew Pieprzyk 02:35
so with that, guys, why don't we start a little bit on

Andrew Pieprzyk 02:39
what we've described as Doctor versus bot, right, credibility and transparency. And I know Dr. Zur, you and I talked a lot about this, so maybe you can give some perspective on, you know, what is our responsibility to doctors and patients?

Dror Zur 03:04
Yeah, I think so. First of all, I'm Dror Zur. I'm the CEO of Magentiq Eye. We have an automatic polyp detection system for colonoscopy, which works in real time.

Dror Zur 03:22
I think one of the main things in AI is the dataset. A lot depends on the dataset itself. And it's important, you know, as our responsibility from the industry, towards the doctor and the patients and the regulators

Dror Zur 03:34
to be transparent with regard to that, to show, like, on what type of data the AI model has been trained and tested

Dror Zur 03:41
that we took into consideration all the aspects of the application that we are developing, and, yeah, and to show that, and if there are limitations, to also present them, you know, so everybody can know this AI model will be very good in this, but maybe you have to be careful or not to use it for that and stuff like that. And I think that brings up. And for those of you that were here yesterday, I was on the women's health panel, and we talked a lot about health disparities, right? And your initial data selection, if used outside of your use case, could lead to, you know, generalization, exacerbation, or false alarms or false readings,

Dror Zur 04:17
you know? So I think there's an opportunity in that transformation, and I'd love to hear everyone's thoughts on that, on how companies, depending on their application, choose that initial dataset, communicate that, but also build that to be more inclusive in the future. And that might be a roadmap thing. So, you know, Pau, especially David, and mental health, you know, how have you thought about that?

David Zakariaie 04:52
Yeah, that's a great question. So I think there's a couple of different questions built into that. So I'll start backwards. And I am the CEO and founder of Senseye. We're a startup based in Austin, Texas, essentially working on building the first diagnostics for behavioral health. I think, candidly, I think the inclusive question is actually relatively straightforward. Like, I don't think it's rocket science.

David Zakariaie 05:14
Like, we know all the different. Like, when we're looking at a certain indication, we know all the different demographics that it affects. And I think building a dataset that's inclusive is as simple as just going and collecting data from a very diverse population and being really intentional about it. Like, I'm not trying to discount the difficulties that may come with it. In certain types of disease areas, there are difficulties to try and build a very diverse dataset. But I think, like, the answer is just very simply that at least, like, in our experience with the FDA, has been working with them, they have started to take a much more, like, they like to talk about things that are recommendations but not requirements, but really it's a requirement. But they've kind of, like, at least with us, around conditions such as PTSD, they’ve taken a very intentional approach of like, what does the market research kind of say? Is the population, the real world that actually, you know, like, is actually affected by these particular disorders, and then what parts of the population do we suspect suffers from these disorders, but it's, you know, it's severely underdiagnosed, and then making sure that we actually are intentionally enrolling those folks, you know,

David Zakariaie 06:31
into our trials, both as we're building a dataset and training on it, and then also from the validation

David Zakariaie 06:39
perspective of it. And for

David Zakariaie 06:41
in our space, in terms of the kind of, you know, AI, I think it's a different dynamic from spaces such as oncology, because we're not trying to augment or replace an existing process. Like, there's currently no diagnostics in mental health. It's all kind of, it's all based on these very pen and paper approaches. If you ask a patient a couple of questions and based on their answers, you come to some sort of a diagnosis, and the majority of the time it's wrong or, you know, you get a certain piece of it. So you're getting, you know, you diagnose them correctly with anxiety or depression, but most mental health patients have anxiety or depression comorbid with whatever the actual issue is. And so you kind of, you know, are missing what's actually going on. And so

David Zakariaie 07:29
for us, it's been really important to kind of penetrate a reference standard and actually, like, find these different silver standards in the field that we're able to kind of put together to form a proper ground truth. And then in a lot of our communications with the patients as well as the clinicians, kind of make it explicitly clear that we're not trying to replace them with AI. We're just trying to, you know, for once, build a very powerful diagnostic for the space, put it in their hands. And then, if they're able to more quickly and accurately figure out what a patient is actually suffering from,

David Zakariaie 08:07
it will then allow them to kind of figure out the treatment side of things, which is where they're really effective in a much more efficient manner.

