Perimeter Medical Imaging AI, Inc. (PINK) Earnings Call Transcript & Summary

Perimeter Medical Imaging AI is a medical technology company driven to transform cancer surgery with ultrahigh resolution, real-time advanced imaging tools to address areas of high unmet medical needs. I'm very pleased to welcome CEO, Adrian Mendes and Perimeter Board member and Anantha Kancherla to discuss the company's leading-edge AI technology and medical imaging systems. Both speakers have extensive world-class experience in developing leading artificial intelligence systems at some of the world's top tech companies. Today, we're going to be discussing the role of AI in health care and specifically how Perimeter is developing its proprietary OCT Imaging Technology, coupled with AI, with the aim of improving patient outcomes and reducing health care costs. Gentlemen, thank you very much for joining us. We're going to start off with a short presentation from Adrian giving an overview of perimeter. Adrian?

Thank you, Martin. Yes, we'll go through a few slides here to introduce you to Perimeter, introduce you to Anantha and myself. Just a reminder that we may make some forward-looking statements throughout this presentation. And so to me -- so I joined Perimeter about 9 months ago as the Chief Executive Officer here. After a 25 year -- I've had about 25 years in technology down here in Silicon Valley. I started my career at a company called Cypress Semiconductor, where I managed various functions. And just recently, most recently before I joined Perimeter, I was a Chief Operating Officer at an AI hardware and solution startup called Groq Incorporated where I was -- where I worked for about 7 years. At that company, we grew that company from -- it was a start-up. There were a handful of engineers and employees there when I joined. And by the time I had left earlier in 2023, we have grown that company to about 250 people. We have raised a significant amount of money. We had hit a $1 billion valuation on our latest raise, and we had a product in market and we were working on second-generation market. And so when I was introduced to Perimeter with this opportunity, I was extremely excited because this would be an opportunity to take what I had learned over the past 7 years in the AI industry and then apply it in a brand-new area in a place that's much closer to the end users in a medical field, which allows me to take what I've learned with AI and then bring it into a whole different area of the economy and of the industry that was helpful then to a whole brand new set of users. And so that's my story. And with that, let me hand it over to Anantha to give a little bit of background of his work experience.

Hi, everyone. Great to talk to you and Adrian and Martin as well. So my story starts like in India, like I went to IIT Kharagpur where -- actually I worked on the last generation of AI. And when I came to the States to pursue for the studies, I was taught that AI is dead. And that was mostly because like the last wave of AI died, and we were in this what was called AI [ VIKTOR ] and so all of us started working on alternative things to do what we call IA, which is Intelligence Amplification. The idea was that people are pretty smart, so why not just build tools for them to like to lot more with what -- to make them even better. So I worked on graphics and ironically there is a connection with what I did. So I was at Microsoft work for many years with companies like NVIDIA and AMD, so now you know what's coming. So we worked on the GPU architectures back in the day. And today, like fast forward, this is really where -- this is the hardware on which like AI is happening. And along the way, somehow I reconnected back to the AI world. First, when I was the VP of Engineering leading the software at Lyft's Level 5 organization. We were building a robotaxi, which is the most and hardest problem that one can solve with AI. And so it was a lot of fun and a lot of learning, understanding the limits of where the technology is at. That's when I really reconnected back with AI. And then I went on to Meta where I led its AI platform team, so Meta has had -- Meta, as you know, like has many products, all of which use AI and whether it is the ranking in new speed or whether it is like the head-mounted display to Virtual and Augmented Reality all of them use AI at different scales. And the scale is enormous. And I also got to like work with people who actually invented AI back in the day -- back in like early 2000s. And very, very recently, I joined General Motors, so to lead their ADAS organization. So ADAS is Automated Driving and Advanced Safety. So again, as you can imagine, we use every high-tech AI technique possible to make driving safe and keep people safe on the road.

