Medical Devices

Saso Papp: Robert is also passionate about new technologies, but from a totally different standpoint. He's a legal professional with more than 10 years of working experience, spanning from time as head of the industrial property department of the Slovenian Intellectual Property Office, where he was actively involved in the preparation and drafting of laws and EU regulations, to being a legal advisor on IP matters. And let's not forget to mention, he will soon receive his PhD in Intellectual Property Law, Competition Law and Licensing. Robert, welcome.

Robert Kordić: Thank you very much, Saso.

Saso Papp: Okay. Let's jump into today's topic, guys, medical devices. What's hot at the moment? Can you tell us what are the design and patent trends in medical devices sector, maybe Matej?

Matej Vengust: Sure. Yeah, definitely. These days, medical devices have grown from what are usually conceived as a, let's say, stereotypical devices that invade the human body in some way or another to a completely new level, either in terms of radiology like MRI, CT scanning machines, or even x-rays, which have advanced a lot. Then we are talking a lot about, of course nanotech, micro, let's say, robots that would be inserted in your body to do some type of evaluation, diagnosis or even cure cancer cells. I mean, kill them to be precise, but it definitely doesn't stop here. It's not just about the integration of new machines that are in contact with your body. We are talking a lot, of course, also about artificial intelligence that helps doctors do the final, let's say, diagnosis on different states. Of course, AI helps to do that a lot faster and of course also more precisely, but the human doctor is still the one who makes the call.

Saso Papp: Robert, what do you think about it?

Robert Kordić: Yeah. I also believe that, for example, the field is very dynamic and innovative. For example, only last year, despite the pandemic, the overall numbers of European patent applications are almost the same. And statistics published by the European patent office show that innovative healthcare-driven patent activity is as important as it was ever before. The medical technology was the leading field in inventions in terms of volume and together with the pharmaceutical and bio technological spheres, they were the fastest growing areas. Medical technology accounted for the most inventions in 2020 and surpassed, for example, the digital communication industry as such. And the key aspect that it's not just a very innovative, but also a very R&D driven field. The amount, the numbers spent on software development or basically also on the development of the AI aspects in the medical device industry is the leading one. I would say it's an integral part of the industry as such. Still, large amounts of data and information on patients is not structured. So this is one of the areas where AI can and will improve things dramatically.

Saso Papp: Okay. Yeah. And what we have to mention is one case where AI was used to interpret images that are usually sent to radiologists. Experiments have shown that AI outperforms experts in terms of efficiency and accuracy by an average of more than 11%. That's big, right?

Matej Vengust: Well, yes it is. But we still must consider that this goes to some degree. If they are repetitive or so to say standardized, cancer for standardized illnesses, then that's okay. But when it comes to some type of new form of, I don't know, I wouldn't say really mutation, but when it comes to cancer, for example, a lot of times there are some, let's say, signs that are not just seen basically from the scanner, from the picture. Computer vision can help a lot, but still, we must not forget that human intuition far surpasses computing power, right? So this is-

Saso Papp: So doctors don't be worried, right? Yes.

Matej Vengust: Definitely not, yes.

Saso Papp: Okay.

Matej Vengust: So it goes to some degree, it is a huge tool that helps doctors, but just to add all the diagnosis to a computer at certain moment, it's definitely not possible yet.

Saso Papp: Okay. Everybody's talking about big data and artificial intelligence. So what do you think the importance of big data and AI is in the medical devices sector?

Matej Vengust: Well, when it comes to big data sectors, we usually call the ... We call it simplified, just big data, right? Well, this is the, let's say, the food for the AI algorithms to function on. So the more data, the more precise it gets. It's just statistics. It's simple as that. So when it comes to crunching big data and tons of tons of terabytes of data, then obviously computer is faster here. But when it comes to let's say punctuality or so to say emotional intelligence, when it comes to interpreting the data, then of course humans still have to have the oversight, let's say, over the decision-making of a computer.

    But if you ask any doctors that worked before, let's say, the AI era, when it comes to computer vision and image recognition, they would tell you that they would probably not go back to the days when they just have to take the paper and circle what's important and then compare it to maybe some other pictures from, I don't know, some other patients. Yeah, this helps a lot surely. And of course in some type of illnesses when really time is of the essence, we are talking maybe even months or even weeks here, that definitely is a lifesaver and a huge, huge aid to the doctor.

Saso Papp: Matej, at Nervetech you are working with a driving simulator, so maybe a question also is relevant: how far are we from self-driving cars?

