It's really just remarkable to hear the talks from industry and from Dr. Anand Kumar about all the capabilities of technology. And I think it's really behooven upon us as physicians to figure out what is the interaction between technology and industry. This is a very important partnership and obviously we need to work together to bring forward progress in the field. And so as part of that, we've organized this last session on how industry and academics can partner. We've invited several of the leading companies in the field of AI and endoscopy to the table. Here, our panelists for this session will be, or our moderators, excuse me, will be again, Dr. Alessandro Repicci, Dr. Pratik Sharma, and Dr. Ulas Bagsy. You've heard previously from Dr. Repicci and Sharma. I'll introduce Dr. Bagsy. Then one of my major research partners in the field over the last seven or eight years, we met as part of a symposium on academic partnerships. Dr. Bagsy is a PhD in computer science and computer vision specifically, currently at Northwestern University and has substantial expertise in cross-sectional imaging as well as endoscopic imaging. So I'll turn it over to Dr. Sharma to introduce the session and our panelists. Thank you. Okay, thanks, Mike. And Mike will remain online as well as part of moderating this along with Alessandro and Ulas as well. And you may not be able to see Mike's video, but he's there. So it's my pleasure. I mean, just again to introduce the panel, we have Giovanni DiNapoli from Medtronic, Kurt Hein from Olympus, Andrew Ritter from Docbot, David Pierce from Boston Scientific, Michael Byrne from Satisfye Health, and Anthony Borelli from Fujifilm. So my welcome to all of you. And now I'm going to turn it over to Ulas to lead off initially because Ulas also I hear has ventured a little bit onto the dark side maybe. So perhaps from his previous days and his current role, he has a very good perspective on it. So Ulas, I'll let you lead off with some initial questions to our panelists here, and you can pick and choose who you want to either ask the question or perhaps we can get folks to answer the question in order. So over to you, Ulas. Sure, thank you very much for the introduction both to you and also Michael. It's very actually good to be here. I missed some of the talks. I was traveling internationally, but I catch this panel. I'm very happy to be here. So I want to actually talk about from the trustworthy side. So in GI, actually we have probably more data than radiologists, so we are more lucky. I also work myself in radiology side as well. So having more data, highly likely we may have more diverse data, but data inherently include bias. And many of these AI algorithms are trained on this bias data set. So I want to start the discussion that how our panelists see how should we trust this data and the trust the algorithm predictions. Can we just rely on like accuracy, specificity, sensitivity without having any trustworthy metrics like fairness and all other things. Michael Byrne, probably we can get started off with you and your comments since, I mean, you've been extremely successful in your initial ventures on that. So please. Thanks Pratik. Nice to see you today, Ulas. So I think one comment I read somewhere recently was that the AI machine learning can improve our human accuracy, but it can also exacerbate or amplify our human inaccuracy and bias. So of course it behooves everybody involved in this space to look at the current state. How have we trained the current commercial? We've got three or four huge companies here on this panel. And I feel like a little minnow here in the corner who've got commercially available, fantastic tools. And yet I think we all know, and I'm not expecting Gio or Kurt or Anthony to say so, but that the datasets definitely need to be expanded and we need to see these solutions working robustly because there are undoubtedly biases in any even currently commercially available tools. So I'll let everybody else speak, but to me, it's also around how we use these tools. For now, they are gonna be CDS or clinical decision support tools, which is probably right in that we can spot hopefully the errors, the issues early before they become maybe not standalone diagnostics, but certainly more pivotal and central to the decision-making. So we've got a ways to go to eliminate bias for sure. Yeah, I would agree with you, Mike, completely on this. And I think you said a critical thing there, which was that this is really a decision-making assist tool in its current form. And probably in the next couple of years, it'll stay that way, but access to data is critical. And that's gonna be the only way we can continually improve these tools because we know that in the next couple of years, it'll be more decision assistance, but in the next five years, it'll speed up to be much more. And I think that the world will look much different in five years than it will be in two. But the only way to do that is to make sure we continue to get good access to data. And bias is gonna be something to make sure that all companies check themselves on making sure we minimize any bias. To me, it's about real-world assistance too. It's real-world exposure and testing that will also help mitigate it, mitigate the bias in this. Again, chiming in