On our panel, we have both Dr. Berzin and Professor Mori, and I want to introduce three other members of our panel that we're lucky to have with us. Michael Byrne, who is a clinical professor of medicine in the Division of Gastroenterology at Vancouver General Hospital, and UBC has a lot of experience with AI development and research. Dr. Hong-An Yu is Chief Division of Gastroenterology at the Renmin Hospital and Professor of Medicine at Wuhan University, the key laboratory of Hubei Province for Digestive System Disease in Wuhan, China. And Dr. Michael Lang currently works at, I'm going to mess this up, at Sahlgrenska University Hospital, Longdao Barum Hospital. And he does research in gastrointestinal endoscopy and colorectal cancer screenings, currently engaged in a project related to improved diagnostic yield in GI endoscopy by the use of AI. So I'd like to ask the first question to the panelists, which is, we know now that some systems for detection have the CE mark in Europe and in Japan. I'm not sure if there are any Asian countries outside of Japan. But so far, give us an update on what is actually happening. Are these systems being purchased and used? And what feedback are you getting from users? Alessandro, are you seeing anything in Europe? Yeah, I was going to answer, so things are going very slowly in Europe. So presently, we have four different systems approved with the CE mark, which is Metronix, Fujifilm, Pentax, and Endoangel. So four of them are approved, three of them are commercially available. So you can purchase them to the market. But if you look to the adoption, if you look to the numbers, so the number of units that have been sold in the last six months, they are growing very slowly. I think one of the points is being taken by Yuriko Mori, because there is no remorsement presently. So physicians and hospitals are struggling to understand how they can introduce this technology and how this can be paid. And there is nobody with that issue right now in any European countries. So I do expect a difficult time for this, because as long as we do not have a reimbursement code, we are going to struggle. Of course, it's important that we have a strong clinical evidence showing that as much as we adopt in clinical practice, we are saving lives because we are detecting more polyps and more adenomas. I wonder about the discussion of selling these as plug-in add-on devices. When a lot of electronic chromoendoscopy became available, and it still is, it's just part of the device. When you buy an endoscope, you have electronic chromoendoscopy available to you. What's the discussion about the best way to approach this, the add-on plug-in device, which has an extra cost, versus trying to figure out ways from a business standpoint to incorporate these programs directly into the endoscope systems when they're sold? Any thoughts on that? At a European level, I do not have any thoughts, because the strategies are very different. And also, we have companies with different business models, because Fuji and Pentax can go in this direction, probably. Endoangel and Medtronic, they probably cannot go. They do not have any partnership with the Scopes company. So I think it's difficult to make a prediction. Also, it's not totally clear to me what is the basic cost of building this technology before approaching the market. It is something that has not been reported yet. Thanks. Professor Mori, any uptake yet so far in Japan, just in regular practice? Okay, thank you very much, Professor Rex. And from my perspective, Japan is a kind of unique country, because the use of NBI, or optical diagnosis, is currently being reimbursed by the Japanese public health insurance system. So you can get a 20-euro or $20 for one use of NBI, amazing. This can be applied to the use of CDX or CDE. That is what we are thinking about. The answer to your first question, I think the CDE is much more popular in the customers or the users compared to the CDX, because the CDX is possibly reducing the income of the hospitals. However, if you use the CDE, you can get more money, because you can get more polyps or adenomas. Therefore, I think we should apply some reimbursement exclusively for the use of CDX. Otherwise, the widespread use of CDX is hindered. That's my personal perspective. Please, panelists, if some thoughts come to you, questions that you want to ask each other, just put your hand up, and we'll get you in. Michael. I just wanted to say in relation to your question as well about whether this is a plug-in device or incorporated into the scopes, I think we have to make the distinction as well between detection and differentiation, because of course, some of the industry groups have gone with a platform agnostic device, particularly for CADI, where one size fits all. I think that's going to be very difficult to do right now anyway, and the next few years for CADX, because as we've already seen, those of us who work in this space, as you well know, there are many different scope models, and there are many different settings on each of those scope models. Where CADE may be much more platform scope type, scope setting agnostic, CADX is not. CADX needs to be much more black and white, yes or no, whereas CADE is just an improvement to our detection. So I think we need to make that distinction as well if we're