Thanks, Anantha. Okay. Let me flip through about our company a little bit. So this right here is our mission. So you can read it here, but we envision a world where patients no longer experience the emotional and physical trauma up being called back for a second surgery due to cancer left behind. So this is obviously a big mission. But it's one that sort of drives the company, drives all our employees, drives our Board of Directors to help solve this problem. And we actually believe that we've got the technology to do that. The combination not just the AI, but really that overlaid on top of a somewhat novel imaging technology. And those two things together are what we believe is going to allow us to do this. And I think we're well on our way there. So this is our mission, and this is what we're working towards. This right here is a picture of -- on the left side of the screen is a picture of our product. So it's a cart device that sits inside the operating room. What you're seeing here is what is on market right now with FDA clearance. It's called the S-Series OCT device and what this uses is a near infrared light source that allows penetration of tissue to a couple of milliliter depth. But with that, what it helps surgeons do is visualize the microstructures that are within those first couple of millimeters. And why that's important is you can see with the OCT technology, this Optical Coherence Tomography technology, you can see the differences between different types of tissue, what's key to this technology at the OCT level is the resolution. And so as you can see written here, OCT with this near infrared light source has much higher resolution than both X-ray and MRI and the images that you see on the right-hand screen here kind of show the -- what the surgeon can see when they use OCT, which is the image on the top and then how that compares to what the pathologist sees later on, which is the image on the bottom. And really, the difference is that the pathology data comes back to the surgeon a week later after the patient has already gone home. Now when we have the device in the operating room and the surgeon can see this in real time, the patient is still there. So if they observe something that looks suspicious they can look at that and then make a clinical decision while the patient is still on the table to do what they call a shave, so what I've just described is our base technology, the OCT technology, nothing here has to do with AI. But really where the power comes in is when you can overlay some AI on top of that. So what you're seeing here on the picture on the right side is an image of our B-series device and this is what's in clinical trial right now, not FDA cleared. So we're going through the trial to get it cleared. We should be finished the trial later this year. And what the AI does is it has an Image Recognition algorithm that reads all those images that are taken. We've trained those images. We have a database, a very, very large database, over 2 million images that we've collected. And so we've trained our algorithm on that. And the algorithm can go through -- look at all those images and highlight those images that are most suspicious that there's potential for some cancer to be in the margin. That then allows a surgeon to flip through those -- a very small subset of images when they're looking, and it helps speed up the workflow. So think of it as sort of as a copilot or an assist for the surgeon as they're looking through the -- everything that was scanned to be able to really look at the dozen or so highlighted most suspicious areas. And so this helps speed up time in the operating room helps bring confidence to the surgeon that they're not missing something, helps improve the usability of this technology to many more surgeons and really helps really make this device takes to the next level of usefulness. Right now, this clinical trial is ongoing. Here are some of the hospitals that are using it. Dr. Alastair Thompson is our lead PI, a lead investigator working on this. And so I think this is -- we're very excited in the company for getting the results of this trial later this year. And as that trial progresses and if we're successful with what we're trying to do there, this will allow us really to bring this AI algorithm into the marketplace and then provide it to our surgeons to allow them to benefit their care, their patient care for the patients they have coming through their operating room. And that's it for the presentation part of this section. So hopefully, that gives you some understanding of what we're doing from a base technology and then how the AI is helping our customers use that technology even more.

All right. Both of you have exceptional Tech and AI resumes. You've got -- you went over your histories, but with this kind of resume in this kind of market for AI, you could have chosen pretty much to go anywhere and to lend your expertise anywhere. What was it specifically about Perimeter that drew you to this company and this opportunity?