Gokce Gun 08:17
No, actually, I totally agree with both of you, and something I would add on that is the importance of that post-market follow-up, right? So you can define this dataset that is representative. I mean, it's never representative enough. I mean, you would need huge amounts of data to validate that, but I think that post-market follow-up is very, very useful to actually see how this is evolving and try to adapt in that if that's the case. So that's one point. On top of that, the AI typically for us is a product to try a stroke from a plain CT without contrast. And the idea is that this is a triage tool that needs to be combined with clinical exploration, right? So each center may have different criteria on how to combine both things. So we as companies, we need to focus on the AI, but also on how do we deploy our technology and how is it of best use for the physicians?

Gokce Gun 09:44
I agree, and I would like to add maybe something from about the beginning, and then for later on, how we cope with the variety in the data. So for example, in our case, in colonoscopy, we knew quite from the beginning, based on studies that were done and also from talks with the doctors, that colonoscopy is being done quite similarly all over the world, and there's no real dependency on geographics or demography.

Gokce Gun 10:00
But what we were debating about is the influence of the different colonoscopy devices, how it would influence and as the data that we collect from different devices, and we have also the system has to work with these devices, so we had something like two.

Gokce Gun 10:00
Options, for example, whether to have a specific version for each device, and then maybe to train a general model, and then to fine-tune it to each device, or to train on all the data as a mix, and then to test it. And you have, and you don't know at the beginning what to do. You don't know what is the right answer. So we had to try. We ended up with a version which is actually a model which works with all the colonoscopy devices. And to show that it really works well, we did subgroup analysis to show that with the different devices, the results are similar, like there's no device with which the data, the performance on its data, is poor, stuff like that. So this is an example of how to take consideration, maybe with regard to the data and to cope with it towards the performance and the post-market after that, and stuff like that. So on the data piece. And Dr. Gun would love your opinion on this. You know, from the VC and strategic side,

Andrew Pieprzyk 11:04
we've heard reference to the FDA, right? And again, from a geographic perspective, that's already a target population for your data. As you think about broader opportunities for your platforms and your technology, how do you think about geographic expansion, right? And how does that impact your dataset as you move through those phases?

Gokce Gun 11:27
I mean, one important part, obviously, when you get the application to the US. I mean, you need US data, at least half of it. And the same thing in Europe. However, the one important part is the clearance, but then you need to do sub-studies across the world in different areas that really support, I mean, it's not about just having the FDA or just having the CE mark. It needs to be consistent across different regions, and that comes along with a post-market follow-up, right? So it can be in a way of post-market follow-up, but different studies, all this data that you gather is what is relevant and what builds trust with the physicians. So that's fundamental, not necessarily in the FDA or in the CE mark.

Gokce Gun 12:12
No, yeah. So I mean, in a broader perspective, I would say how the data is inclusive, diverse, large, then it makes the result more precise. So beginning defining the first data as inclusive, as big, as diverse, is the best way to go to the precise solution. But on the other hand, we are talking about solutions that are actually progressing every day. So in AI and in machine learning, the end product is learning from the data that we put in. So I agree with, like, the sub-data and subdivisions, but in the beginning, how big and how strong the data defines the whole process actually in a more healthier way.

Andrew Pieprzyk 13:00
Yeah, good. And knowing the audience is probably more on the industry side for me that one of the takeaways is both market evolution and when you think about big data and AI is the importance of the capability of medical and scientific affairs in your organization, right? That needs to be strong, right? We talked earlier at lunch around statisticians on the team, how you go about designing that data and then conducting those clinical trials becomes even more important as you're in market and continue through. And I think that's been some areas that, from an industry perspective, I feel a lot of organizations can probably do better in that evidence generation, and there are gaps in that one, and there's a fine balance there, for sure, because you want speed of innovation with evidence-based medicine, and some of those large-scale data studies are often expensive. So that gets into another question, so maybe I'll pause. Is there anything anyone wants to add?