Matej Vengust: Well, from what we see now, quite far. I mean [crosstalk]-

Saso Papp: Okay. Elon Musk wouldn't agree.

Matej Vengust: Well, he's been saying that for seven years now. Each year he says, "Next year you're going to have a self-driving Tesla." Well, for first part, and I think my colleague here will completely support me, his legislation part, whose fault is it? Is it government's for the regulations, is it the insurance company? Who's going to pay the damage. Is it car makers? Nobody wants to have the responsibility, right? So it's still under human and it will be for some time, at least 10 years. Of course, there will be some separate routes. Well, autonomous vehicles can drive, but then again, how is that there different from a train? If you are just on a two-dimensional, straightforward route and you can't interfere with other traffic, then not much is accomplished. When it comes to self-driving or driving per se, the whole body is involved.

    It's not just about intelligence or the computing power of your brain. Of course, they play a huge role, but the whole body involved, your hearing, your eyes, everything. We are trying to replace that with certain sensors like collider sensors, computer vision cameras, even sonars, sound-based devices. But the problem here remains, what you've accomplished throughout your life, not just driving from when you get your licence, but your orientation and this, so to say, emotional intelligence can solve especially critical situations much faster and much more efficiently than any computer right now. So even a bad driver can resolve a critical situation in most cases or in unpredictable situations that wasn't planned by the programmers, much better than any, so to say, AI in an autonomous vehicle.

    And especially when it comes to certain situations where there are maybe no markings by the road or even tracks, the signs on the road like lines or stuff like this, then these systems completely fail. Or maybe, for example, having snow when there's no visible marks, or heavy fog, or even heavy rain. This proves difficult for even lighter or computer vision cameras. Because if you want to have a 100% computer vision-oriented autonomous vehicle, then you must know that there's a huge, huge, super computer in your back, I mean in your trunk, processing and struggling to process these huge amounts of data, which it compares to some previous, either simulated or real life experience that it's stored, but it's never, never as fast as human brain.

    So here we are still coping with these problems and of course in perfect world, when, for example, way more Tesla test their vehicles on a closed ... It's not close to the public, but there are pre mapped roads up to... Scans up to a millimeter, it's completely different than putting these vehicles in Kolkata or New Delhi. They will completely fail. And traffic culture, if you go from, let's say, San Francisco to New York down to Texas, we learn that through our road shows to US, that driving culture is completely different. And this is also a huge part of why an autonomous vehicle that's produced and tested in, let's say, Silicon Valley won't function well in London, for example. It just doesn't compute. Yeah.

Saso Papp: Okay. And Robert, from the legal aspect?

Matej Vengust: Well, I believe as Matej already mentioned, we are still in certain aspects or only at the beginning, and there are various questions which open up. For example, only in April this year, the European Union proposed an artificial intelligence regulation, which sets out structures that ban the use of certain AI and heavily regulate other high-risk aspects of AI. And I believe as Matej already mentioned, the amount of data is the crucial aspect of those things, because it's not only autonomous driving, it's various aspects, because data has become the essential source for economic growth, for competitiveness, for innovation, for job creation and other social aspects.

    So the amount and the necessity for data sharing will become also proof we have just entered a new phase in this, basically, idea of data governance act, which is basically a legislative proposal establishing the data governance framework for sharing industrial data. The initiative intends to bring together the vast amount of data produced by European industrial-based companies, and to basically create this effect that analytical and structural data is shared and also transferable, because most European companies and especially small medium enterprises are reluctant to share their data because then you're breaching the privacy law, the confidentiality requirements and also their own investment in this aspect.

Saso Papp: We will come back to breaches and data protection, but is it at all possible to patent AI inventions, Robert?

Robert Kordić: Well, we have to basically consider here two aspects. AI as such is considered basically a branch of computer science and therefore inventions involving AI are considered so-called computer implemented inventions. In this context, European regulation and also the Slovenians show that there are certain opportunities. Computer-implemented inventions are treated differently by patent offices in different regions and in the European Union, for example, an innovation or basically an invention of covering software is not possible. However, there is the possibility to patent if they have a technical character. So to simplify, you cannot sort of patent a single program, but the program together with its hardware, then yes. Over the years, we have established a case law in disregards, but in recent times, a different question pops up. And that's the question of whether the AI as such can be the inventor. And in general, from the European perspective, you have to be a human being to cover this.

    But there are several currently ongoing cases in the United States and Australia and even in South Africa, where a person Steven Taylor filed two patent applications and he identified the inventor as DABUS, an artificial intelligence system created by him. Through various loopholes, he basically provided the necessary documentation and provided the reason why DABUS should be the inventor. But currently, this question is much debated and it opens basically a follow up question: is our current patent system enough to cover all these aspects? Does it need modification or do AI created inventions at the end need their own legal system?