Yes. Great question, Mart. Maybe I'll start. I'll give my answer, and then I'll hand it over to Anantha. For me, when I was introduced to this company, my first reaction to it was that I don't have a lot of experience in the medical devices space and Medical Imaging. But as I -- but I do have a significant experience in the AI side of the house for my last 7 years at Groq. And really when I started at that company, AI was just starting to find its footing. And so I've seen the industry develop over the last 6 or 7 years. And through that, what I've observed is that when I started at Groq, really, there were very few companies that were taking advantage of AI and it was really the high-tech companies, the Google, Facebook, Netflixs of the world, Microsoft. But as the technologies matured, as the hardware has become better as the tools enabling folks, engineers really to develop have become better. What I was getting more and more excited about over the years was the ability for this to get propagated beyond a small set of companies, high-tech companies, and so this was already, this AI in other industries was already sort of on my mind. And so when I got introduced at Perimeter, I was primed in that way. that I think the next phase of AI evolution is going to be able to bring this to kind of the everyday people outside of the tech world. And then specifically with Perimeter, what I was very attracted to was the fact that there is an Imaging Technology called OCT that Perimeter uses that's actually in widespread usage for --not for the application we're using, but mostly for ophthalmology and some cardio -- some vascular applications. But this was the opportunity to take an imaging modality and apply to a brand new area, number one. Number two, the company had a lot of patents around some unique applications, unique ways of using OCT that allows us to really carve out a niche for ourselves. That's fairly well protected. And then the third thing was the fact that the company had already built up a pretty large data library, a data set which for me, in my experience, is very attractive because that's actually where the value you see for AI comes in. The ability to have proprietary data sets, the ability to train models and then to use that in a way that is very hard for other companies to replicate. And I think the last thing I'll just kind of say before I hand to Anantha, was the team was very strong. So Anantha was on the board before I joined. It's very impressive with Anantha's resume. I looked at the team internally at the company and seen what their resumes were, what they've been capable of producing both here and in other places. And may be very excited about joining a world-class AI team layered on top of a world-class medical imaging team.

I'll go next. Yes, I joined Perimeter like maybe like 9 months before Adrian did and well, like I've always been fascinated by this AI thing ever since I was in undergrad a few decades ago. And the -- in the -- in my early part of my career, it had just kind of faded away because the technology didn't live up to its promise. But when it came back in early 2000s, early 2010s actually, using the same stuff that I was working on with -- in the interim. It just kind of felt like an absolutely natural thing for me to gravitate towards it. And once I kind of started working on it, like I said, I worked on a really hard problem first, which is sefl-driving cars. That's where you deploy every possible imaging technique and try to understand what the world looks like, and you have to use AI to like really understand what the surroundings are and what the machine is able to see. So -- but that gave me like a view, but that was a really hard problem. But then it also gave me a view into what the modern AI is capable of. And like very similar to Adrian, I was also thinking in terms of, oh, wow, if we can do this, what else could we be doing? And how else could it be beneficial to mankind? And there are so many different places where we could be employing it like all the places where we really need help. And so I was thinking what the spaces where it could be done. And like health care, climate, there's a number of different places. And all of them are like very, very worthy of like the -- to benefit from AI. So when this connection with Perimeter happened, and I saw like this is something that can genuinely help improve people's lives and that to people who are -- who badly need help, it was a no-brainer for me to say, "Okay, how can I help this company? How can I help them like to use AI to transform?" So that's basically what drew me to it and it did help that like there were very smart people who knew what they were doing. This OCT team was brand new and being familiar with like LiDARs and cameras from my self-driving car world, just kind of felt like, oh, yes, this is just yet another modality of imaging. And AI, should easily be applicable to that. And it did help that they have a pretty Perimeter has like a pretty good proprietary data set that could be leveraged to build some pretty cutting-edge applications here. So that's basically why I can to help Perimeter.

All right. Obviously, with -- especially Adrian bringing on someone with your AI expertise into the leadership role at Perimeter, the AI Image Recognition is a key aspect of the company's technological path going forward. The key benefits from AI with the imaging system, can you elaborate that, I guess, with -- as you said, the -- your the surgeon is trying to identify when all the cancer is out of the patients, so they don't have to cut out too much cancer and that they also don't have to worry about having to go back and do another surgery. So that reading of the image is very important. And so AI is being used here to augment the surgeon. So I guess, could you -- the benefits would be, a, to let the surgeon make better decisions where the AI helps them identify what is and what is not tumorous or cancerous tissue in there? And then, I guess, to speed up the whole process, can you elaborate how the AI benefits of surgeon and ultimately benefits the patient, most importantly.