Dror Zur 13:56
I maybe can add one more thing, I mean, related to what both of you have said, like, about the learning, learning based on the data, because then and the responsibility, I think there is the issue, because theoretically, for example, we could learn from the data online. You know what I mean, like, the AI system could learn from the data online, but then there's a question of responsibility. Do we want to allow the AI model to learn online and then it is with less supervision and stuff like that? And I think at least from what I know, I probably don't know all the aspects of all the options, but at least from what I know, most of the systems, you know, do not learn like on the ground, on the fly. We collect data in an ordered way. We supervise the data that we collect, and then we train the AI model offline, and then we release it in a very controlled way, and stuff like that, like a software version, as we did in the past, adjust the AI. So this is maybe an additional aspect,

Andrew Pieprzyk 14:56
all right,

Andrew Pieprzyk 14:58
so I'm going to change gears a little bit.

Andrew Pieprzyk 15:00
So Pau, you and I have talked a bit around and you referred to this earlier, around AI as a task, right? Doing something quite discreetly, but the delivery of that value through an ecosystem. And I know this is front of mind for you relative to your business, and it's a little bit different than Dror's in real time. So maybe share kind of your view of navigating that Bermuda Triangle within your business, absolutely. So actually, when first applications in AI for stroke, the idea is that, in our case, it started with contrast imaging, advanced imaging, and then they created these communication systems to allow to coordinate healthcare systems, I would say, is as important the AI and the coordination system and most of the companies tend to become this platform, this one-stop shop with all the AI. The thing is that to scale worldwide and rapidly is really complicated to build your own communication system and try to spread these worldwide. So something that we thought is, instead of developing another communication system, we've integrated the CT Scan Viewer, a DICOM viewer within Microsoft Teams. We don't want a communication system in 2000 hospitals. We want it in 100,000

Pau Rodriguez 16:20
so you focus as a company on the AI, what are we solving the best AI possible with everything that we said? But then let's partner. Let's not reinvent the wheel. Let's find a communication system or a way that we can leverage, and that was very important for us. Then you need to see the different synergies. And then the other aspect, in particular in stroke is a complex workflow, right? Because you have the radiologist, the neurologist, interventionist, so there are many different ways in when you get support, and we forget about the IT department, when it's becoming more and more relevant, so you can convince the physicians, but then you need to go through cybersecurity. You need to go to compliance, you need to monitor all of that. So let's leverage what is existing out there, because at the end of the day for us, AI needs to be affordable and needs to be scalable. Otherwise, we are investing the money in reinventing the wheel, and we're going nowhere. So hopefully we're successful in this endeavor.

David Zakariaie 17:24
David, how do you see that in mental health and your application

David Zakariaie 17:29
in terms of the business model, or in terms of,

David Zakariaie 17:33
you know, the business model, and how do you plug into that ecosystem, right? Is when you think about a segment in healthcare that is ripe for transformation, right? We were talking as it's very paper-based and a little subjective in diagnostics today, and you're now coming in on a clinical level with AI in a very enhanced, probably mobile-based system. So you've got quite a bridge to cross from that perspective. Yeah. And so the reason why I brought the business model is we've kind of learned, as everyone here knows, it's like you could have the best deployment mechanism in the world or the worst, but it has to also like a we've kind of been trying to work back, working backwards, figure out what the incentive structure is in mental health are to figuring out how to plug it in a way that clinicians can use it, but also allow us to actually get paid. So our thought process has been, you know, we are,

David Zakariaie 18:28
we're being regulated as a, as a, as a custom medical device, but we're 100% software, basically, you know, it's an app that works on a mobile phone.