Saso Papp: Okay. What about the synergy between man and machines? So if we are talking that AI can be the inventor, this is kind of a gray ground, right?
Robert Kordić:    Correct, correct. So the synergies now come... What do we want basically with an AI? Well, there three options. First of all, I have a problem, find solution and the AI should verify the solution. The second option is basically, I have a problem and the AI serves as the problem solver and gives me the solution. These are currently possible. But I believe we are currently the farthest away from the aspect where the AI asks itself or creates the problem and solves it [inaudible]. We are still far away from this aspect and the aspects of man and machine show exactly their differences. Because at the end, it can only work with the data we provided and we feed it.

Saso Papp: Yeah. New technologies are something inevitable as it seems, but how safe are those technologies? There have been numerous cyber-attacks already. So how do we prevent breaches and stealing of very personal medical data, especially if we talk about medical devices?

Matej Vengust: Mm-hmm (affirmative). Well, of course, first of all, there's legislation. We have GDPR here in Europe, for example. So there are certain levels and, let's say, layers of protection of what can or can't be distributed freely. This was, I guess, the first step. Then of course it comes to the... How vulnerable is an actual mainframe and with what would you attack it? And it's usually tactics of fight fire with fire, right? If there's AI helping hackers, then on the other side there's AI helping the ones who are trying to protect their data sets or whatever. But here again, it comes to human nature, I guess. Why would somebody want to steal somebody else's data? Well usually, they're up to no good - a kind of blackmailing. All these attacks have certain, let's say, end costs in the end, usually money. And of course we have also industrial [inaudible]. There's numerous reasons why this can happen, but obviously if a computer algorithm can attack, it can also protect. So here it's like a [inaudible] game. It's ongoing. Usually thieves are one step ahead.

Saso Papp: Unfortunately, yeah.

Matej Vengust: Yeah. But no, here of course, this is... Data has definitely become synonym for fortunes, for money, as we can see with crypto, as we can see with huge amounts of data that companies are in industry uses. Definitely without that, we wouldn't know the world as we know it right now. So there is a huge value definitely in that. So this is also wide opportunity even to start thinking about stealing.

Saso Papp: Yeah, yeah. But it's one thing if somebody steals the data about what I want to buy or my buying history or my health data, Robert.

Matej Vengust: Yeah, exactly.

Robert Kordić: Yeah. That's basically a huge difference or at the end also not because data is data. It's become a crucial commodity today almost like gold. But as you said, the impact that the ransomware, spyware or various other attacks have, are quite more devastating on certain facilities, for example, hospitals. In recent years, there have been several attacks on hospitals, ambulances, medical providers in Ireland, France, Germany [inaudible] basically at hostage because their data was coded and could not be any more excess.

    And it took sometimes several days, weeks that basically the system went back to normal. And if you can imagine in a hospital where every second sometime means life and death, the access to the data, to the crucial data of the patient is of utmost importance. And unfortunately, the IT systems and such public institutions have sometimes been neglected due to various budgetary restraints and they have stretched their services. And that means that at the end, they're the last one to receive sufficient financing and support to prepare against such attacks. Now, or we can change it as well, yeah. As Matej said, the AI is the... The computer is the problem. The computer can be the solution also to this.

Saso Papp: Okay. So we have new technologies that will help us in so many ways, but we agree, we need to be careful in developing not only those, but also develop the data protection systems to avoid breaches of data. But let's be honest, introducing new technologies has never been easy and without a cost and we, as the human race, always find solutions to tackle side effects. Well, not always perfect, but close, right. So Robert and Matej, what are your thoughts on that?
Matej Vengust:    Okay. So yeah, of course, new technologies bring great new opportunities and wellbeing for us, for humanity. So how we tackled for example, because at Nervetech, basically it seems like the core product is hardware, but it's actually not. It's the algorithms that capture and interpret the data from the patient, for example. We just got positive opinion from our national medical devices association or agency that we can use our product in medicine. We just sold the first simulator to one of the biggest rehabilitation centres here in, let's say, Southeastern Europe. So it's called Soča, you probably know it. And of course we collect a lot of data that's connected to humans’ health, biometric data that shows a lot about your efficiency or deficiency. And one of the simplest ways in the beginning we used, how to protect the data was simply moving the data to a different location than the person's name, and it was coded.

