Okay, yeah, on to our last session for the day. And, you know, this is something that we have been alluding to all day. And a common theme which has emerged from most of the sessions as we are discussing is, you know, the role of education and training. We have john Cohen who is already joining. John, welcome. John Cohen is in New York, and he's an ASG Governing Board member and will be serving as one of the physician panelists. Another panelist we have is Chuck. So if you Chuck, if you want to come on stage, you already know Dr. Khan. We have Dr. Srinivasan, who's going to be representing the GI fellowship community. And Sachin is a fellow at KU in Kansas City. And then we have the majority of our industry partners again, joining us. We have Bill Huffnagle from CDX, John Temple from Endosoft, Andrew Barbarino, Fujifilm, Dustin Atkinson from Medtronic, Sean Huff from AIM, Peter Morton from Odin Vision, Frank Feliciato again, from Olympus, Brian Bannon. Brian, you're there from Boston. And Jonathan from Iterative Scope. So guys, a full house, find the chair. If you don't, you're unfortunately going to be standing. And then finally, not the least, I'll invite both Raj Keshwani and Sravanti Parasa to come up and moderate this session. Just I know there was a lot of discussion prior to this about, hey, what are the questions? What is this about? Is there anything, any specific preparation that's needed from our partners? And the answer is no. I mean, this is just a discussion around how do we move the needle forward? We've heard about AI all day long, but how do we get physicians right from, you know, fellows to people who are in practice almost about to retire versus those in their mid-career to sort of learn about this and how can we collaborate together as a society, as the ASGE and the industry to move this forward besides symposia or summits like this? So that's the key question here. So Raj and Sravanti, I'm going to turn it over to you. You guys can be up here standing and leading the discussion. And I will share the mic here. We have a lot of tough questions for you guys. So get your mics ready. First question from Raj. No, from Jonathan. Jonathan, do you have a question or should we shoot off? I'm ready. No, I have, I have lots of great thoughts from today and, but I'm eager. You start the questions and we can start there. So I think one of the title itself is a question right now for me is the title is the Roundtable Enhancing AI Education and Training and a Collaboration with Academia and Industry. So I'll start with the fellows. Of course, as a clinician in practice, I need to know how this AI thing works, but I want to know from such in your perspective, what would be your expectations from either the societies or from industry and what kind of collaborations would be nice? You know, you can ask, but we don't know if we can make it happen. Right. Thank you, Dr. Parasa and the ASGE committee for allowing me this opportunity. I think, I think we've been sort of talking about how to integrate AI into fellowship and that's kind of, it's, it's, it's heartening to see that, you know, there are some, some of us who are, who are lucky to be in programs where there's some exposure to AI and, you know, I've also talked to people who, who there hasn't been that much of an exposure and, and it's, it's, it's amazing how much research has, has come around and just to see AI the way it is in terms of other fields and in GI, I think, I think we're probably at the, at the helm of something. And for us to understand that is important. Just as, as Dr. Keswani was also mentioning, I think AI in some ways can be our training wheels. We also talked about how many of our attendings are watching our screens while we're doing our endoscopies, but AI could always be watching. So I think there's definitely some role. We also, we also talked about, you know, about GPS and how that works. And I think, you know, we use GPS five times to get to where we go and then we know where we're going. So we don't need the GPS as often. So AI can be something analogous like that. But I think, I think what we'd really like to see is in addition to sort of patient care, are, are, are the industries able to sort of open the doors to training in terms of one there is patient care, but also is there more that the industry can offer to start this process early on for trainees? And I think that I, that'd be something that I'd be curious to find out. I was going to just sort of maybe add to your question that you're starting with and maybe bring in Jonathan Cohen as well, just to, I can ask the question one of two ways, since we have so many panelists, I think I'll ask it to the panelists. So whoever wants to, to, to answer and then John can respond is what does industry looking for from societies, because everyone has to figure out how to focus their efforts on education. And I don't think it's fair to say just for the trainees, but for, for the general population you know, what sort of input are you looking from society and sort of conversely to Jonathan, you know, what are the, what are the priorities of, of ASGE in terms of education and what, what you think that the industry needs to focus on in terms of training first? Is it, is it the basics of AI? Is it sort of the issues of bias and ethics? Where, where do you think the education focus needs to be? So maybe I'll start with the panel and then sort of reverse it back to Jonathan and see if they answered correctly. Yeah, maybe I can take this. I'm Brian Bannon with Boston Scientific Endoscopy, the global director of marketing within our single use imaging business and spent some time in the US. So