David Zakariaie 18:39
And so we've been trying to figure out how, like, how to take advantage of the fact that we're software and the way that we deploy while also kind of hitting inside of the existing ecosystem. And so what we try to figure out was, fine, if there's aggregators of mental health patients that have much larger clusters of patients, instead of selling directly to the providers, because the kind of traditional issue in the mental health space, about 75% of all mental health in the United States is diagnosed and treated exclusively inside of primary care. And primary care in the US is a little bit too,

David Zakariaie 19:17
is a little bit too, a little bit too fragmented. And so what basically discovered was that about 80% of all mental health interactions between provider and patient is occurring on telehealth and post-COVID. It's the only area of telehealth usage in the US where the numbers are still going up, because it's a patient population that traditionally does not want to show up in person, and so, you know, and that's everything from your general doctor who just has a Zoom account to a lot of these more specialty platforms, like Talkspace, BetterHelp,

David Zakariaie 19:55
Pillodoc, and so on and so forth. And so our thought process was, and I think it's like, and you.

David Zakariaie 20:00
It, it in a very similar manner. Rather than trying to build a standalone app and scale it, we've built our product as a software development kit. So basically, it serves as a, you know, as a, as a, as an app inside of an app. So our, you know, our customers are essentially the telehealth platforms. We serve as a plug-in inside of their platform. And so, you know, as you're hopping on for a therapy session or a psychiatry session or whatever it may be, you know, you log on, you know, five minutes ahead of time, the Sensei test will launch inside of that existing environment. You'll take our test, and then at the point of time that the provider shows up, they'll, you know,

David Zakariaie 20:44
essentially receive a diagnosis. And so that's kind of, that's not we still don't think that's the perfect way of doing it, but for at least, kind of our initial launch, it allows us to kind of fit into the existing clinical workflow, at least for the majority of those interactions. So in these two use cases, Pau and David, thinking your initial assumptions on your business model and revenue ran to what you don't now, what are a couple of lessons as you think about interacting with strategics like myself or VCs, or as you think about the expectations your business so taking the business model and applying the expectations to revenue, what are a couple of lessons we can share, or observations?

Pau Rodriguez 22:31
So I would say, in our case, the most effective treatment in stroke is the endovascular treatment, which basically is removing the clot from the brain, and actually is a very cost-effective treatment, but only, I think it's one in 14 patients have access to that, right? So our thought is, how can you focus on basic imaging, basic modality and technological scalability? And the AI doesn't need to be perfect, it needs to be good enough, and in combination with the physicians to do more treatments. So in our case, medical device companies can benefit from more treatments, and we believe that's the way to go. We need to make it again affordable. We need to make it scalable, and we need to work with medical device companies to really make sure that every single part of the world has access to those treatments. So that's the way we can partner with medical device companies or pharmaceutical companies or whoever.

David Zakariaie 23:31
Yeah, I can add that, I think really in order to make it affordable or to make it beneficial for the company, for the investors, and for the customers. So we really want, as has been said, we want it to be prevalent. We want it to be scalable, and stuff like that. So I think there are several things that we can think of, for example, first, of course, the performance. You want the performance to be good enough, so it would be beneficial. But then things like, what I mentioned before, like to make it, in our case, in colonoscopy, to work with all the colonoscopy devices to be agnostic, and then you can deploy it in more places and stuff like that. And then also, as David mentioned, I think it is important to think from the beginning about integration in the clinical flow of the doctor. You don't want to change the clinical flow. You want to analyze from the beginning we are talking at least in our case. But I think most of the diagnostic system, decision support system. You want to support the doctor in the clinical flow that they have, not to change it. So you would like to analyze the clinical flow that they have and to see how you integrate your

David Zakariaie 23:37
your decision support

David Zakariaie 23:37
comments, with the existing clinical flow. And then, of course, the business model. In our case, for example, we found that the leasing model or subscription is beneficial for all, you know, to all sides, like

David Zakariaie 23:53
you're talking about ARR like repeating, recurring revenues, and also it's something new. And then you don't have to buy a device. I mean, from the customer point of view, like the physician. You can, you can, you can lease or subscribe, and then try it from as a company, which we do, of course, the best to have customer retention, to keep them, but they have the option, and then they like it, and stuff like that. So you have to think about the business model also that will help you to attract clients and to be beneficial for you and for your investors.