Yes, absolutely. So I think let's start with focusing on the image recognition piece, which we've talked mostly about. If you think about sort of the situation of surgeons in, it's a very high stressful situation. They need to understand what's going on inside the operating room in many different dimensions, and they've got a patient that's under -- on the table. So there's a lot of things buying for the surgeon's attention. So now there's this technology that helps them make sure exactly what you said. They're not cutting out too much tissue, but they are getting all of the cancer out, which is exactly what the patient wants. So the easier we can make that for the surgeon, the better, What the AI does so it does a couple of things. One is it gives that surgeon confidence that they are seeing everything that has been imaged and paying the most attention to the parts of the image that is the most suspicious. So if you think about it, when you image a tumor, there's going to be hundreds and hundreds of images that are being captured from the OCT device from our device. Well, if the surgeon has to look through all of them, if they can, they can flip through a kind of like a movie. They can flip through the whole volume, but it takes that amount of time to kind of concentrate and make sure that they're catching everything. From their standpoint, it would be great if they had and assist saying, "Hey, look, here, look there, look in this other place because these are the areas which is most likely to be the place that there might be some cancer in the margins. Great. What does that do for the surgeon? It takes a mental load off of them or reduce the mental load while they're in this -- in the operating room. So that's number one, makes that job easier. Number two, it helps speed up. So if you have to -- if the surgeon has to go through the entire volume, that's going to take a certain amount of time. If they have to zoom in to, let's say, 12 images only, Well, that's going to be much quicker. And for them, they're really trying to get the job done, get the patient closed and then send them off to recovery as quickly as possible. So these are the two main areas from an image recognition standpoint, this AI is going to help that surgeon. I think there's an element of it also which plays to our market from a business standpoint, our market expansion where there's going to be some surgeons that are going to be more willing to adopt the technology once there is an assist and AI assist there that's going to help them do their job a little bit more easily. Get them more comfortable with it quicker. So for me, getting this next -- the B-series the Next Generation device, cleared by the FDA and onto market is going to help us with our market expansion quite significantly, I think. So I think that's it from the image recognition standpoint from the surgeon. There are other areas that we are using AI deeper inside the technology stack to help speed up the image capture to help improve the clarity of the images, the quality of the images they haven't really touched on, but our AI team is kind of very active all the way up and down through the stack.

One of the areas in AI has benefited in many of its applications across industries is -- and you hinted at this is the skill gap reduction, where when there is an excellent radiologist or a surgeon who has a lot of skill and practice and analyzing the images, they could do a better job at it where if there is a newer surgeon or a surgeon who doesn't go through has the same number of surgeries to go through, this can help increase their expertise where and their confidence as well. Is that one of the sort of improvements that the AI adds to the process?

Absolutely. So if you think about -- just take a thing about America, right? There's many people live close to a metropolitan center where there are surgeons that are -- that do hundreds of procedures hundreds of more procedures a year -- hundreds of lumpectomies a year. There's about 8,000 surgeons that are doing at least 1 or 2 lumpectomies a year. So if you live near Dallas or San Francisco and New York, Boston, you probably have access to those surgeons and those surgeons are going to be well practiced and going to be top of their game. But there's also a lot of people in this country that live in more rural areas where they don't have access to those types of high-volume surgeons. And so the surgeons that are helping those patients are not gonna have that type of volume. And so part of our goal is actually to be able to bring this tool out to those lower-volume surgeons that don't get the practice that don't -- that maybe aren't getting as many reps, so to speak, In lumpectomies and be able to bring these tools out to them to help them actually get the same results that the high-volume surgeons are getting. And so this AI really helps with that, where it helps allow some a surgeon that's only doing, let's say, 20, 30, 40 surgeries a year. Lumpectomies a year, get up to closer to the skill level in sort of assessing the margins as those surgeons are doing hundreds per year. And we see this as being a way to help that.

And the surgeons missing some cancer within the patient isn't a rare phenomenon. There is a significant [ re-excision ] or reoperation rate. Could you dig into that and what potential benefits your -- the technology you're doing can to improve that?