here, Gio. First off, thank you for the invitation, Pratik, and to the ESG, very nice initiative and content throughout the day. I think a very important factor here is to prove the algorithms through robust clinical data and has to be done not only in one country, but multi-countries with different level of expertise in real world. I think that's the way, in my opinion, that you can remove bias out of what you collect. Okay, so maybe just a follow-up on that to the comments which have been made, like real-world we are talking about, which lends us into going outside of clinical trials at the tertiary center. So perhaps maybe, Anthony and David, starting off with you, maybe, I mean, not just for AI products. I know you have other products too. After they get tested in randomized clinical trials, which are mainly at the academic centers, what are your plans traditionally, if you look in the AI space, as to how well do they work in the community sites and the real-world data? How do you plan that in terms of your own company investing into those kinds of studies? Well, I'll go, Prateek, and I'll thank you and the group for inviting us to participate on this panel as well. It's interesting. I think it's evolving, especially with EUMDR on the horizon. And we talk to the FDA all the time and the demands for real-world data are only going to increase. So I think perhaps in the past, it wasn't as keen of a focus as it will have to be in the future for devices, as well as solutions like AI and other innovations that are on the horizon. And Dr. Sharma, I can maybe add to that as well. I think it's product development and clinical studies, I think are very closely tied together where it's critical that we are getting the feedback of community physicians, physicians from the ASC, as well as the tertiary centers, just to make sure that whatever we're bringing to market, whether, again, it's AI or new products, that it's meeting the needs of clinicians across that landscape. Ale or Mike, any other discussion, questions you want to lead off here? Oh, yes, Prateek. So nice comments from everybody. My comment is what we can expect as a physician from companies in terms of support to develop better AI programs. There are different ways that the company are using the past to support physician advancing science. So this can be at the level of societies, institutions, academic support. What is your plan for the future for all of you? I think, I mean, Dr. Rappicci, I think it's all of the above, right? I mean, if you look at the ASG's data science priorities, we as industry, I think, have to get behind those priorities and support the development of those priorities far as standardized methods and that type of thing. I also think that kind of one-off smaller type relationships to prove principle on smaller things has always been part of the ecosystem that we work within. And that typically can lead to bigger and more broad interactions. So I think it's kind of both ends of the spectrum, supporting the priorities of the ASGE to get this data up and off the ground and usable. But then also each individual company, I'm sure, has multiple relationships on different smaller initiatives. You know, I'd like to add, this is Andrew Ritter. I'm the CEO of DocBot as a matter of introduction. You know, for us, a big thing that we focus on and center on is exploring utility of use. How physicians can use this under what confines? How do we meet the needs of the physician with all the challenges they face? And to me, that's really the key is being able to deploy this properly in ways that meet the needs of the physicians. And there's a lot of needs that they have that this has to fill. Hi, this is Mike Wallace here. We have many of the largest companies involved in development of AI technologies here. From your perspective, historically, and history here is the last three or four years, all of the development has been with, you know, smaller groups of academic endoscopists, usually in a clinical trial setting. Going forward now, as we mature this field and we scale it up, what is the most effective model for academic industry partnership and community industry partnership to accelerate further development of new technologies and further implementation of recently developed technologies? You know, for example, is it creating a large data set? We've termed one called EndoNet. Is it developing large panels of data Is it developing large panels of experts? Like our radiology colleagues talked about 80 experts that we're scoring. Is it setting up competitions to compare systems head to head? Please let us know how we can partner as physicians and academics with industry to move this field forward. You know, I'd start off here by it'll be, we find it helpful if there was more universal benchmarks and measurements that as a group, we all follow or can improve on as ways to show consistency in our technologies, as well as ways to communicate to academics and GI physicians on the performance of these tools that could really help them. Any of the thoughts from our panelists on that? I think it's an important question about how else do you see from Mike Wallace about this relationship moving forward in the AI space? Any other thoughts? Yeah, you know, I wanna add one more thing and it's