Andrew Pieprzyk 24:29
You guys all have discrete users. You get into an account, right? Or you might have a physician using your technology,

Andrew Pieprzyk 24:37
but all of your revenue is based off repeat usage, right? How do you make sure they are using it more and more once it's installed? You know, thinking about the compliance to use, or growth in the share of those cases, to your point, earlier. So the way we see it, have in mind the stage of our company. So we passed from a research company to have the CE mark and being cleared.

Andrew Pieprzyk 25:00
Now we're growing the organization. Actually, we hired

Andrew Pieprzyk 25:03
five new people to focus, particularly on customer success, and we've been thinking through these a lot. It's not like hiring a sales team that is a part we like that the same person that actually comes from the clinical evidence and the relationship you build potentially, with these medical device companies, but the relationship we will with the physicians, and you combine them all the process, right? So you spend time with them, and it's more, even more important the customer success than the sales itself. I'd rather focus on a small group of hospitals that perform well, and you are with them, you also learn a lot, and that's fundamental before you scale, because if you scale too quickly and the solution is not consistent, or whatever, you may have issues. So I think that the customer success party is crucial.

David Zakariaie 26:00
I completely agree. I think customer success is a very important function in the company. We have it as part, for example, as part of Operations. We're beginning now to work on it, but I think it's very smart, what you have said, really, I think you don't have to look for very rapid scale-up and stuff like that before first, you know, trial system. Get your feedback. Get the feedback from your customer, from the initial customer, see what you have to fix, and we always have to fix, say, you know, and make your product easy to use and then, you know, push it to a scale-up. And we also, I think there's also mechanisms to do, you know, to follow up, like post-market surveys and stuff like that, to follow about the use of the system.

Gokce Gun 27:00
Yeah, actually, if I may add, is something interesting the customer success role is. It's not, it's not easy to find, right? So we've hired several people with several backgrounds, and we're trying to define these best practices. So you can start hiring somebody that has a lot of experience in clinical studies, because at the end of the day you're doing clinical studies as a way to start engaging, then you have somebody that is more expert in technical implementation, because you need to get the implementation done. Then you have a sales person that shows you the poor medical affairs. So we try to hire people that have like, common values, what is important for us, but at the same time, bring different angles. And they all, we want to train them to all go end to end, because we need one person to answer the needs of that system. Either, if it's technical, then we have our specialized units, but, but you need one person to understand everything to a certain level, because, if not, it's too complex to manage the relationship with but we're implementing it so we see we do it right?

Andrew Pieprzyk 28:00
So I look at the audience here, and the AI and diagnostics is broad, and I think what you just said is important. The opportunities for transformation are broad in the clinical sense, but they also are deep in the delivery sense. So how do you prioritize one or two of those areas and those earlier stages in particular,

Andrew Pieprzyk 28:05
I would say,

Andrew Pieprzyk 28:09
I think that more than prioritizing you to just it's difficult to prioritize it. To be honest, I think that you stick with a small group of people

Andrew Pieprzyk 28:21
that is diverse, and you try to adapt. And because prioritizations change over time, so you need to prioritize, but you need to adapt. And you solve one piece, then you focus on the other one, then you need to go back. So it's not a clear answer for me of what you need to focus on in particular. It may evolve.

Gokce Gun 28:59
Dr. Gun perspective from you, because you probably have a lot of companies come across your desk at all stages that you review. What are the two or three things that you look at relative to their thinking, their maturity that would qualify them for investment, or the feedback you would give based on their thinking?