the, the, the training and education piece, you know, to me, it's almost you need to establish the goal with the end of the goal with the end in mind and describe what that learning journey might look like based on what learner you have in front of you. So I think there's an expectation from, you know, industry as well as the societies such as ASGE about describing what the common set of beliefs are agnostic from vendor to vendor. So we all have the same set of beliefs in terms of training education, but you had described a study early on today. It was a thousand patients, you know, 57% of those patients said they they wouldn't adopt or the biggest hurdle was the lack of experience or their, their, their not understanding of the AI in terms of what that, that, that offering was. And so if you take that lens, I would almost say it's a basic and fundamental what AI is and minimize the misconceptions as you go through that learning journey. But to me, this is almost a process of go slow to go fast. And we're at the foundation of something that we should build the criteria in terms of what is AI today versus what you want it to be in the future. So one perspective. Okay. One thing I was going to add to that is if you're going to talk about basics, something that might be super helpful is educating about things like confidence intervals and how to interpret those or how, how that should influence decision-making at the point of care. On top of that, I think some of the basic training that happens early on in fellowship can really help fellows and later attendings get used to making the most of AI, right. And establishing the conditions that are necessary for AI to perform its best in terms of the prep that is taken, the amount of time that is provided, how many times to use or basically consult the AI. If you would like to differentiate a particular lesion three times, two times, how do you interpret the confidence intervals when if it's just one time, how do you interpret multiple results and make sense of those? I think these are all kind of scientific decision-making that has to happen at the point of care and developing a sense for those things and developing a feel for those topics is important before it gets to that stage. Yeah. And I just wanted to add, you know, this is something that's been talked a lot about today's data. So Fujifilm, you know, our background is imaging and a lot of capital equipment. We've become good at training our customers and prospective customers on those types of products. Data, you know, being at the foundation of the decision-making with a CAD algorithm is quite a different process. So I think what would be helpful is understanding, especially from academia, because this education will then also trickle down into private practice and community. What are the, aside from, of course, what's the regulatory requirements from a data perspective, what are the key data points and how do we define them? And that way we know how to educate customers and prospective customers. And I think it was, was it the radiology that claim, I can't remember who presented that earlier, you know, some type of, you know, decision-making like that. Thanks. So I think one of the things that we found when we started to talk to patients about artificial intelligence is they get confused quite quickly and quite often they'll start thinking about robotics and they'll start thinking that artificial intelligence means a robot's going to come and do the procedure. And so what we found is you have to be very careful about the way you communicate with it. So I think it's also important that we're training the doctors to be able to communicate to the patients as well, not just that they understand how it works, but also how they should be speaking about it with the, with the doctors, sorry, with the patients. From our point of view in industry, what, you know, what we found was we started to develop tools that could analyze everything that goes on during procedures. And that's incredibly exciting to be able to see. And suddenly we had, you know, people coming to us and going, can I use that for training? Can I use it to review cases? So having clear direction on how you want to use this sort of new world of information is also really important for us as well in developing the right tools that are going to be usable for you. Yeah, I have a question at that point. So let's say as industry, you deploy or you give us these algorithms and as clinicians, I may or may not know what exactly this AI device is doing, not to any detail, but I know it detects something. Do you have any instructions for consumers or end users as to how we need to navigate? Because this is a new box or a new bracket that is coming onto our screen. And all we know is just going to detect, but there should be some kind of parameters in terms of distractions or something. Please expect some kind of a distraction, but that's normal. Do you come up with those kinds of things? I mean, Chuck, I mean, you probably can answer like when you train radiologists to start using AI, do you give them a, do they know how to use this tool? Short answer, no, they don't know how to use it. That's a really interesting question about the distraction and as part of it, because I think part of what we really want people to understand is with all of these things, there's a trade-off, there's balance between what the system does and how it adds to your ability to do things. But it's, yeah, it's a challenge. I don't know of anybody who's provides that as kind of education that goes along with the adoption of an AI product. I haven't seen that anywhere. Just wanted to give John a chance to