Yes, absolutely. So right now, for lumpectomies, the re-excision rate is on the order of 20%, so if you think about that, 200,000 lumpectomies a year in the United States, 20% of those patients have to come back again for re-excision a. So if you run the math on that, it's 40,000 re-excision surgeries per year on a situation where, if unfortunately, if the surgeon had been able to understand those margins, be able to see that in real time, that's 40,000 surgeries, that would not have to happen. 40,000 women that don't have to get that phone call a week later and say and hear that, "Oh, I'm sorry, we thought we got it all out, but we didn't. 40,000 phone call surgeons don't need to make. So it's a pretty big deal. Now our goal is to try to reduce that and try to reduce that significantly. And so we're starting to see some very interesting results from the surgeons who have this have our machines in the field.

Is well many women get a full mastectomy, where the entire breast is removed for fear that some of the cancerous tissue is left behind. And could metastasize and spread to other parts of the body? Does it have an effort to both the surgeon and the patient that the full mastectomy is not required and that's just a less invasive or less intrusive surgery is a viable and safe option?

Yes. Yes, you can imagine yourself in the situation of being a person being told that you've got cancer. And the only thing you want is to make sure that you don't have cancer after everything is said and done. And so there -- there is a portion of women that will say, "You know what, just let's do the mastectomy. And that way, we can have the highest chance of not having a recurrence. " Now of course, if you speak to women that have gone through this, the vast majority of them actually want to preserve as much of their breast as possible. And so they choose lumpectomies but they do have to face that trade-off of 20% of the time that there is a re-excision. And so that's a very, very hard decision for a woman to make. And so our goal is to try to help reduce that probability of having a re-excision to help make that decision of the patient and their doctor much easier. And the other thing I do want to highlight, Martin, is this technology is not just applicable to breast cancer. So this isn't just women who are facing breast cancer. This tech -- although we spent so much time talking about that, the same problem exists across pretty much all cancer types. So 20% is number for lumpectomies, but it's in the teens and higher across prostate, head and neck cancers, colon cancer. So this is something that everyone -- not everyone, but a lot of people will face throughout their lifetime. And the goal if we can help surgeons reduce our re-excision rates, it will really help a large portion of the population you have to face these decisions.

All right. reading the headlines of the world of AI and how it's evolved so rapidly over the past 18 months since ChatGPT first burst onto the seems what people have learned it's not just the algorithm is that you have to train the algorithm with good knowledge and good information. And you referred to that, you've got over 2 million images to train the -- your algorithms with and your whole tech stack with. Could you -- that seems pretty large and impressive. Could you just elaborate a bit on that, the significance of that database of knowledge, and I presume it's a growing database as you do additional surgeries where -- and your algorithms improve, that the accuracy and the quality of your decision-making improves.

Yes. Maybe I'll just start, and I think Anantha probably has something to say here as well. But we've collected data through, obviously, on tissue types, many, many different tissue types and both with cancer and healthy tissue. And so when I refer to 2 million, I'm actually just referring only to the breast cancer -- sorry, the breast tissue data we've collected. We've actually have more across other tissue types. And those images are created multiple ways. So every time we scan every time we scan a tumor, we actually have multiple images off of that. So that goes into the database. There are ways to augment the images that we have, so that adds to the database. And all of this just increases the intelligence of not only the algorithms we've got, but also opens up the doors for us to create even more sophisticated algorithms, we have fear of overfitting and things like that. So it's key to any type of AI, I think that's pretty common knowledge. And so we do have a very strong focus on trying to grow that database continuously.

Yes. And I want to add that like it's more than just simply the image recognition work that is happening at Perimeter. It is like the one that you mentioned to [ D-NOISE ] that's actually like a very clever and pretty cutting-edge application of AI. So it actually uses a model called the diffusion model. And so if you heard about generative AI these days, have you -- I don't know if you've had a chance to use something like Midjourney or something like that or DALL-E like they create images. So they use something called a technique called Stable Diffusion. So what we use here with Perimeter is very similar to that. Like, so instead of like generating new images, we generate clean images without the noise. And that's basically like how like the surgeon is able to really clearly see without actually like cranking up the resolution or taking double or four times the amount of time to scan. So everything that you're doing in a different part of the stacks is like very, very AI-related. And you mentioned also like data augmentation. There's so many other things that are that are going on inside Perimeter that push beyond the traditional image recognition side of AI.

