Podcast Episode 33

AI in MedTech: Pioneering Innovation and Business Growth

Get valuable insights into the transformative potential of AI in MedTech with Head of AI and Technical Product Management at APRO Software, Kanstantsin Vaitsakhouski, in this episode of the Tech Business Roundtable podcast hosted by Ryan Davies. They discuss the current landscape of AI applications in healthcare, its impact on medical tests, clinical trials, and mental health solutions, data integration and collaboration with the scientific community, and the importance of a science-driven and data-driven approach. Tune in to stay ahead of the curve and unlock the limitless possibilities at the intersection of AI and MedTech.

Kanstantsin Vaitsakhouski Founder at Adventum.ai

Listen to the Episode Now


Ryan Davies: Welcome everyone to the Tech Business Roundtable podcast show. This is a podcast show dedicated to shining a spotlight on tech innovators, entrepreneurs, founders, and the compelling narratives behind the movements they’ve established. I’m your host, Ryan Davies, and I’m hosting today’s discussion on AI in MedTech, pioneering innovation, and business growth with Kanstantsin Vaitsakhouski. Kanstantsin, how are you doing today?

Kanstantsin:  I’m doing well. Thank you. How are you?

Discussion on AI in MedTech

Ryan Davies:  I am good. I am good for whatever this airs. This is Monday, early morning for me. So this is the start of the new week, and the energy is high for me. And we’ve got a really great topic here, Kanstantsin, you know, an expert in this area. And there is so much, I mean, in our pre-talk before we started recording here, there is so much we can talk about with AI and MedTech, and all of the areas it touches, all the barriers, the challenges, the opportunities. But right now, the best is for you to kick us off with a little bit of, you know, a current overview of the landscape of AI and MedTech and how it’s changing the healthcare industry.

Kanstantsin:  Yeah, thank you. So, the current landscape of AI in healthcare and websites is something great, and it’s a really big thing. The good point is that this is only the beginning of AI applications in this domain. So, for example, this year, Google launched a really advanced state-of-the-art enlarged language model for healthcare, which is called Med Palm Two. So the second, the first one, was also launched by Google. But anyway, it’s a real beginning so that it can operate with answers with medical information as a great good student from the USA and all from India. So, it was hardly tested on the examination questions from these countries. But at the same time, it’s not as good as experts. And at the same time, there are a lot of questions behind it. This is only a question-and-answering system that also has a huge demand for different image processing. AI is for the MedTech for MRIs for histologic hexology for x-rays. There’s another story that this domain is also really huge and really interesting. This direction is much, much better developed than large models, and the first algorithms that were applied for the X-ray analysis could do this even several decades ago. Also, we have a huge set of different intelligent algorithms that are responsible for finding drugs, the necessary proteins, and the correct molecules. That’s another direction. So it’s really a huge un-boundary domain.

Importance of AI Applications in Healthcare

Ryan Davies:  I mean, This conversation. We could have five sub-podcasts in each of those areas. Because AI took each one very differently, right? You’re the head of AI and technical product management APRO software, right? So, you’re the leading figure in the world of AI innovation and entrepreneurship. You’ve been doing this for 12 years. What motivated you to focus on AI applications in healthcare? Is it just that there’s so much growth and opportunity here? Is it a lagging industry from that standpoint? Was it for you that kind of drove you to this space?

Kanstantsin:  You’re right, that’s the point, but that’s the main point only from business prospects because, in health care, that’s really a direction. That’s really a nice domain when small start-up companies can compete with such huge corporations like Google, Microsoft, and Meta. So, at the same time, this direction, this domain, is so responsible and is so important for everybody’s life and my life as well. So, these technologies will change healthcare, which shows the future to be something much better than we have now.

Challenges and Opportunities in Integrating AI into Healthcare Solutions

Ryan Davies: I mean, there’s so much to cover here. Again, I’m going to be broad when we start here, but there’s a lot of uniqueness in this space, right? Compared to other areas and other pillars, what are some of those challenges and opportunities that businesses face when they’re integrating AI into the healthcare solutions space? And again, this could go on for a while. We’re going to have a few questions about this one, but I’ll let you kind of kick it off here.