licensing. For companies like ours that are trying to develop tools to help clinicians, the data is a lot of it's housed academically and being able to share that data with companies is gonna be very important for us to be able to use that data to then develop the appropriate tools. That sharing is quite frankly, pretty challenging with academics today. And opening that up would make a huge difference in being able to move innovations forward. I think that's a great point, Andrew, that the licensing component of it could be an insurmountable hurdle if we don't figure out a different way. It's a hurdle in a lot of areas in particular with AI. And where the challenge comes in and I can appreciate academics concerns, it's the value of owning the data. And while there is value in the data, the technology is where the really value is being able to take the technology and bring it to market. And so there needs to be sharing in that added value, but it's taking that, the licensing arrangement, the data, the innovation and putting it together that's gonna ultimately create the greater value. If I may on Andrew's point there, I think the great thing, the great opportunity is that as Mike just said, history for this space is three or four years really in many ways. So we haven't got a huge amount, a decade of wasted data. We haven't really been collecting many video type data sets in endoscopy. Let's be honest about it. Until recently, it's still very haphazard and sporadic and piecemeal and hard drives in different units and people plugging in USB things to take videos off endoscopy tiles. Let's be honest, that's the current real state. It's changing thankfully, but it's moving forward now how to use those prospectively collected data sets, how to have patient permission on maybe on a nationwide basis, how to extract data from clinical trials in the drug space. I'm embarrassed to say I missed most of David Rubin's talk. I was in the gym, but I know he talked about the utility and clinical trials in IBD. And those data have been around for a long time and industry has worked out how to use those for drug discovery and end points. So why is it any different for AI in some regards? But I think in terms of the endoscopic data, the video data, we're at an inflection point now where, if we can quickly grasp how to do that properly for permissions and usability going forward, we haven't wasted too much of the last, we haven't wasted too much ground. I think I have a question for Michael Byrne. Michael, you are doing a great job. You are across the two sides. You are a physician, entrepreneur, scientist, and how you see the need for training for physician to get used to AI? Do we need a specific program for AI? What we need to do? Yeah, Ali, very good question. I don't think clearly everybody who's using these tools doesn't need to be an expert in AI. One of the catchphrases that many of us use is that these AI solutions in endoscopy will put the expert on your shoulder. So do they really need to know much about how the AI works? I think maybe you're talking about the practical use of those tools. Hopefully, most of us are building user interfaces that are pretty straightforward. And most people are fairly savvy these days with technology, and basic technology anyway, on-screen type stuff. So I don't think it's a huge challenge to have most people use it. In terms of how it affects training, it came up earlier. Tyler Burson made a very good point around the pros and the cons of having that on-screen assistance and tool all the time. Does it make you lazy? Does it make you more aware? I agree with what Tyler said. I think it just increases our attention to what's on the screen. And even if it's a detection tool, as with many of the industry groups on here, all have amazing CADI tools, when you're looking at a polyp and you see it, you're probably also doing an optical biopsy by default. So I think it's only helping with training. But I don't think there are too many barriers to the average physician using these tools. They're fairly straightforward, even as is, and they're going to get better, clearly. Another question to all of you is, what about medical legal issues? It's the physician, it's the company, it's the algorithm, it's both of them, just in case. Who is in charge of the final responsibility? I think that's a great question. I think, as I consider that, as these technologies evolve, they really have to become integrated into the guidelines. And I think that creates a safe space, if you will, if they're implemented into the guidelines and well understood, well studied, proven to be effective. I think that gives you a safe space to operate within. I think outside of the guidelines, using them to make specific decisions and not using something that's in the guidelines is going to be risky. Yeah, I agree with David. And I think some of it's just time. Time will create more credibility, more awareness, more stability for programs like these. Yeah, I also agree. I think the critical piece here is that, is to understand that this is a decision assist tool. It's pointing out things. It's not dictating pathology or saying it's a red light, green light type tool. It's more built to give assistance so the physician can make a more informed decision based on that, not to