looking at built-in solutions or whether we're looking at a plug-and-play. I think it'll be very difficult for the foreseeable future to have a plug-and-play device that is platform agnostic for CADX. I want to ask the panelists thoughts about how AI is going to interact with documentation. Really, what we're after with these programs is improved prevention of colon cancer, and the fundamental thing here is improved detection. In endoscopy in the U.S., at least, there's no financial incentive whatsoever to do colonoscopy, anything other than as fast as you can and as often as you can. I think there are a lot of people that don't do that, but there are some that do, and they have resisted increased documentation. In endoscopy, it's a very opaque procedure. When you look at a previous procedure and you're trying to figure out what happened, you have very minimal documentation, namely, you don't have a video recording. So how, as we bring these programs forward, both for detection and for diagnosis, what are the demands going to be on documentation, and is this going to require a move to universal video recording and storage of the records? Tyler, any thoughts on that? I think we are absolutely still in the stone ages of recording quality for all of our units. Just as an example, if you try to pull out the ADR numbers for any particular unit in the country, most people still are doing that as back-of-the-envelope calculations. Occasionally, you can pull it out of the medical record, but a lot of the parameters that we really, really care about, we can't, and we rely on physician documentation, which is highly, highly variable. So I actually am absolutely hopeful that AI will help standardize the data that we collect out of procedures. Dr. Thakkar, who's going to be speaking a little bit later today on a panel, has been developing some software that automatically documents quality and key landmarks and prep quality, and to standardize that across 100,000, 200,000, a million colonoscopies across the country would be a really powerful thing to do. So absolutely, quality improvement is a major goal, I think, of what we're doing. I just wonder if there's going to be a big segment that is going to resist it. Oh, I'm sorry. I'm sorry, Alessandro. Go ahead. So I have a comment. I think this is the next step of AI. So I think that after detection and characterization, the next step is just to optimize colonoscopy documentation to allow everybody to have at least standardized system, because as you say, everybody reports differently, nobody's taking videos or editing storage of videos. So I think that AI will force everybody to have a homogeneous way to report colonoscopy, to describe lesions, and also to make videos as video storage. So I don't think this is going to be human. This will be technology forcing us to be better colonoscopies and better reporting. Yeah, I agree with you. I'm just concerned that there's going to be a big segment of endoscopists who are going to resist it, because they don't really want to see all the polyps. They don't want to take the time to remove all the polyps, and they're not going to have the skill to remove all the polyps, and they don't necessarily want to have a record of that. So I think we have to figure out ways to get these things in the hands of the people who need them the most, because the people with the most... As long as we have guidelines, as long as we have robust clinical evidence, randomized trials, robust data, and these data will be incorporated in national, European, American guidelines, everybody will be forced to go in that direction. So whatever is the resistance of these old generation of colonoscopies, everybody has to follow guidelines. So in the future, we need to have a very, very dedicated specific guideline to be incorporated with AI. Doug, doesn't it behoove societies like the ASGE to really drive this? So we can... Look, AI can undoubtedly help physicians with documentation, and they won't see it as a challenge or a threat. But you mentioned a few minutes ago that people will resist this. They don't want to see everything. They want to do as many colonoscopies as quickly as possible. And I was shot down almost on the podium last year in San Francisco when I suggested that maybe if a particular physician does not reach a certain minimum ADR, that they won't get paid, they won't get reimbursed. And by whichever tool they need to use, whether it's a cap, slowing down, or AI, or all three to get that ADR, then they will do so if it hits them in the pocket. And maybe that's actually a key change that also needs to be made, is people won't slow down unless they are forced to do so, or they need to attain a certain standard by which they get paid. Right? Yeah. So I think the comment here is that guidelines are going to be key in moving forward. And I think that probably is a pretty solid answer, at least for right now. There's some questions about CADEX and as we think about the costs and the cost savings associated with reduction in pathology fees, the so-called resect and discard paradigm. How far away from that do you think that we are, both in Europe, Asia, and the U.S.? Should we be including that in cost analyses? Thomas, any thoughts on this? Yes. I really don't know. But I think that one of the main issues is, I think this CADEM