Kanstantsin: Let’s start from the very beginning of the first challenge. We need to understand how to make this area, how to build the architecture, and how it will work. For example, in all the others that way, if you would like to make I don’t know road traffic. So it’s quite an obvious task, we can understand it. It’s clear how to do this. We only need to build a reliable model of healthcare. It’s totally different. For example, for histological diseases like ulcerative colitis, you can ask 100 different experts about the diagnosis for one and the same patients and only 50% of experts will agree with each other about the test results, the right diagnosis, and the right degree of the illness itself. And we have many such areas where experts don’t know how to combine knowledge of each other. Also, the assessment criteria can be really blurred. For example, even in scientific papers, you can read something like this level of this disease will be characterized by a medium to high presence of neutrophils. What is medium? What is high? We don’t have any explanation. So it’s really, really difficult to explain how to calculate how to build. So what do we need to do? So it’s really difficult if we somehow, that’s the first challenge if we manage to understand it. The next challenge is the data because if you want to build AI, you must have a lot of data, and this data must be reliable. So AI, that’s a data AI development, this is like 80% resources and time, that’s about work with data. And in health care, this is highly regulated. It’s extremely personal data, and it’s extremely difficult to get this personal data, especially if we speak about rare diseases, which are not COVID, not something like this, but really diseases like genetic disease, which appears maybe on 11 in what you when you have one case per 1000 citizens. So that’s extremely difficult. Also, you can’t just come to a random hospital and say like could you please give me like 10,000 examples of data for this disease? So, they have much less, they have specific data formats, and you can’t expect the data formats from one medical organization to be really similar to the data formats from other medical organizations. By data format, I mean not only the file format but also the way they work with these files. So and also sometimes and not sometimes, but truly us if we get this is just data about the disease itself. You need to have metadata, which is basically information about a person. For example, somebody has, let’s say, problems with the heart, and you would like to make some kind of AI prediction about the future of this disease or this kind of person. Next, you need information, not only about the disease itself but about metadata about these men and women, the age, the risk factors, the gender, the smoking or alcohol addictiveness or something like that, and analysis of other things. All this together can take a really, really long time to be collected. Also, it’s extremely difficult and expensive to collect data in this domain. This is an exchange, and this is just another point of the story. I just had to mention it. This difficulty also provides nice business opportunities for new companies. So it means that you can create a company who will help to collect such kind of data, who will be able to make an annotation of such kind of data just and this be really, really important and this will help you to solve the issue number one with an understanding of how to measure this issue, number two this data and understanding what you can measure. So next, you go to the diseases themselves, the different variations of these diseases, and medical tests. Do you know that the Medical Health Organization has about 55,000 different codes for different diseases? Together with this, we have at least 10,000 different medical tests. Together with it, we have several 100 different file formats for storing the results of these tests. If we simply multiply everything to calculate combinations of all this stuff, we will have the answer about several billions of combinations. These AI systems have several billion directions to work, be this only healthcare. And this is why it’s so difficult to create AI healthcare. This is why even there are such big and powerful companies, like Google, that move extremely carefully and only take small and really reliable steps in this direction because this demo is just too big and too difficult. So, issue number three and issue number four regulation. So even if we forgot about any regulation in the world, regulation of technology in healthcare is the strictest in all countries for, I don’t know, several decades, but maybe even half a century. It’s extremely difficult to prove that your step is really safe, and it also takes time and money. So here are four main changes. The first one is understanding of the problem itself. The second one is understanding and collecting the data. The third one is understanding, let’s say, your domain; it’s when you work with data out what you assess and how you assess what the result will be. And the fourth is the regulation. So, altogether, this domain is one of the most difficult, and at the same time, it has the biggest prospects. So I really believe in it.

Ryan Davies: There’s so much to take away there, and I think that was an absolutely perfect summary of some of those barriers that exist. But it means that it’s an unprecedented opportunity in this space, almost more than anything in AI, because the challenge is so big and so difficult. But a great takeaway there is to break it down; you don’t have to try and solve every problem at once, start with like the really, you know, just OK, that’s what I’m trying to solve. Now, how do I even break that down? Smaller and smaller and smaller to find these solutions as you build up?