take the decision making capability away from the physician. So it still lies within the physician, at least in the early stages of AI until it becomes into the guidelines and gets more adopted and can be more to the point where it's proven, it's well studied and proven. But in the foreseeable future, this is gonna have to be strictly an assist tool without dictating pathology for a physician. Ali, can I challenge a little bit? I agree with everything that Kurt just said entirely, as is always the case. But I would challenge a little bit to say that, this is the kind of conversation that we've had now for a year or two or three, which is when can we use them and when can we use it beyond, let's say a decision support assistance or what have you. When can we think about it being not maybe a standalone tool, but a lot more pivotal to the decision? So again, I missed most of David's talk. I've recorded it, but central reading in clinical trials for colitis, for example, that is an offline situation where physicians sit at home at night, dare I say, with a glass of wine and read two or three videos. And there's a great shortage of skilled physicians globally to do those reads. And they usually need two reads. Can you foresee a situation in the next 12 months where you have man and machine in perfect harmony? You have one physician, one machine, acting as a double reader and the machine maybe even abbreviating the video read for the physician. So not only giving its own answer, but also curtailing the video a bit like the capsule endoscopy software that's available. You know, maybe that is a situation where in the next 12 months or 18 months, we can have that. And that's much more away from decision support to almost co-decision or joint decision. That's where I see, that's where I, on space, I see it moving quickly so that we're not having the same conversation in five years, which is, well, it's a nice to have, but we'll see where it goes in the next 10 years. Removing all of my bias. Okay, good points, guys. So one of the things the panelists, unfortunately, can't put in some questions. So Brett's just asked this, which I think is, again, getting comments maybe from Joe and Anthony on this, is that how can working together is how can sort of you help us help you, which is help accelerate, for example, some of the innovation processes for video databases, such as real-time labeling, et cetera. I mean, how do you see that happening? Is that an area of potential partnership on the annotation side? Yeah, it's a good point. Or anyone. It's not sure about Anthony. Go ahead, Anthony. So thanks, Kurt. So very simply, you guys will be instrumental in helping us to educate the algorithms. I think there has been some work done, Fuji is a Japanese company, a fair amount of work done in Tokyo and in other regions to help that initial phase of that algorithm. But I think that continual education with different disease states and different characteristics is going to be critical. This is where I think we have to work together. It's literally a team effort in developing the algorithms. Companies can't do it alone. Yeah, and if I can add one thing, if I understood, there's the imaging piece and the ability to detect and characterize what you're seeing. And then there's the reporting, the real-time reporting while you're in, and more of a AI. If this was, this is a question around how do we improve the reporting for a physician or how do we improve the annotation piece of the puzzle? I mean, the AI gets bigger than just the imaging, right? So it goes, it's what you see and then it takes it to reporting. And I think for some of these companies, and Olympus is one where historically it's product introduction to product introduction and moving more into the ecosystem approach, the software approach of agility and being able to change and do things in a much more agile way. It is critical to be arm-in-arm with the physician and societies in some of these different areas that there are a need. And gone are the days where it can take eight years for things to come out. It has to be more agile in the approach and more of a teamwork approach. And I think all the companies that are on here, big or small, would adopt that philosophy or have adopted that philosophy. I guess one question that I have, I hate to answer a question with a question and I know we heard from some radiology folks earlier in the session. Is there a lesson there? Is there a model there that we should emulate? Well, I'll, David, take that. I mean, from the society perspective, and this is sort of like an ongoing conversation that Mike, myself, Sravanti, Tyler, and Seth, I mean, within the task force, and then along with the ASG leadership we've had is this concept of having end on that type of image library, you know? So it's an unbiased data set which could be used for benchmarking, has videos, images, you know, from all different companies, all different processors, all different types of light sources, whatever you can imagine, you know? Because one of the questions we've been toying with all day long is that, you know, are these results generalizable or not? I mean, one question was, can you use, you know, a device which was built on X endoscope used for a Y endoscope company or not, right? So that's one