    So you just got the code and the name and this data without any information about the patient or about the subject, which is not worth a lot of money. Especially because you need a region, you need demographics and of course you need a name for the medical institution to be able to use that. So yeah, one thing is just keeping these things apart on different servers, coded of course, names, and this is as simple as it gets. But of course hackers know how to work this way around this coding. So definitely there are numerous encrypting services even available online, then you have advanced routers. But since those machines were programmed by a human being, then obviously human beings can also break into them.

    So yeah, as I said, it's a never-ending story. They figure out new encryption, algorithms, then hackers find new ways to get around it. And it's going to last for some time. Quantum Computers of course promised great encryption capabilities. They said they would be unhackable, but since there's no... There are some use cases of, of operational functionality of Quantum Computers. They have great advantages over, let's say, regular, so to say, binary computers. For some tasks to other tasks, for example, they're completely uncomparable. So yeah, it all depends on technology here and how fast the progress will go.

Saso Papp: Robert?

Robert Kordić: Yeah. When you ask question, if it's worth it, of course it's worth it. From a legislative point of view, we are at the start, but probably headed in the right direction. And what Matej maybe failed to mention is that a product like theirs cannot only help but also save life. Mat and I had the opportunity to test their simulator and the information that the simulator gathers, the amount of data and for what it can be used is almost limitless. And if you can use it also to detect various illnesses and cure persons from various diseases, then yes, the use of data becomes every day more and more important, every day more and more vital for us. And we only need to follow up with it in a proper way. So I don't see any Terminator scenario in the near future, I only see the opportunity.

Saso Papp: Great. Matej, do you want to add something?

Matej Vengust: Yeah. So yeah, there's no Skynet yet. Not like the one in the movies at least, but yeah. As you mentioned, yeah, maybe I didn't explain enough what we are actually into with the data we gather. So the point of gathering data while the body is active, the brains... The whole body, it's very crucial actually, because when they put you into MRI machine or a CT scan machine, you're basically in some type of stasis, right? Maybe if you watch some movie to get reaction from you, but it's completely different if you are driving and practically your whole body, basically whole body is involved into this operation and subconscience comes out. So you're doing the, let's say, reactions and trying to avoid maybe hitting a kid.

    Even in simulation on a subconscious level, it's not that you have even time to think, "Okay, I see a kid, then going to do that and I'm going to do that." No. You do multiple things at one time. So vision, hearing all four of your limbs are involved into that process. And from this data, this data is much, much... It tells you much more about the current psychological and, let's say, physical state of a person than just lying in an MRI machine because your body's completely active and it's under stress and this is what we want to provoke. And then you can actually see we use really high end eye tracker from the pupil size, or let's say pupillometry, we can predict diabetes in certain situation. For example, we can go into details of how a certain person solves a certain situation for early signs of Alzheimer's or even Parkinson's disease. So there are really, as you said, limitless possibilities of what this data can mean to us, not just to us, obviously, for medicine.

    And we are partnered with Stanford University, for example, and the head of the Big Data Science Lab said to us guys, "You just collect as much as data as you can, even if you don't know what are you collecting. Just collect it. If it's EEG data, if it's ECG, if it's even gastro sensors for kind of sensing your kind of... If you're feeling well or you are upset or something like this, then of course, skin perspiration, then obviously skin temperature. Everything, you can just record it, there will be a day when machine can tell you even some things we don't want to know, but yeah, it will be able to tell you." And when it comes to early signs of diseases, there's this machine, which you can see in the back, it has great potential in the future. And we developed it with help of doctors, with physicians, with the University of Ljubljana, the Faculty of Arts, Department of Psychology for example, and of course driving instructors. And it all comes together as a whole set of new data sets we can gather and help medical industries.

Saso Papp: It's so great that you can see how much of big, big, big, big data that you gather, you can then put together and find new stuff from it about the person, right?

Matej Vengust: Yeah.

Saso Papp: It's amazing.

Matej Vengust: And you need AI for that.

Saso Papp: Yeah, yeah.

Matej Vengust: [crosstalk] these raw data, for example, one drive can take up to 20 gigs of data and you would just put a person behind it and I'll tell him-
Saso Papp: Maybe Excel sheet.

Matej Vengust: ... what's my score, what's my evaluation, what's my diagnosis? Yeah, it's pretty hard work. I think two months wouldn't be enough for one person. [crosstalk] it happens in what? Two, three minutes, yeah.

Saso Papp: Matej and Robert, big, big thank you for your time and big thanks to all of you watching or listening. Until next time, stay safe, drive safe and enjoy life.

Matej Vengust: Thank you very much.

Robert Kordić: Thank you.