Kanstantsin:  Yeah, you’re right. With this approach that you mentioned, when you go from small steps to small pieces of this and the big structure of the medical AI, we can combine it into something bigger and bigger. This approach is really promising and will give the best opportunities; just recall that a few years ago, there was a huge story about IBM Watson. This is because they invest so much incredible money, many incredible resources, and time in the development of this system. So basically, IBM Watson did. They believed that this system would be advanced, state-of-the-art, and the best AI assistant for AI medical diagnostics in the world. So, unfortunately, the approach was wrong also because they already would like to build something big, just not this, not from bricks but from big stone. Unfortunately, the expectations were already overestimated. I don’t remember all the details of this story, but I have already been watching for several years. It’s definitely not in newsmakers in the mass media.

Ryan Davies:  You just talk about that. It’s like breaking down the human genome project all over again, trying to crack these codes. As you said, there are so many variations you’re collecting from so many different sources. You’ve got test reliability, you’ve got data formats that are different languages that are different regulations that are different, you know, like where different countries, different regions that have to try and do all of this, there is so much to be able to that AI can help bring this together. Have you seen the integration of AI impact into the medical-tech space and, like, some of the impacts on business growth that have come about from that? I’m sure with APRO, you’ve been a part of this already. Could you share a story there, too?

Kanstantsin:  Yeah. It definitely impacts just. I’ve mentioned a very general example currently, like each MRI scanner, each CA scanner, and each model X-ray machine has AI inside it automatically detects tumors and makes some kind of classification segmentation. It can do really a lot of things, and it can do things much better than humans. It does not voice humans, but it really simplifies the lives of doctors and saves the time of such important people as medical doctors. So that’s great. That’s perfect. From my experience with APRO, we help to improve medical tests, clinics, and clinical trials. So, we participated together with our partners in the creation of the AI system for the assessment of some kind of histological indexes. So, we worked with images, cells, and tissues. This direction is really not well developed. So it’s really quite new. So we had to come up with the idea of how to build the architecture, how to summarize the opinions of different experts, and how to make that data validation and data labeling. So because, again, this is the dumb way, this is the direction when different experts simply do not agree with each other. We can’t find any gold standard here. It was an additional extra change in this domain. So, the project was successfully done, and I’m really proud of the quality and the opportunity that this project will give to humanity. So, the other direction. So it’s very personal. I’m currently at the very beginning of developing another start-up. It’s about healthcare, it’s about mental health, finally. And AI can really be used here. So because people because if we speak not about this, like problems with data problems with regulations, which are usual, usually if we work with some kind of medical tests, medical texts. But what if you work directly with humans? For example, a psychiatrist or psychologist so that you can be biased, you can be biased because maybe your lunch was not very delicious. You can be biased because you have a bad mood or something else, so the opposite person can also be biased against you the same for different reasons. You may have shoes of the wrong color. That’s possible so that you may have a meeting in the wrong place, something like this. But sometimes when people buy, they don’t understand the reason. They don’t even understand what happened to them, what’s going wrong, or what’s going on. So, have you tried to solve it? At least all it is to solve, to try to solve it. So, I’m very passionate about this.

Ryan Davies:   I think it’s obvious that there must be, I don’t want to, maybe I shouldn’t say obvious, but maybe I’ll ask you the question around, you know, how receptive is the health care sector to embracing AI solutions? Like do you see that this is kind of the trend? Are they ready to get on board, or is there still a lot of resistance to bringing AI into the healthcare sector, or is it kind of again maybe regional dependent or industry dependent or things like, you know, sub-industry dependent there?

Kanstantsin:  Yeah, it’s a really interesting question because I don’t have a simple answer. Yes or no. So because I like the complicated answers, I love these. Because I can see everything that you mentioned, I know people who really believe that AI will change the healthcare they tested for something better and that it must be integrated into it. At the same time, I can see, and I know, stories about a huge resistance when people think it’s just some kind of strange, useless black box. So this attitude, like it hardly depends. It’s highly different, even between different hospitals. I’m not talking about different countries or regions, just between different hospitals in the same country. So again, it’s not about technology. That’s about humans’ perception of this technology, right? So, but if we ask like medical experts, if we ask entrepreneurs, if we ask, I don’t know, for example, Bill Gates, so all these people will prove to you that AI and healthcare, this is the future, this is a perfect match and they will be together. 