chime in, especially from a society perspective and someone who's very interested in education, what you think. Well, first of all, this is a fantastic discussion and a great session. And everyone on this panel has really been so supportive of education. Training is part of what you do, you develop AI systems and training is key, but it's interesting where the discussion is going. We're talking a lot about training in how to use the AI. And part of this question is what is the best approach to the entire, how is this going to be changing education in general? And I think we've been learning that we're going to be, it's going to be all the more important in getting, for example, colonoscopy in teaching the proper technique of doing the colonoscopy and examining the whole colon as we get this additional support for the cognitive aspects. The other thing is cognitive skills teaching is, I think, under-emphasized in our training programs. Look at the image that Raju showed earlier today of the serrated sessile polyp that an expert could pick up in a second. Even if an AI shows you that abnormal, it doesn't say why. And you're likely not going to miss it the next time, unless you have someone telling you what is the features about it. And they need a fund of knowledge in terms of image interpretation. So as we develop programs, and I think what we really knew in terms of cooperation between society industry is to support and develop better early learning curriculum for fellows early on to understand the AI packages you described in your lecture, but also to refocus and dedicate to image interpretation as an art, because that is something that will be very important. When AI shows them something, they have to know if that box is something real or something that's something to pass over. So I think that we do have to relook carefully, and the society would need to do it with our training committees and our educational council. And with supportive industry, how do we invigorate the learning process? When I was a fellow, we were learning band ligation. I'm dating myself, but I remember teaching all the faculty and the private practitioners who came in how to use bands. I think that if you get the fellows understanding AI's capabilities, and also refocusing on image interpretation, I think that's going to be a key to wider adoption. Yeah, one of the things I was going to bring up regarding training is the responsibility of the manufacturer to provide an interface that allows for a more intuitive engagement with the product. You know, we have to do for medical devices, human factors testing. And I think when it comes to products like AI, the human factors testing must address how easy it is for a user to interface with the device so that you don't have to go through hours of training as to how to use something. It should be very intuitive. And there's obviously a trick to that. I think the other thing that needs to be considered in training programs when it comes to AI, you know, when you think about AI, my first reaction is it's doing the hard work for me. So the concern by the FDA is they don't want the physician or the user to become idle during that process. They want the physician to be engaged. And I think if AI is able to do that, and I think if AI is developed to demonstrate that the user actually does better as a result of using AI, I think that's very much a part of the inherent training that should be associated with it. Yeah. Hi, everyone. Thanks. I'm John. I'm the CEO and founder of a company called Iterative Scopes. Thank you for having us here. I really like the conversation that's going on right now, but I want to reference to, I think, Raj, you mentioned, you know, you made this analogy around AI as, you know, GPS kind of thing where, you know, it's just showing, you know, to me that's, if only our daily medical practice, and I used to be a physician, used to be a surgeon in my training as well. If only our daily practice was that straightforward of our daily commute to work, right? One way in, one way out. We kind of, by the end of five, 10 years, we close our eyes, we can get there. Unfortunately, biology is such a way that it takes us in twists and turns that we don't kind of, we can't anticipate a lot of what we see on an everyday basis. And that's what keeps medicine exciting in some part. But the way I view AI is also like, and its, you know, relation to education and to the ability to influence behaviors, there is a lot of information as we look at the GPS on a daily basis, there's these twists and turns, unexpected, you know, roadblocks that we come into, that come into play. And education, the way I look at education is a function of how fast can we create, how short can we create these feedback loops to provide, you know, point of care information at the point where physicians are making these decisions, really important decisions that impact so many lives to be able to get them to make a well-informed, well data-driven decision, right? It may not come in the form of numbers, it may come in the form of a bounding box, a purple box, whatever it is, it's still data-driven in some way or another. It was trained on a certain amount of data. And how do we be able to connect those two dots? And how do we, you know, looking at the larger picture here, look at how many, you know, like Google today will tell us, okay, this road is blocked. This is the other way of going to your work in the fastest, in the shortest way possible. How do we anticipate that? And how do we provide that information to as many clinicians as possible in the clinic while they're practicing on a daily basis? I think that's an open question. I