of the things that, you know, we are trying to figure out how best to develop that. And, you know, so I think that's in the works. So that's one lesson learned, I think, from radiology. I think they've struggled with it over the years. They had a hard time getting up even initially with their pneumonia use cases, for example, which you might think, hey, everybody gets pneumonia, it's easy to do it. But just like with endoscopy, getting high quality endoscopic videos is difficult when you start doing it for this process. And then of course we have the annotation problem with it. So that's one thing. The second is these challenges, and you can do these challenges, this gamification, once you have such a well-defined, unbiased, and annotated, you know, video set. So to me, at least those are two or three easy lessons that we've learned from radiology. And in fact, we've had, you know, one of the radiologists join us on our task force call to tell us about how they've done. And then of course, requiring the infrastructure. I mean, it's just not easy to do it. I mean, you need to have a big infrastructure for a digital library to work and stuff. So it's, you know, we are at the initial stages, but I think we'll get there. You know, talking about radiology for a sec too, on the regulatory front, the FDA and regulatory is so much further along in understanding radiology with AI, and in GI, the GI division, obviously, is still learning. And so I think we have a long way to go from a corporation standpoint, as well as an academic standpoint of educating the FDA and bringing them along with the development of this field. Right. No, absolutely, Andrew. And I think that's, you know, critical. And for all our summits, I mean, you'll see that the FDA is always well-represented. Mike and I had actually written an initial letter to the FDA, you know, just offering the expertise of the ASGE and, you know, the experts within it, and the task force to help them along in understanding this, because you're absolutely right. This is new. I mean, if it's new for us, we talk about three years history. I mean, it's brand new for them, you know, and I think they're ready to admit it. And you heard some comments today, which, you know, you can see that they're a little bit more clear thinking. Again, you know, it may not go well with what we are thinking, but at least they're better off today than they were last year, for example, when we were discussing this. Ulas, I'm going to turn it back over to you for any questions there. Again, I mean, some of the things you may have learned in setting up your own enterprise, I mean, how do you see this relationship working? Yeah. Following up for your question, actually, I would like to also mention, since I actually missed a few talks, maybe it was mentioned. So in radiology now, one of the also trend, especially in other fields of AI is also to overcome the difficulty of sharing the data through like federated learning kind of algorithms. In GI worlds, our panelists, do you see in the near future, like federated learning algorithm can be used to share data without really sharing the data, but just parameters having better models, having, like Fatih said, it's more generalizable models. Do you see this can also happen in GI? And this will also probably avoid other issues like the data privacy and other things can be also avoided. So what do you think about this? In radiology, I think it is getting better and people already start using federated learning to share large scale data from different centers without sharing the data, having much better generalizable models. You know, in many ways, I feel like this is uncharted territory for us a little bit. You know, every day, every discussion Docbot has with academics or other industry partners, this always topic becomes almost like a brainstorming effect of how we want to handle it. There's no real uniform way. It feels like it's a preference or it's a case-by-case scenario. I'm curious what the others think. Mike Wallace here. So one potential solution here, we've got two competing issues. One is these, I think we all would like a large federated data set that we could test algorithms against. And at the same time, the companies have tremendous amount of investment in their own intellectual property that they don't necessarily want to share. But a common point of interest may be, you know, developing these large license-free data sets with some trade-off where industry might gain access to those to compare their own algorithms, but in exchange, there's some willingness to share the details of their algorithms. So, you know, again, I'm trying to explore how we can move the field forward. I'm curious as to whether, you know, this would be an appealing arrangement to industry to, you know, essentially in exchange for having access to these large standardized data sets, they would be willing to share some component of their own algorithms to further improve them. You know, I can say from the Docbot side, we're always looking at ways to collaborate and share our technology and key pieces of our technological know-how with academics because it is a shared environment. And so I think having a shared data set of some sort would make a huge value for the whole industry, especially actually, quite