We're going to open up to questions. If anybody has questions, please text it to the poll 22333. We'll start off with one for you, Dr. Bersin. How well does AI detect flat or serrated lesions? So the performance of any model is just all about the training set. The original training sets for most of the polyp detection systems were not enriched with flat lesions, but that is changing. And so with each subsequent model that is published and ultimately released in the environment, you'll see higher and higher rates of flat lesion detection because, frankly, those are the things that we need most help with. The tricky part about the published data is that anything we published two years ago is on a model that is already outdated. And so the systems in 2023 and 2024 are just going to outperform those original U.S. and international studies from three or four years ago. So if you as a human can identify sessile serrated lesions, trust me that a computer system will be able to do that as well as you and maybe better. I have a question for Tyler as well. So I'm married to a radiologist, and he's also, like, equally worried about robots coming for his job. But it's interesting, when our families hear us talk about AI, there's a split. They're nonmedical. And so some people think it's really cool and exciting and very innovative and think that it will better their care. And then some people have a fear of it, that, you know, they'd rather just have a human doing their procedure or reading their CT scan. Do we know anything about how patients feel about the use of AI in their care? It's a great question. I think there have been a number of surveys, both for physicians and patients, around what AI may mean and perceptions. And I will tell you that part of it, honestly, part of the complexity is the term to start with. Artificial intelligence sounds in some way that it is like it's sentient, like it is thinking on its own, that it is replacing us. And I really think that that makes it honestly too complex, too sexy a concept. These are all literally just prediction tools. They're probabilistic models. And physicians, for many years, have used new computer tools, new technology tools. This is another one. So physicians are not going to be replaced by AI. Physicians who don't use AI will be replaced by physicians who do use AI. That's a classic quote from radiology. But we're going to, I think, seamlessly incorporate these into our practice. This is not a matter of replacement, from my perspective. Excellent. Thanks. We have a couple more questions from the audience. Let me see if... One last one for AI. I think you covered it, but I think it's, should our hospital rush to get it? I don't think we have to rush to do anything. I think that AI, again, I think we have to think about it just like any other tool. We have to think of it thoughtfully, coldly, in some way, objectively. And we should be having, for now, around AI polyp detection, the same types of conversation we should be having around endocuff, or whether we should be retroflexing, or whether we should be documenting withdrawal time. They're, right now, the same discussions. And so, yes, in terms of using tools to improve colonoscopy quality, we should be talking about it. AI is one of those tools. It happens that the opportunities that AI will present will grow very dramatically in the next five years, and I think there are advantages to beginning to learn about the technology. But outside of that, I would judge it just as coldly as you would judge any other tool or device you might bring into your unit. Excellent. Thank you. So, we have some questions from the audience for Dr. Kothari and Dr. Ladd. If you could comment, in your practice, for cold-snare EMR, how to decide whether to do piecemeal cold-snare EMR, what features of polyps would suggest that that could be a good option to use? And then, also, maybe comment on the use of underwater EMR, and whether you use that in your practice. Yeah. Yeah. So, salcerated lesions actually render themselves very well to do a piecemeal cold-snare EMR. For adenomas, as I said, the recurrence rate is higher, and studies have shown that. In terms of preference for cold-snare, if I have a patient, they cannot stop their anticoagulation. I'd rather see the bleeding there and then. Those are the patients, actually, where I will sway towards piecemeal cold-snare EMR than using hot-snare and having the patient come back with delayed bleeding. Underwater EMR, I know there are firm believers and there are non-believers. I'm somewhere in the middle, where if you have a polyp, there are some little scarred areas. If you do fill up the colon with water, you can have the polyp raise up. However, it is difficult to keep that kind of water insufflation to resect the polyp, so I do use it for some scarred areas of a large polyp that I'm removing, where I'll use the underwater technique. I don't purely use underwater for my lesions, but that's just my practice. Yes. I do ESD for everything. No, so I agree with what Shivangi said. I think the main decision for me about doing cold-snare EMR is basically the size and maybe the location of the lesion. I am also in that group of endoscopists who I like. I like swimming, but I don't really like doing my procedures underwater. I know there's literature. I'm not denying the literature. I just prefer to do it in that way. To me, it's either a classic hot-snare, lift-inject, resect EMR, or if I'm going to do cold EMR. Very rarely I inject. I just go straight with the snare, and yes, in my limited experience with cold EMR, because I just recently started adopting it, I've tried to do it mostly in sesal serrated lesions because, like she said, they're easier to grab with those snares. So one question that came up regarding rectal lesions, and you both had beautiful pictures of large EMRs or ESD for these rectal lesions. What is your opinion regarding TEMS or TAMIS versus ESD for these rectal lesions? It's interesting you ask the question. I was trying to find literature on that for a committee work that I was trying to do. There was, from what I remember, one paper that I could find by the San Francisco group, and I remember, if I remember correctly, there were no major differences in terms of complications, and I think mostly they were looking about cost, TAMIS versus ESD. You know, TAMIS is, at our institution, we have, I'm recently moved there, but I think some institutions have a higher use of it than others. At my previous institution, TAMIS was almost never offered. Over here, they use it a lot more. That's my answer. I'm sorry, I don't have that much experience with that. And I should clarify for the audience, for those, I mean, we're talking about transanal excisions that would be able to be performed, and along the same lines, you could talk about endoscopic full thickness resections as well for some of these lesions that might have some evidence of submucosal invasion, either from the lesion itself or from scarring from previous attempts at resection, which could be helpful. Thank you. I probably do. Just one last question for Dr. Deer about endoscopic management of fistulas and strictures in patients who have IBD. If there's any data for endoscopic suturing of fistulas, and then also stents for Crohn's disease-associated strictures, covered versus uncovered. I will say, like Tanvi's IBD patients tend to be so complicated, and most of the time, at least for suturing, I can speak to, you know, it's very difficult in these patients to suture areas that can be very inflamed. You know, oftentimes, they don't respond very well to suturing if the tissues are not sturdy enough to actually take to suturing, but I'll let Dr. Deer comment on any other additional comments. Yeah, I agree with what Dr. Patel had mentioned, that, you know, these patients have quite a bit of fibrosis, so I think suturing would technically be challenging. I'm not aware of any data looking at the use of that type of modality for management of fistulas. In terms of placement of stents, you know, there's data on using partially covered stents or fully covered metal stents. Again, I think the jury's out in terms of how we're going to utilize that particular modality just due to the efficacy. I mentioned about how balloon dilation seems to be more effective. PASS-2 as well may be prohibitive as well for utilizing that procedure on a standard case-by-case basis. Thank you. I wanted to ask you a question about, and excuse me, this may be an obvious answer, but what is the difference between an anti-grade and retrograde dilation, and why is a retrograde better based on your slides? So retrograde dilation is the ability to pass the scope actually through the stricture, so you're placing the balloon past the stricture and then bringing the balloon back into the stricture and dilating it that way, versus if you're doing an anti-grade, sometimes there may be more inadvertent injury because you're doing it somewhat blindly. We just want to thank all of the speakers for this session. It was highly informative, and was going to say I know we're a little bit over time for this session, so we'll still plan to do about a 10-minute break, but if everyone could be back ready to start our next session, our last session of the day at 3 p.m., we've got some really good topics, so stretch your legs, stick around, we'll see you in a bit.

artificial intelligence

medical procedures

AI models

training set

physicians