think professional bodies can certainly help us navigate that. And we're looking forward very much to hearing back from professional bodies around, you know, I think a big question for us is access to voices, access to as much as residents, as much as fellows want access to companies, we also want access to their voices. And I think facilitating that two-way exchange is something that can be filled by professional bodies in some way. That's exactly this, right? Two-way communication. All right. Bill, do you have a comment? You have the pathology and AI there. So that's a little more complicated for our GI brains to wrap our heads around, but CDX. So it'll be interesting to get your- I guess my comment was stated early on, and that is from the societies at large, there has to be a comfort with AI. You got to get your membership, an understanding of what it is and what it is not, and letting them know that it's a tool and that there's always a physician at the end of the day, and that it's not a replacement item. And there's got to be an overarching call to that from the society that will help all of these gentlemen teach on this optical platform. But hearing it from you and within the membership and getting the strength of that understanding will make all the difference in the world on the next three steps of teaching on AI. Yeah, John? Yeah, I just wanted to add from my private practice perspective and my history going back with narrow band imaging and trying to get a lot of my colleagues and in private practice to start using it, just a little button on the back of the scope. You have to really go over, show how easy things are and what the value is and how it can make you better. And I think that's simple. I think that one of the lessons I think is, as one launches these new things, is one really needs to develop efficient and even validated training modules for folks, for the attended audience. It's a different module for fellows or for people at an academic institution perhaps than it for in private practice. But you've got to figure out a way to get folks understanding how to easily use the technology to give themselves more time, to get better yields of their diagnosis from Barrett screening, to finding more serrated polyps. And how that, you know, they can see the value in that education quite intuitively. But really, how they actually do it and how they can do it in an efficient manner. And it's not a long thing, but we need to develop very precise and test-out models of how that can be conveyed. And that has to be part of the launch. Let's make one comment and then go to Sean for his question, which is that, you know, as I think about this, one opportunity for education and what we're not doing a very good job of right now is the learning that happens is we suspect that people are learning when you highlight a difficult to see polyp or perhaps your test diagnosis, high-grade dysplasia, but it wasn't seen endoscopically. But that's not necessarily reinforced learning of complicated things that people have missed. Early gastric cancer, you would have missed if you didn't have AI. So a learning library to highlight to trainees of subtle lesions that are missed and that have been identified with AI will both help people or humble people who don't think you need it, which I think is very important for both people in practice and trainees, but also reinforce the morphologies we should be looking for when we're looking at this in terms of computer vision. So I think that education on a broader level, rather than just hoping that that immediate feedback makes you think, oh, I should be paying attention to this, may be helpful. And so I don't want to hog the mic. I would invite you up there, Sean, but the stage might break. So I'm going to have you skip a question from there. I mean, I agree with your comment, Raj. And actually, I would, I would kind of further suggest, you know, should there be some guideline around fellowship training with respect to AI? Should there be many fellowships around AI? And if so, perhaps industry-sponsored in order to make that happen, or industry-specific training courses that happen in this so that we can measure and see exactly what's happening in fellowship, or to the fellows from the utilization? Because as you're right, it's right now, it's as you learn, as you go, as you use, kind of like in the early days with endoscopic ultrasound. But I think anyone would argue that our current endosynographers are probably much more faceted and better because of the structured training they have now, compared to those that had to learn it as they went. Just to add one comment, I agree with what was just said. Beyond just a tutorial or a module that helps people understand how to use the AI or make the most of it, there's also, you know, the option for kind of an add-on of sorts in which there's transparency, not only around the type of lesions that were included in the test data or in the training data, but also types of lesions that were commonly missed in perhaps human comment or expert endoscopist comment about what those features are missed, what locations are missed, and why AI has perhaps a better time analyzing them than we do. I think that type of, as I mentioned, as an add-on into a module before use, but perhaps actually, you know, moving both directions in which the AI is pushed out through cloud, and then examples in which there's discrepancy between the human judgment and the AI judgment can go back into the software company, and they can implement that into their kind of set of training data, and hopefully improve the imaging capabilities and image recognition capabilities of