Yes. And I think I'll just add to that, Martin, that this -- when we talk about the attraction for me to the company and then what keeps me excited about working here is that the well, both the AI team and then that we have and then we continue to grow as well as the different ways we can use it to help the business it's sort of multidimensional. It isn't just an image recognition algorithm and we've got a teams to totally focused on that. Actually, the team is able to and then they continuously do look in the world of development, AI world of what's being developed out there, from an algorithmic standpoint and a technique standpoint, and then thinking super creatively about how to use that across all elements of the business. So when you got a tech -- and I've observed this, since I've been involved in AI since 2016, is you sort of -- it's kind of this green field right now for brand-new tool set AI and brand new technology continuously getting better. And what limits -- sort of what the limit is right now is the creativity of the team of how to apply these techniques and tools to various parts of the business. And this group that was built even before I joined here, and that will continue to evolve. It's just very, very good at that. both on the classic sort of image recognition piece of it, which is the most easily observable part of what we do, but even under the hood in many, many different very interesting ways, like Anantha alluded to when he spoke.

You've talked about the different levels within your AI stack. And you have a clear objective and purpose with your -- the current technology, but it sounds like it could be used for other sorts of images, where you're cleaning up the image and identifying it. Is it limited this technology or developing to the OCT platform? Or are there other modalities like x-rays or MRIs or I don't know whatever other types of imaging sources that this technology could be used for as well?

Yes, it's absolutely not limited to the OCT technology, Imaging Modality. So -- and it's an interesting question because if you think about it, cancer in the whole sort of treatment and diagnostic treatment of cancer. There's lots of different images that are taken. There's X-ray, mammograms or a type of X-ray. So it's x-ray imaging, there's OCT imaging, there could be different modalities, ultrasound or MRI. All of these put together would make for a very interesting data set for -- to layer AI on top of -- so what we're very, very careful of here is as we think about how we evolve the technology, not just making sure we don't back ourselves into a corner where it's really only really good for OCT, really making sure we've got pathways open that allow us to take different data sets in various libraries and then train models that are multimodal. And I think there's a lot of power you can get in terms of helping surgeons, helping patients in their diagnosis and treatment, if you can pull some of this together. And -- so...

Yes maybe can I also add an analogy here, like, so I mean, my background is like working on autonomous vehicles. And the way these vehicles work is that they use like different type of imaging technologies. They use radar, they use LiDAR, they use cameras. And they see the same object with these multiple different modalities. And what AI does is that it can even learn if some things are easier to see in an image, and then you know it's the same thing that is seen in the radar but in the night your camera might not be working as well and then maybe you can identify it with the radar. So you can actually transfer the learning from one to another. And you see some similar things like when these models are able to understand a structure of a language, whether it is French, German or Spanish or English and it is able to like figure things out based on that. So this technology is incredibly powerful. So it just process cuts across -- it's trying to really understand the underlying patterns and so it goes deeper than just simply the imaging modality that is there. So the potential is like vast and to be able to make a multimodal system that is able to learn from one modality to another and then take it back and try to interpret things in a much more deeper way.

I would think also we've -- many of us have seen the AI-generated videos now and generated images of pads or whatever. And when you look at them, there's always some glitches or something you look at and say, "Oh, there's something a little off here. You're operating in a environment that's FDA regulated, where you can't have to sort of random bad facts thrown in to your images or your analysis. So you're working with the surgeon itself. So there are sort of two sets of eyes on the image. I would think that developing a model under a highly regulated and a sort of limited -- you're not allowed to make mistakes, so to speak, kind of like the car driving situation. You can't blow through a red light. I would think -- could you just discuss a little bit how under sort of a high risk or a zero-tolerance environment how the AI process is different than if you can be sloppy in making cat images or videos?

Yes. Anantha, why don't you take that? Because you've had some experience in it.