frankly, with the FDA, if there's a set database that we all could go to from a benchmarking standpoint, that could make a huge difference on the regulatory front, which can make ultimately a huge difference when it comes to commercializing and supporting GI practices across the board. Andrew, I agree with that, but I wonder that Gio and Kurt and Andrew and Anthony being very quiet here, because I wonder from a, you know, we're not, this is no different than drug discovery and device manufacturing, novel, you know, cameras, et cetera, that everybody on this call at the industry level is clearly global experts in. This is not a philanthropic space. There's a lot of thought in the last two or three years because AI is software, that it should be all about collaboration and cooperation, which I absolutely agree is gonna move things forward. But there is also, again, it's not philanthropy, and I'm sure industry here, you know, is very protective of their own pathway and IP and data sets. So I don't know if it's a token gesture question to ask this or whether there's really an appetite to have a significant shared pathway from industry. Or whether it's just to have a data set to compare against, for example. You know, maybe it's a philosophy, but we have our IP and we protect our IP as a company, but we also recognize that it's a shared environment with academics. And so we, from a philosophy standpoint, wanna be more flexible with our key opinion leaders and the academics and work with them. And so maybe that's just a philosophical difference between some corporations. Yeah, I don't, I'm not gonna try to step over the line here on this one, but this, you know, I think that companies have shown and continue to show that there's benefit in collaborating with other organizations that have a competence and a capability in a space like AI to help advance it. And you had mentioned, Mike, that it's not a philanthropic piece, but we're all trying to advance medicine. So I think that those things will continue to happen. And, you know, collaborating with companies big and small. Where I think that this could help is to break down some of the walls, is if it's a society-driven exercise potentially, where we're trying to attain the same goal and the only way to get there is if we all sort of come together. I don't have a perfect answer for this one because we're all working on our own strategic plans and we're all working in our own pathway. And it's, and to, you know, to put all that down and say, all right, let's figure out how to do this together could potentially create harm than good. But I love the idea of it. I just don't know exactly how we would get there, but open to discussion. You know, I think it's a fine line, but, you know, I'm gonna say it. I think it's, you know, the academics, you know, Dr. Wallace, Dr. Sharma, they have a lot of data. And we just talked briefly a second ago about sharing their data with us. But as corporations, we need to give something back too. This is a, there needs to be some giving and some receiving. And so our philosophy at DocBot is if we can provide insights or some tools that could be helpful, we're gonna be more sharing. I think that the easiest give would be to support something like EndoNet, right? So you have that common library to work with. The gives after that become more complex to the points Kurt was making. Okay, thank you. Great discussion, guys. Now, there's a question from the audience and, you know, it's more specific about CADx and saying that besides finding polyps, you know, it would be extremely useful if you characterize them and save pathology costs. So, you know, it's obviously stating, you know, what you guys would be interested in, but any thoughts from industry on that as you are looking both into your CAD and CADx devices is what are your thoughts, you know, around that? Seems a pretty straightforward question and an answer, but would love to hear all your thoughts on it. And again, I mean, I see Shani's still here from the FDA, you know, so it's also a good opportunity for us to see, you know, where we are and what the thoughts are around this process as we move forward on characterization of polyps. I mean, I'll make two quick comments for that. I think that's nirvana, right? Because you're saving all the time, energy, and effort and costs associated with pathology if you can get there. I think you really wanna make sure you have a really high specificity. So somehow prove out your specificity is very high so you don't have a lot of false positives. But I think, you know, we haven't really talked about it in this session. I think one of the big advantages of ultimate adoption of AI is workflow optimization. And as we continue to have resource challenges with a number of clinicians capable of doing the procedures, AI has gotta help increase efficiency so fewer doctors can do more procedures. I think, Prateek, I think David made a key point around, you know, the need for high specificity and sensitivity with, let's say, I presume he was talking about CADx tools. So to me, CADe, and I don't want to incur the wrath of any of the major industry players on the call here, but the CADe in some ways for polyps is the Trojan horse to get this into practice. And that in itself can be now and remain