endoscopists. I think it's a good two-way conversation to have. Yeah, I mean, I'll just, it's a great suggestion, Dr. Tech, right? I think there's definitely a role for industry to get involved more early in fellowship programs. I think, you know, as a specialty, we're on a journey here. So creating that baseline of acceptance of these type of algorithms and building trust earlier in the process will allow, when a physician's gone into clinical practice, will allow all these toys that we're talking about to be accepted and we can actually put them into practice. I think if we can establish a baseline and meet the trainee or the customer in our regard at where they're at in the process. So that's starting from the basics and as an industry, whether it's society, industry or physicians move along this journey together, we'll all be successful. So I said earlier, I think it's a collective effort. It's going to take time to get there. But I think there's definitely a role for industry to start the educational process earlier with academia. If I can jump in and just share a quick comment. I think the idea of having it part of a fellowship program, AI that is, makes obviously perfect sense. I don't think that's going to be our challenge. I think we're going to have receptivity from fellows transitioning into fellowship, residents transitioning to fellowships. They're going to want to learn the new technologies. I think our challenge, industry and the ASGE and all the societies is going to be the other 99% of practicing endoscopists that are out there on the hamster wheel every day, that are pressed for time, that don't have the time to learn, that can't fly from the Midwest or the East Coast to San Francisco for a one day AI course. I think we've got to use technology. We've got to use a cloud-based learning, whoever just said that. I think Sean, I think we have technology at our disposal outside of the medical world that we can use in the medical world from a training standpoint that may or may not have to go through the FDA, but that I think we can facilitate that adult learning much more easily than we have done historically. I think it's an incredible time and I think if industry and society can truly embrace this opportunity and work collaboratively together, which will be hard because capitalism gets in the way of that very often, but if we can find a way to work collaboratively together, I think we can make great leaps in the education and training of the existing endoscopists as well as the future endoscopists. Quick comment on that as part of the ASGE AI taskforce that's been on our radar and we're trying to put up some kind of small videos or educational videos so that people can open access once you're an ASGE member, you can just log in and get some of that information and some of that training because we want people to get up to that minimal standards and then you talk about the complexities and other things. My question to all of you here is, would you be willing to partner with societies to do this? Yes? Yeah. Great. I think we have it captured on video, so we're good. I just have one question, actually this is to Dr. Khan. I know we've had talks here where we've looked to other specialties to see how they've been done. We've talked about how radiology is at the forefront and talking about education, talking about fellowship integration with industry, do you have anything to suggest of how radiology as a specialty has done this or are you thinking about it? Well, yes. I will say the one question I get asked most often is, are radiologists of the future all going to have to learn how to program in Python? And the answer is no. No, actually, one of the things, in fact, I think that's really important is that societies such as this one make AI and learning about AI a priority because what that says to that 99% of people who are outside of and done with their fellowship and out in practices, this is something I need to learn about. This really is a case that it's not that AI is going to put endoscopists out of business, it's that the endoscopists who are using AI are going to put people out of business. It's going to be that you need to incorporate this into your practice. So a number of our radiology societies, our large one that has actually an AI certificate program that we've put together, it's an entirely online curriculum that people can participate in that focuses specifically on radiology AI. One thing that I talk to people about even before they start learning about AI and we introduce it during our residency is just having a grounding in informatics to understand how, and I think as part of that critical thinking skills, I heard someone talking about confidence intervals, about statistics, to have an understanding of when someone starts telling you something performs well to be able to ask the right questions of it. I was thinking kind of about analogies between what you all do and what we do. I mean, we all use technology of one sort or another to form images. Part of my training and board certification as a radiologist is in medical physics. I have to take, you know, that becomes part of our core exam. And so I'm not a medical physicist, I wouldn't profess any expertise in it, but when I see an artifact on a CT or on a plain film, I can talk to my medical physicist and say, this is what's happening there. There's a detector flaw and then that's the result. And I think in the same way that in radiology, we have these medical physicists who every day check the CT scanner, make sure it's not overdosing people, make sure it's calibrated appropriately. You folks have it differently. You have sort of the knobology