Application of AI in Smart Mobility

Ryan Davies:  I know you have a lot of experience in both health care and smart mobility as well. And there’s a critical intersection there that’s happening, that we’re seeing more and more that’s now becoming mainstream accepted. You should give us a quick little insight into that as well. I know it’s a bit diverting, but they come together as well in terms of, you know, the synergies that exist between the applications and those two domains.

Kanstantsin: Yeah, that’s true. Both domains have quite specifics. So they must, both domains, unfortunately, have a lot of resistance. Both domains are not so obvious if you want to build substance. So, we already discussed MedTech and healthcare. But their smart liability is a bit different. I am based in Europe, and in many, many countries, this domain is really well developed. So, anyway, it’s still beginning. So, for example, in this domain, we can use AI for such tasks as a prediction of road traffic and average speed. We can build a logistic path. We can make optimization of this traffic itself. We can count on detecting vehicles to be sure of this; for example, trucks don’t try to pass on, call the roads or near schools or somewhere else, or for example, trucks will deliver dangerous goods. That’s also the point. So we can track such cars and records with dangerous goods using AI technology as well. This is a really huge domain with many applications, especially if we combine data from several sources. So this smart future domain also has a lot of interconnections with different hardware devices, different sensors, Wi-Fi Bluetooth magnetic sensors that are installed, directing the world smart cameras, radars, satellite cameras, and many other things. So it’s really huge, and I have to work with many different kinds of data and sources of data. It’s also linked with MedTech AI, which has a huge number of data sources and data formats. Another difficulty is that smart mobility is usually something that is interconnected with the state and the government. So when you have government, there is also always some not-very-fast decision-making, which is a bureaucracy. And also, you do not have many such projects. For example, some Brussels administration will launch a tender for the installation of such a system inside Brussels. And somebody wants this, we want this tender, and for the next five years, this business opportunity will be closed. So it’s quite difficult. So, if you are the winner, we will have really incredible opportunities. Because of the investment in these projects, it’s quite big because the results are quite important. It’s quite a bit. For example, if we speak about MedTech and AI in med tech, at the end of the day, they have this human life. So we have our house. So it’s something this system which can impact on your death or your life. So, if we speak about smart mobility, it’s similar. So it’s about time; it’s how much time you save, how much money, how much money you save. Also, if we speak about emergency services, so smart mobility management, the way you are able to manage traffic instead of the city or not really depends on how urgently emergency services will be able to achieve somebody who needs this service. It’s also crucial. This is also extremely important. This is also extremely interconnected and leads to climate change topics because the more advanced the smart mobility system, you have the less CO2 emission you have in this region. If you manage to build a smart mobility system, for example, something about public transport, if you manage to make it truly convenient when the buses and trams arrive at the right time. So when people really need it on, when you have the biggest number of potential passengers, most likely somebody decides, oh, today, I will not use my car today. I will take a bus. So, and it’s really in Europe, it’s already the case with the trains, and this really works. So it’s not about traffic management, it’s about the management of public transport. So, it’s about everything except airplanes and ships. So it’s also about selling different tickets. So it’s about parking services. It’s about the security services because this system really respects the modern smart mobility system; they really respect your privacy, but they can really track a lot of information about the board and people in this world, so from gadgets to different details about your car number, model manufacturer, and many other things. So it’s a really interesting talk. So this is about security; this is about saving time. For example, if the police would like to know, if so, if you have, let’s say, a stolen white car in this region, it can open a special dashboard on the smart community system and check and search by car cover on the database from smart cameras. So it’s very interesting.

Collaboration and Partnerships in AI Solutions in MedTech

Ryan Davies:  That’s, wow. You could see how they just say so they impact each other. The commonalities between them and how they’re going to be intertwined are coming to the surface with that. I know we’re running a little long on the episode, but I love the conversation. I want to ask a couple more questions here. Let’s talk about the importance of collaboration within this industry as well because forming partnerships is going to be crucial. We’re talking to, you know, our tech business founder audience and it’s such a fine line between partnering and oversharing and partnering for the benefit of pushing businesses forward. But you know what you have seen in terms of the importance of forming these partnerships within AI solutions in MedTech. 