a decision support tool that helps us improve our performance. You may see what the machine sees. You may miss a few things or disregard what it tells you to look at. Whereas CADx needs to be black and white, yes or no, cancer or not. You know, in Barrett's, in colon polyps, in larger polyps for ESD, in colitis or Crohn's activity level, for example. That to me is where it's gonna make a huge difference in terms of cost savings and delivery of care and accuracy. And where it needs to be at least a co-decision making device rather than a decision support for a lot longer. So CADx maybe started out several years ago and then CADe overtook it, but the CADe solutions I think have paved the way to have truly more diagnostic tools. Okay. Giovanni, Anthony, or Kurt, any comments? If not, we'll move to probably towards the end of this. I think we have about five more minutes. Just a quick end. Yeah, sure. Go ahead, Kurt. On what Mike was saying, because I agree that the CADx piece could support moving toward, you know, supporting the PIVI and resect and discard. The big hill that we have to climb here is that it is still reimbursed and it's still a tool, you know, that reimbursement is not in favor of the resect and discard initiative. We're gonna have to try to work together around how to get to a point where the resect and discard initiative by the utilization of CADx and being able to look at these polyps and making a decision that would save, help save the amount on pathology. It's a tricky road right now, but something that could certainly save a lot in healthcare. Okay, thanks, Kurt. So just the last question, which all of you will have to reveal your company secrets with this one, which is, you know, and again, I mean, whatever you can share, where do you see this going into the future? And you know, what are things perhaps that you can actually share with us that you guys are working on so that our audience can understand that AI and GI and endoscopy, you know, is that the CAD and CADx are just the start and what are we looking for? And this sort of like looking at Anima's beautiful presentation on robotic surgery. I mean, you know, you can easily see us training the next generation of endoscopists on POEM and ESD and stuff using those types of techniques and stuff. So I'm sure you guys have thought through it. So I'm gonna like towards the end, just put each one of you and have you answer that to do that. So just on my screen in random order, I see David appears first. So David, Dave, I'll start off with you and then towards the end is where do you see AI headed into the future? And what are some of your company's personal goals looking into it? So Dave. Thanks Prateek. And yeah, Dr. Anand Kumar's presentation was amazing and so eyeopening. So, I mean, we're looking at kind of three things right now and it's nothing that hasn't already been said, right? So one is tissue detection and characterization for diagnosis, right? The next is kind of critical structure identification and case support. So identifying the papilla, helping to navigate and cannulate. And the third I just mentioned is workflow optimization. I think that's gonna be huge, not only for GI but for all medicine going forward just because the demographics are just not in favor. So we've got to find ways to be more efficient. What we're doing as a company is a ton of work evaluating partners and really looking at where we have strengths and where we have areas of weakness and then trying to find partners to help augment where we have weakness. I will say we have two pilots going on right now. One in the detection and characterization space, another in the workflow optimization space each with a different partner. So actively involved in it, think it's super important and we'll continue to go down this path. Okay, thanks Dave. Next I see is Andrew. Yeah, so at DotBot we're an AI technology company focused on detection and diagnosis. We're looking to transform the practice of GI and what we see is primarily a focus on the upper and lower GI tract and focusing on the whole GI tract and that's really important. Pulp detection is something that I think we all see is we have it, we can do it. And it's really what else can we do to augment and give more empowerment to the physician to make more informed decisions as they do procedures with more objective data. And so we see AI here to stay in GI. It's probably one of the first time, well, it's probably first time in maybe a century or so that GI diagnosis overall is gonna be transformed because of AI. You're turning from more subjective physician using their own experience in the naked eye to make decisions to using a more objective, reliable tools using visual detection. And so the transformation is here and it's really exciting and we're looking to be excited to be a part of it. For DotBot we're focused on pulp detection, paroxysplasia, IBD and workflow improvements just to be specific. Okay, thanks. Anthony. Thank you. And just real quick, congratulations on another amazing summit. But so from a Fujifilm standpoint, I echo all the comments that have been made. And I think Kurt actually kicked it off earlier by saying the landscape today and five years from now is going to be dramatically different let alone 10 years from now. So when