of the scope, right? You're not maybe board certified in it, but you're tested in it, right? You're attending is sitting there looking over your shoulder as you're, you know, controlling the scope and making sure that you know how to use it. This is just another part of our equipment. And I think it's something that we don't have to be experts in AI and data science, but we have to know enough of it to communicate effectively with our colleagues in industry, in our practices who are expert in it so that we can assure that the technology is being used most effectively for our patients, that what we're doing is safe and beneficial. And I think that's a real opportunity. And so we've, I can assure you, there's not a single national radiology meeting that doesn't have these days, some core teaching in AI. In fact, some of them, if you look at some of our journals, you'd think they were AI journals. The third of the papers in some of our leading scientific journals are focused on artificial intelligence. And then there's a lot of educational content that we've matched with it. Just one sort of comment since Shivanthi just started making pleas on the panel, and I'll do that now. I think it's a very computer vision heavy sort of day. And I think that if education is made more broadly, you know, if we decide there's some sort of partnership and Shivanthi sounds like she's got a lot of great ideas to do that about education, understanding that obviously GI, we want it to not just be an endoscopy field and understanding the nuances of sort of the population health and what the EHR sort of already implemented by some of our vendors, what those machine learning solutions are like would be helpful, because I think that'll really skew what our trainees understand about AI. They sort of focus on just the computer vision part. Because I actually think my own bias, I think the non-computer vision stuff is going to be more impactful to our field than the computer vision stuff will be ultimately. You can ask for more things now. Yeah, yeah. So at that point, I have a value proposition, right? We're talking about all these computer vision type algorithms and AI stuff. But one of the things that if you talk to any director of endoscopy or any administrator, if you say, I'm going to develop an algorithm which can help with no-show prediction so that you can overbook or make sure your staffing is optimized in your endoscopy, whether it's therapeutic or general GI, they will be the first ones to press the button on that. So as industry, so I want to kind of move from education to kind of the research pipelines and new AI technologies. Do you foresee this kind of technology, or do you have any pipelines looking at these kind of use cases? I can imagine one for gastric cancer in which high-risk patients who would otherwise go unnoticed, perhaps because of their insurance status, maybe they're underserved, et cetera. Their electronic medical record is basically trawled for certain factors that make them high-risk and that leads to a referral and starting the process of screening for those high-risk patients, which has been shown to be cost-effective compared to population-wide screening. I think that's one thing that can make a big difference in the survival rates of gastric cancer versus cancers for which there are screening programs. If not, Austin, I have a question for you. You're talking a lot about AI training, and one of the things is we do have these meetings and stuff like that, but reaching to those individuals, telling them, hey, this is important, is one of the platforms, and you have a voice on social media. So I'm just thinking how we can accelerate that need for people to understand some basics of AI, whether they're in practice or in training and so forth. Yeah. Well, can everyone hear me? I mean, I think social media in general is just a great tool to get the message out there, and I think collectively we can all raise more awareness about this as there's more that's developed over time. But yeah, I mean, I think that it's really meeting the patients and whatever audience you're trying to target, where they are, and crafting your message to really fit that group, that audience that you're trying to reach. So if that's patients or if that's other physicians, then you may have to choose your platform carefully and craft your message in a way that's understandable and easily digestible. I think there's been several times that we've mentioned explainability today, and so the same goes for whichever audience you're trying to target. One question, but Brian, want to come back, is it coming on social media? Well, yeah. I just, if you're looking for ways to activate your user base, I think one area where Sages did an interesting approach was they created the CVSchallenge.com. It's the critical view of safety challenge. And what it did is it was a call to action for industry to come up with an algorithm during a lab coli with a critical view of safety, a key step in the procedure to minimize any complications. And at the same time, they also asked their user base to donate videos and or pictures. And as an adjunct to that, they actually used the Sages parameters for annotation. So if we're looking for creative ways to actually activate a user base and galvanize people to get more people involved and interested, that may be a consideration for ASGE to come up with maybe the Casuani pancreatic cyst challenge in the future. Just a comment. Also raise the awareness in educating the quality of exam. It's difficult to even match the product with the customer, because if you think of a fee for service practice, why would I want to spend more