Kanstantsin:  So in MedTech, that’s really crucial to building a trustworthy partnership with hospitals and with medical scientists. So because that’s a huge probability that this, we try to solve the idea to the issue on what to build and how to calculate if you have to make a really long consultation with scientists, maybe with people who develop these tests, maybe with people who develop this methodology, this is crucial, this is extremely important. So, this collaboration with the scientific domain method will also provide a huge help with regulation. So we use different compliance things. It’s also very important to collaborate with hospitals because scientists are somebody who is very beginning, so a hospital somebody within the bar basically who will take care of people who are whose wives will depend on your solution. So also, and you know, you must have this feedback, and this is not just about responsibility. So it’s, of course, responsibility distortion because we have the humans at the end of the day. But it’s also about helping with regulation, helping with compliance, and helping with business opportunities because the scientific community and business community are really highly interconnected, so scientific scientists can become businessmen and launch new companies based on his or her recent discoveries and vice versa. A huge company can invest in science to have a new discovery to launch a new business domain. AI is something that can unite science and business here. It’s just in this like health academy.

Closing Remarks and Contact Information

Ryan Davies:   That’s just a perfect way to sum up that area. We’ll kind of close off here with breaking out the crystal ball for you here looking again, you know, as a final thought, key message, or anything that you think, maybe, a trend that’s going to define the future of AI and med tech or how tech business founders can really position themselves for success in this space. Any key takeaways here? There’s been tons of advice in terms of points here, but this could go on for a whole episode again. So I’ll turn it over that.

Kanstantsin:  So it’s so important at the same time. Such a difficult question. If you would like to succeed in AI in the MedTech domain, you must believe in science, and you must rely on science. So because this understanding of what you want to build at the very low level may be on the level, not only the math, because math will be the next step on the level of biology, biochemistry, or psychology, it will be the first and very crucial step. So also you have to understand the disease and people with this in an extremely detailed way. So you must be, you must know what so you must keep in mind that A I can make mistakes, people also can make mistakes of course, but what kind of mistakes is less dangerous for example, if you have like five grades on the same disease when the grade number zero, it’s like nearly nothing. And the grade five you have, you must provide emergency help to this patient. So what if this AI system will create this disease? Make a mistake. For example, what if you make a set? So this person has grade two, but in the variety, the person has grade three or vice versa; which option is better? So you must keep it in mind. The science-driven and the data-driven approach come first.

Ryan Davies: Perfect, summary thereof, a good place to close off, give people. There’s so much to take away from this episode. We could run a few more in this space for sure. And I may not let you off the hook yet. We may have to do a few more discussions here. But for our audience, I want to give you a minute here just to let them know how they can get in contact with you. You learn more about APRO, you know, all of the pieces here. There are going to be people out there who would say I’d love to pick your brain and get a little bit more information, maybe from one of those important partnerships we’re looking at here.

Kanstantsin: Yeah, sure. So you can contact APRO via our website APRO. AIso, or you can find me on LinkedIn. That’s my main social network, and I will be happy to cooperate with you to discuss it with you. Together, we will be able to find something incredible that will change the world for the better. So for the better tomorrow.

Ryan Davies:  Absolutely perfect. You know, that’s a great place to close off. Thank you so much, Kanstantsin, for being here and giving us so much knowledge so much to take away in AI MedTech. For our business, you know, founder audience to be able to understand just this limitless opportunity that exists in this space kind of thing. It’s a heavy lift. But for those that are willing to do the heavy lifting, there’s just so much here, and you know, pioneers like you are blazing the way for us. So, thank you so much for being here and sharing all of these great informational pieces with us today. Really appreciate it.

Kanstantsin:  Great. Thank you very much.

Ryan Davies:  So, with that again, thanks to Kanstantsini for this amazing podcast in MedTech, pioneering innovation and business growth. And I want to thank our listeners as well. We can’t do what we do without you. So until we meet again with another amazing TBR episode, I am your host, Ryan Davies. Take care, and stay curious out there. Thanks so much.

About Our Host and Guest

Director of Marketing – Ekwa.Tech & Ekwa Marketing
Read More

” This difficulty also provides nice business opportunities for new start-up companies. So it means that you can create a company that will help to collect such kind of data. We will be able to make an annotation of this kind of data just, and this be really important and this will help you to solve the issue. Number one, with an understanding of how to measure this issue. Number two is this data and understanding what you can measure.”

– Kanstantsin Vaitsakhouski –