we look at the AI space, what we're looking at is polyp detection and characterization from a European standpoint where we have algorithms released. And then from a US standpoint, we have the incredible importance of the collaboration with the US physicians from an FDA standpoint, just really looking at what that 510K path is. And just very simply stated as far as the future, imaging is everything. And when we're looking at AI algorithms, I think optical information can be leveraged across the entire GI landscape. Great, Kurt. Sure, and I agree with the sentiment of the other companies and thanks for having us to this as well. This has been an unbelievable summit. There's so many things that can be said here, but I'll just narrow it down to the human eye. I think the optics and all of the platforms that are out there are so unbelievable and the human eye can only take in so much, especially after a full day. And the goal would be to reduce variability in operator capabilities. So in doing that, there's the whole imaging side, which at Olympus, we're working certainly on improving the assistance to the operator from that standpoint, from an imaging standpoint, but there's also the workflow management piece, autonomous reporting and helping the physician throughout the entire clinical workflow from patient in to patient out. So the world of AI is gonna encompass that whole ecosystem. And I go back to so many of these talks were so great, but Dr. Armstrong talked about the smart endosuite. And a lot of that resonated with the areas that we're looking at as an organization, because it's not just about the optics, it's not just about CADI and X, it's more about workflow management and reporting as well. So we're working on those. Okay, thanks, Kurt. Giovanni, you're next. You've been the leader in this field and working on this for a long time. Please tell us your thoughts. Yeah, so I think Medtronic all and all... Can you hear me? Yes. Yeah, Medtronic all, not only Medtronic, but AI committed already from the last two years on a company that wants to drive artificial intelligence across MedTech. The CEO, Jeff Martha, said multiple times in the last earning calls that this is the commitment. And clearly also the investment we have been making across the company have been towards AI companies. The portfolio we have as a multinational company in terms of AI is pretty broad already. And how do we leverage it also within GI is going to be a very interesting project for all of us. Robotics, and we have a company that we bought a couple of years ago, Neutrino, and digital surgery. We have a very robust partnership with Cosmo, very, very robust and long lasting. And also within our team in Israel and also in California, we have a people investment that really is willing to revolutionize endoscopy and also use our portfolio. So that's actually the beauty of this is that we are committed to bring AI across the Medtronic GI portfolio. So expect to see more to come and we are very committed, very excited. Smart people working on many different projects and also partnership across the globe with many of you and many societies because like has been said earlier, this is the key foundation of what we have been trying to do. Okay, thank you. And last but not least, Michael Byrne. I'll to go last in this question because there's lots of overlap. Of course, I share a lot of the vision and the plan and the strategic ideas that we have at Satisfye Health compared to what I've heard from big industry and from colleague companies like DocBot. Our aim is to use AI to deliver precision endoscopy and all that that encompasses. Of course, we're working, have worked in the colon polyp space with AI for GI as I think most of you know. We're doing a lot of work in the IBD space, Barrett's, working with some global data partners in the ESD space. Our aim is to really utilize the power of big data because it's not, we're all using the endoscopic images as the pivot point but there's plenty of value up and down the chain with all the various omics and other metadata that we can get from the EHR, et cetera. So how to incorporate that, as I'm sure everybody agrees. We've done several global data collaborations. We announced one last year with AIG and Hyderabad, huge hospital, which is great. So we're open to those. And we've done some academic collaborations. Helmut Messman, I think has gone offline now but Helmut and his group in Germany doing amazing work in Barrett's and now work very tightly with us at Satisfye. We announced that last week. So we're open to working with groups who are doing well at an academic level and need some help to bring this to practical use and commercialization. Okay, thank you. With that, I'm gonna turn it over to Alessandro to wrap up this session and thank our panelists and then move on to Mike for the final words. It was wonderful, really wonderful. And I feel there's a lot of support and commitment from all of them, from all companies. I think this is the perfect time to develop programs with them. And I think ESG is going to lead many of these programs altogether with companies. So thanks to all of you.

partnership

technology companies

medical industry

progress

session

trustworthiness of data

AI

real-world data

academic-industry partnerships