time looking at the stomach while I can just do it in one minute? And that's okay. No one cares about grand sequencer, right? So I think for us to partner with ASGE and other societies are really good and we could talk about it. Also, when you mentioned social media, I think it's really good because the next my generation, next generations is all about looking at Instagram. I thought, well, maybe it's a good idea to do our Instagram or social media, but I don't think we have to split into different social medias. We can just also collaborate and send some ideas and pictures and to do on Instagram stories like, oh, what do you think this lesion is? And yes, access, what is the knowledge of the students nowadays? So, yeah, that's all my inputs. Yeah, I think to highlight your first comment, especially for those online who didn't hear the beginning of it, tying AI into the disease state and understanding why the disease state is underdiagnosed or underrepresented is helpful, especially as we move beyond colon polyps. We've sort of talked about this a few times. We're so colon polyp heavy here that things like esophageal squamous cell cancer, which people think they never have missed, but they've never seen because they've never seen, but they've missed, just like gastric neoplasia, all these things, tying the education for the disease into the AI education is sort of a way to approach this, really, when you think about, I developed an AI algorithm, maybe talking a little bit more about the disease first as well, because I think most of it's an underrecognition and we just spend, as we said, far too much time talking about the same thing over and over again and not treating some of these less represented diseases. Jonathan, I think you had a comment. Yeah, I think that some of these innovative and fun educational methods, short and sweet and really to the point, are going to be increasingly used. We're really doing a lot of work with the learning committee and coming up with gamification. And I think it's cognitive skills and how AI enhances them and how it relates to it. But there's a lesson about the disease, but also about the images and if you have a patient you're following an intestinal metaplasia, how are you going to look at that? And what about... So I think that that's interesting. And there's two separate... There's a dichotomy here between general AI instruction curriculum, which is something we need to do and to enhance our existing educational curriculum and maybe come up with some online modules, as was mentioned, about AI in general. But I think that also some of these integrative, fun, gamified series, whether it's tips or challenges or what have you, certainly with the welcome industry support, are going to be very exciting to enhance both for fellows and for folks in practice as well. Yeah. I think to get to the gamification idea and Brian's comment, I mean, the interactive learning where fellows in their first year essentially develop an algorithm by submitting images and watching it go through iterations, I think Srivanta could take them through that through the first year of their training. Internationally, it's a fun idea. It can start with something that's already been made or they could pick the algorithm they want to make. We want to detect EOE and then they basically go through that education. It's a little bit like that. Not quite the gamification, because I don't think we can have fellows competing against each other making better algorithms, but I think that's the educational committee of the ASG could really, a council could do a lot to help with that. So I think it's a great idea. So while we have Sangeeta here, who's one of the program director, I would like her to get your input. What do you think? Let's say tomorrow we have developed this curriculum and ASG Learning Center has these sort of sources before we take on this task. Do you think program directors would be interested in this kind of education? What's your thought process? Yes. Thank you, but yes, I think program directors will be definitely interested in this. As we all were talking about social media, so social media is a thing a lot of fellows are on, not necessarily the senior program director. So I'm just making an effort to be on social media so that I can see what's going on there. But it's not something which comes very automatically to us, oh, let's go on social media and talk about it. So, but it is like ASG can have these, even the learning modules, like when EUS first came around, you used to have this training session on big models and things like that. What is the concept of EUS? So similarly for NBI to detect the Paris and NICE classification, and similarly we have for AI also, like a structured way to teach AI to the fellows starting from first year because that's the time they're going to learn. Right now, I have to keep reminding them to switch on the NBI. And we have these charts of Paris and NICE classification printed out and kept next to the screen, look at it and say, now what it is. So if there is some automated way of doing it, and also a curriculum that we can teach the fellows, that will be really helpful for the fellows. Yeah, so just on your point there about actually turning on NBI, so what I think we're sort of, all of the things we're talking about here are really important and it's really great to put those things in practice, but there's a bigger opportunity here. And that's the fact that by taking that data to the cloud and using machine learning to process that and understand what's going on during procedures, we can get a totally new understanding of what's happened during a procedure and a very