Gokce Gun 29:16
So I think, since we're a CVC, the corporate venture capital arm of our fund, the most important thing for us is the strategic angle of the company. Like, if it's a good strategic fit, then it's a good investment. But besides from that, I think team is very important,

Gokce Gun 29:30
IP and tech is also very important. And like the market entry, like if the company can go into market or scale up, is also important regarding business models and everything.

David Zakariaie 29:30
I think the market is a very important thing. I can talk not from the VC point of view, but from the entrepreneur point of view. And I think what I recommend is for everyone who begins to begin, or something like that, is really to look maybe a few things, you know, to see that you have an unmet need, that you're going to solve something, you know that, and then it's like a pain point for the customer that you are targeting at.

David Zakariaie 30:00
And then to see, I think in many cases, it's important that there is a technological challenge, because you don't want, okay, you found an unmet need, but you don't want that anyone can do it immediately. You want to have a technological challenge and to see that you can have an advantage there. You know when going to solve it. And then the market, I think you have to see that you already have a business case. And to look at the market is big. In our case, for example, we have data analysis, and now we are, but we didn't know much, I must say, now we know much better.

David Zakariaie 30:30
But the colonoscopy market, you know the endoscopy market, but also the colonoscopy is, is very big, and we now understand that, for example, AI in colonoscopy or endoscopy is, as you know, prospect of more than $2 billion or something like that, and this is on top of the money that is now rolling in the area without AI and stuff like that. So yeah. So I think it's important to choose a market which would be beneficial. It will make your life easier, your investor will be happier, you know, and stuff like that.

Andrew Pieprzyk 31:11
What problem would you solve next if you weren't the CEO of the company you are today?

Dror Zur 31:16
Oh, I see

Dror Zur 31:16
I have a few of my questions that I didn’t give you guys originally.

Andrew Pieprzyk 31:30
Interesting. Yeah, it definitely will be in the healthcare because I think it's, I will tell you, what I, what I, what I don't know what exactly it would be, but it will be, as I said, something that

Dror Zur 31:30
that first of all to find an unmet need, probably in the healthcare domain, and something that will make good, you know, for people, for the world, that really helps. And then, you know, I would have to check that really, there's a business case and technological challenge and stuff like that.

Andrew Pieprzyk 31:30
I guess this will be the priorities that I will give.

Gokce Gun 31:30
Actually, you raised a very good point, right? Because we often discuss with the team, okay, what's the vision of the company and what, what are we really trying to achieve? And at the end of the day, our vision is to unlock life-saving treatments. How can we simplify the imaging modalities so that you can enable more treatments? We decided to focus on stroke, because this is important, because non-contrast, because we can do more endovascular treatments, more intravenous treatments. But if you think about other areas, is, how can we simplify the imaging modalities we discussed a bit on mammography, for example, and these are things that we need to explore. It's not a matter of building AI for the sake of building AI. It's not a matter of becoming a platform for the sake of becoming a platform. You need to look at the medical need and you need to pursue that, because ultimately, we want to become a team that is capable to manage these challenges from a clinical regulatory standpoint without losing the perspective of the medical need, if not.

Andrew Pieprzyk 32:30
I believe that we're all trying to chase the best that we can, because it's a struggle. What are the revenue streams? Where do we need to go? Who do we need to partner? We all partner with everybody, but at the end of the day, we don't generate the revenues because they are not paying enough. Nobody is paying. Let's go to the basics. Let's go to increase the treatments and to simplify that process.

David Zakariaie 33:00
I recently finished my MBA. Recently completed my MBA. So I have a lot of ideas when it comes to how the impact and everything, but if I were my physician hat, and if I answer this, I think if we add anything to the mainstream clinical application, and if we just shorten the time of diagnosis or increase the population to reach a treatment or a diagnosis, or saving a time in the physician's office for a patient, then the impact of it is huge. So even like small things make a big difference in this area.