characterization may just work for smaller polyps, maybe polyps smaller than one centimeter, and of course, that is the majority. And I think it would also be helpful for informing the patient directly after the colonoscopy because now they are going to wait for weeks and months to have the answers and report. So I think that will be helpful there. And that will also be a part of the cost. Sure. Any other thoughts about resect and discard? Is this really going to enable this? Are there still going to be obstacles in terms of medical legal risk? Yeah, that's a problem. Please go ahead. Ale? Yeah. I think the use of AI is really beneficial to facilitate the implementation of optical biopsy, including the resect and discard strategy and the diagnose and leave strategy. I think that Tyler wrote a really nice article, which was focused on the questionnaire in the United States. And according to his paper, the use of the optical biopsy will increase if you have a chance to use AI from 40% to 60%. So I think the CADx can be a facilitator in terms of ACDx. Tyler, what do you think about this? I sort of thought of the data a little bit differently, which is that there's still a lot of people who are not comfortable doing resect and discard or view and leave. In the United States, I think it's going to be very, very hard to get gastroenterologists in the practice environment, in the legal environment, and so on, to rely on softwares until we get to an incredibly, incredibly high accuracy threshold. So I think the bar to change behavior in the United States is going to be very, very high. I have a question for Michael. I have a question for Michael. So based on what you have seen in the issue of medical-legal problems, you think that the CADx evidence must be built differently in terms of clinical trial? We need more evidence, more robust, more data, more patients, more trials. How is it going? Yeah, for CADx, we definitely, as we said earlier, it's black and white, unlike the CADe. You know, we can have varying levels of performance of improving the ADR with CADe, but with CADx, it needs to be right or wrong. It's black and white. So more trials will be needed for sure. You know, the ASGE, the ESGE, most of the main societies have already endorsed formally if you are appropriately trained, the use of optical biopsy in the eyes of people like everybody on this panel here, if you've monitored your own results and you can do an accurate optical biopsy. So I see this no different, in fact, obviously more proven out that AI can make that next leap for beyond experts like yourselves. And again, you know, the elephant in the room is that everybody knows that there's lots of unofficial resect and discard that happens all the time or diagnosed and leave in place. I think everybody on this panel would agree that they commonly leave behind very small polyps in the rectum all the time. But whether or not that is documented is also another contentious statement, but it's true. So when I think when the studies come forward showing that optical biopsy with an AI tool is as good as the best human operator or maybe even matches or exceeds pathology, then we can move forward. Doug and others have done studies looking at the error rate even from the pathologist for these diminutive polyps. So that so-called gold standard possibly needs to be challenged. But we're already doing it, Ali, right? We're already doing unofficial resect and discard and diagnose and leave in place. Why not try and bring it with much more evidence and robust quality by an AI tool that can do it accurately? I think there's a lot of diagnose and leave in place going on in the rectal sigmoid. I'm not sure really how much, Mike, there is resect and discard. But I think most people, when they take things out, are giving it to the pathologist, and I think that's where the real cost savings are. And I'm concerned that to drive it forward, we're going to need reimbursement, which is just a challenge in the US. Japan seems to be much more sensible about developing avenues for reimbursement for new tools, but it's been very hard in the US. There's a question here about, are there avenues for endoscopists to collaborate with data scientists? I think this has to do with regard to image collaboration. Hongyang, are you familiar with any ways or others on the panel that we can do this? In fact, I think it's very difficult to cooperate with other groups about the imagery collection. In fact, you know, if you want to get data from other hospitals, maybe other departments, also it's very difficult in China. I think that's the situation. But in fact, what we want to do is use limited image, limited data to develop very good model. I think that's the only way. Thank you. Doug and Ali, are you guys okay if we end it? We're right running on time or a minute late. So Doug and Ali, any final thoughts before we wrap up, please? Just want to say thank you to Ali and to the panelists. These presentations have shown we're actually here. This is happening. And so that's fabulous. And at the same time, there are some issues that need to be overcome. So thanks, everybody. Ali, any last thoughts? No, thank you. Thank you also from my side. And stay safe, everybody. Thank you.

artificial intelligence

gastroenterology

current status

challenges

panel discussion