simple thing that we do. So we take all of our data to the cloud and we process it. So what we were seeing is in one of our sites, they weren't switching to NBI before they were using our characterization module. And we could detect that automatically using another AI algorithm, and that's just sort of the start of where it's at. So there's a whole new world of understanding how doctors are using technology to do data driven product development that we're only just starting to touch on. And I think that's a really exciting step forward. And that really comes from the power of not just putting an algorithm on a box and throwing it into a room and never going back. It's about building a relationship with that hospital and a relationship between the technology and the end user. Just one comment. And then I think that the session's titled AI education and training, and we heard from other program directors as well, that just in general, the data that comes sort of from what we were talking about, you use NBI in 1% of your procedures, but your colleagues use it in 5% of your procedures or your insertion time is this number of minutes or colleagues is this. You know, all these sort of things. You use a cold snare much more or use cold forceps much more than your colleagues. That is education both for the independent practitioners and as well as the trainees. So making it easier to give that feedback, I think, would be valued. So it's not the same sort of thing about, you know, how do we teach people about AI, but how does AI help us teach? I think would be very helpful as well. Please go ahead. It was just another idea to share. I'm a person with ideas. I remember one DDW, I saw, I think, a panel with Medtronic and they were also wanting more ideas so they can invest. And I think a lot of the students here had and also meeting with your question, like, do you have other algorithms that can leverage AI to teach people about AI? So we are looking for more, better outcomes for patients, for detecting or NLP, CDS, other things. We are not looking to do like some kind of design thinking and pitching. And maybe we all have investors. We can even like a thousand, twelve thousand dollars for us who are, I'm not residents anymore, but residents, it's a lot like, oh, I can travel or invest on something. So they can do their whole framework and even the abstract and they pitch, they can develop something and all of you can invest because I talked to some people here, they have really good ideas that can turn into real products. So I think it's also a win-win situation for everyone. To keep everyone in mind and we think for the next. Thank you so much for that comment, Caroline, just so that the people here, as well as the ones virtually, we do have a feedback for specifically the SGAI, we have a web page where you can send us your feedback, your questions, and if you want to collaborate or you want to start a research project, how do you get started? You can always reach to one of us, not that we will get your wishes granted, but it's something that we always take these feedback very seriously and try to improvise on how we do our programming and so forth. So, John, before I close the session, do you have any comments, questions or anything like that? John, from the virtual. No, I thought this is a really exciting question. I mean, this is the last comment, you know, it reminds me with the development of notes and the NOSCAR grants and how we inspired a lot of people to do a lot of research. Some of it led to other things. I think that we're all thinking proactively about a future that we see is coming and we're realizing that education is going to be key at all these different points of the, you know, from the folks that are in practice to folks that are at the big hospitals, developing relationships and especially to fellows. And I think that coming up with some creative ways to enhance that early on learning is going to be both about AI in particular and the foundation of it, which is image interpretation and management. You know, I think those are those are key. I also think that the competency assessment potential that was talked about earlier sessions is fascinating, objective competency assessment. We always look for that. I think it'll be very interesting to have that and compare that to our manual tools, saving up time and teaching. Very, very interesting. And lastly, I think that the whole issue of competency, independent practice is going to need to be rethought about because we have now we have we have some assistance. And even if it improves our ability to do things on our own, if we're going to continue to use AI in practice, I mean, I don't know. You know, we need to rethink about competence procedures, maybe if you have good outcomes, you know, with or without AI. So all very exciting. And I think that the ASU is going to be really, really excited about partnering with all of you and in coming up with specific, you know, projects in this area. Well, I thank you, Jonathan, and thank you for the panel for committing to the education and research. So I think I'll kind of conclude and close this meeting. It's about four o'clock now or actually four or three, and it's a beautiful day in San Francisco. So if you are staying here for a little longer, try to enjoy all the good food up here. So I would like to thank all our industry sponsors and the ASGE staff and the ASGE task force and all of you who made it all the way here to be part of this conference. Thank you very much.

education and training

artificial intelligence

AI in medicine

fellowship programs

industry collaboration

social media

gamification

patient outcomes