Andrew Pieprzyk 33:30
I agree, and from one of our perspectives, and if, if you guys didn't hear yesterday, the Women's Health Index, the cancer burden around the world is coming from low and middle-income countries where they don't have technology at all. Right, so the promise of AI delivering more affordable solutions that can get to where the patient is to get a good enough outcome to make sure they don't get into trouble later. It puts us on a different trajectory from a total health perspective. So we often think about, and I like what we talked about yesterday, as incremental to what we have, but maybe only 30% of the global population has access to what we have today. So from a transformational perspective, how do you cross into populations that don't even take or have access to what we take almost advantage of today for from where we are? And that's one of the big areas for me, particularly for AI and diagnostics, that we can look at all of the technological innovation that's coming. But how do we apply that to populations that don't have even the basics, the good enough that we have today, and that the impact on that disease burden will be massive?

David Zakariaie 34:30
Well, and we can a.

David Zakariaie 34:30
Yeah, I'd love to hear your perspective on mental health. This has been a topic in my world for a few years, on a personal level, but, you know, it was a little taboo for a long time, right? Coming out of COVID, I think it's, it's been easier to talk about. So how many you know, what's, what's your aspirations there?

David Zakariaie 34:30
Yeah, and honestly, I actually, I think COVID gave the people that were already talking about mental health like a bit of a louder voice, but at least in the circle that I run in, I still think it's just as taboo as it was before. Like, especially outside of the United States and like, I think, like, we were a bit more relaxed in how we talk about depression or anxiety, but anything more severe than that, I think, is like, still as taboo as it gets, if not more taboo. And we're not. We don't need to go into that today, but I have a lot of I think, like my personal opinion, is that mental health will be stigmatized until the point that inpatient mental health care gets kind of stripped down and rebuilt. Because it's like, as long as we basically treat inpatient mental health care from like, in every way, shape of the form, like a prison, it's going to be as stigmatized as things are now. And I think like that, like that, that's not necessarily what Sensei is doing. But I think we from like a general mental health space. If you can rebuild inpatient mental health care,

David Zakariaie 36:24
that is how you can kind of slowly start to strip the

David Zakariaie 36:30
the stigma away. But in terms of, like, what we're trying to a mental health like the way we think about the mental health space is that the way we both manage and diagnose patients today is basically the equivalent of diagnosing someone with diabetes, but you never measure their blood sugar, which seems pretty ridiculous, but that's effectively the standard of care. And I use that analogy for a couple of reasons, because the same way it's very simple by measuring someone's blood if they have diabetes or not. It actually has always been that simple

David Zakariaie 37:06
to understand if someone is suffering from some simple mental health indication as well. The problem is, you needed a functional MRI. It's very expensive. It's, you know, and so we'd never be able to do it at a scalable manner. And so what we're trying to do at Sensei is essentially build the first diagnostics and severity monitoring tools for mental health. Our initial focus is on anxiety, depression, and PTSD, and we're not trying to replace the clinicians. We just think that, you know, there we have very powerful treatments available in this space. They're very effective, but the actual matching never happens. And so if we can put really powerful measurement tools in the hands of the providers in the space that will allow them to kind of

David Zakariaie 37:52
actually understand what a patient is happening and kind of start to solve the mental health crisis, patient by patient with an actual understanding of what they're suffering from.

Andrew Pieprzyk 38:11
The vision you all have is amazing and aspirational and inspirational.

Andrew Pieprzyk 38:11
And I think I speak for the panel today, the opportunities are abound. How we get there becomes the formula that I think we're all trying to solve and in the end, even just our conversations over the last few days, is we're stronger together. So talk to your colleagues. It's the power of meetings like this. It's the power of networking,

Andrew Pieprzyk 38:36
because no one entity is going to do this transformation on its own. It is complex, inherently on a technological level, but even more so on a delivery level. So guys, thank you so much for everything. It's been fantastic. And guys, thank you.

Andrew Pieprzyk 38:52
Thank you.