Yes. So this is a great question. And I think this is why I'm like drawn to Perimeter because it's so municipalities with like the kind of work that I do because you cannot make a mistake when you're putting a car on the road. Similarly, you should not make a mistake or cannot make a mistake in diagnosis or health care. So it's very, very similar. But I think the way these things work is that in actually most fields, AI is not yet there at a point where it is a human level, which means it cannot do like everything that a human does yet. It will maybe down the road, but that's like in the future. But today, the best way to apply AI is as a copilot. So it is basically working hand-in-hand with the human. And it is actually like taking the people much more efficient and faster. And where it works very well is that like people tend to lose focus, and we don't have the level of energy or speed to look through a huge amount of data and data sets. And what AI is able to do is it's able to narrow things -- it can actually do those things very well. But where it can't do very well is like deal with things that it has never seen before. And this is where humans are very good at and so this combination of putting an AI and a human together, like in the case of Perimeter, we are doing surgeon and the device together actually is the right way of doing it. And this is where -- like you would see the most success in applying AI. Adrian, do you wanna add anything?

Yes. No, I think that's it. We're very careful about ensuring that part of the reason of having such a big imagefd libraries to be able to train a model to a very high degree of accuracy. So the latest one, we just published a paper demonstrated over 98% accuracy. And there is an element of it where you don't want it making stuff up, of course. The good thing in this -- like in our application to Anantha's point is that it operates as a copilot, so from a business standpoint, if we don't have a high degree of accuracy, the customers are not going to adopt it. So before it even has the ability to create a problem for a patient we need to meet an even higher bar, which is the surgeons or else they're not going to adopt it. And the good news is we are getting the adoption on the device before the AI is available on the S-series and so we're seeing great market traction on that front. And through the clinical trial, we haven't seen the results of that yet. But we do know that the AI itself is up above 98%. And so when we do bring that to market upon FDA approval, we have high confidence it'll get adopted. And if it doesn't, we will know what we need to work on. And that's sort of the quality filter, both between ourselves and the surgeon to make sure that we don't impact patient care.

And I would imagine AI is new for the FDA as well, and they're trying to wrap their brains around it, how best to apply it to increase the skill level of the surgeon and the efficiency, which can have better outcomes and lower costs and greater efficiencies throughout the whole health care system. And I would think just you're on the cutting edge of this FDA approval system as well, I presume that the knowledge you're building from that is applicable to many new modes or modalities as you've discussed as well and that you're creating this real knowledge base of not just the AI, but how to apply it to real world life and death, literally life and death situations.

Yes, that's right. And we maintain a good relationship. We have a good relation with the FDA, working through this trial, especially. And they're learning how to regulate technology like this. I think it's very important what they do it is brand-new technology. You want to be able to bring to bear on -- for patient care, all the best newest technology, but you don't want to do it in a way that's sloppy. And so I think the FDA is doing a pretty good job balancing these two. And they know what they know, and they know what we're still learning as an industry that we have to go through these cycles. But like every other regulated industry of new technology, we'll figure it out. And we understand this. I think the FDA has been very has been a great partner with us through this. And so we'll continue working through it with them.

Excellent. We've covered a lot of topics here, gentlemen. I really appreciate your time. Is there anything you want to highlight or emphasize or anything we missed in this discussion before we wrap things up?

Actually, I want to add -- go back to the point that we were talking about earlier, like about AI being able to replicate the best I think that is the attraction for me with AI. Like, so for example, like when -- so the style of AI that is used is called Supervised Learning, so where like we get a bunch of data and then we put experts to go and label the data. Basically, we use experts to train the system, the AI itself. So better the experts that we bring in and more time that they're able to dedicate, better the system gets. So now that's the brilliance of the system. That means like you are able to bring the best expertise to every surgery. Regardless of like the experience of the surgeon. And this is true for like any application of AI where like you can collect the best data, and you will see that in the AI world, we always talk about it as like data matters the most. So better your data, better your AI performance. And so I actually think that this is actually really great, like we can being the best experience to the most people.

And that's best for everyone. Health care outcomes, economics, makes the world a little bit better. So that's great. Well, gentlemen, thank you very, it was fascinating. I've learned a lot. It's great the innovation that AI is taking to the world of health care because sometimes some of the things that the big AI models are doing, not sure if that really benefits anyone, but what you're doing is pretty clear that it can help a lot of people out there. So thank you very much for taking the time.

Thank you, Martin.

Thanks Martin, appreciate it.