Would you like to start off the panel discussion? I have a few questions, but we'll wait to see if there's anything from the audience or Dr. Messman. Yeah, sure. So maybe what may I ask Dr. Kahn as a radiologist who showed us that GI is really behind radiology. Do you realize that in the future, the physician of radiologists will no longer exist if the computer takes over all this diagnostic tools? Are there any data that radiologist numbers go down? So I didn't quite hear the very end of the question. The question is, is the future of radiologists at stake? Oh, it's always at stake. And I thank you all for your concern for our well-being. Well, first off, if you follow Jeff Hinton, you would know that I don't exist at this time anyway, right? Because he predicted, I believe, seven years ago now that in five years there would be no need for radiologists. By the way, I still have a job. All of our trainees have a job. I think for all of us, quite frankly, I would love to be starting my career over again. It is the most exciting time to be in medicine. And particularly if you have an interest, if you're willing to use AI, I think it offers to us this phenomenal opportunity to extend our abilities. The only thing I guess I would say is I regard AI as another imaging modality. Those of you in the room who were around when MRI first came out, there were a lot of predictions. Not only will MR tell us if the lesion is benign or malignant, but because we can do MR spectroscopy, we'll know all the molecular happenings inside on every pixel. We'll never need to biopsy a lesion again. Didn't quite work out that way. I think AI is going to be phenomenal in what it's going to allow us to do. But I think it needs the direction of physicians. It needs people who understand it. And quite honestly, if AI can put me out of doing one task, I can move on. I can start doing other things. And I think that's kind of the case for all of us. Thank you for your answer. I mean, I'm also not worried. I'm also asked as endoscopists, will we lose our jobs because now the AI system will screen? No, I think the number of experts will increase with the use of AI and the effectiveness and the quality will improve. But not that the physicians will lose their jobs. Thank you. Michael, you raised your hand. Yes, I have a comment and a question regarding, again, this issue of AI and how we learn from AI and how AI learns from us. We've seen in the past that when we've introduced new technologies, for example, when narrowband imaging came out in colonoscopy, there were studies that showed that over the period of introduction, the endoscopist actually became better at detecting polyps. Because, for example, the narrowband imaging in that case revealed polyps that the endoscopist was not previously aware of. And it taught them that there was a polyp they were not aware of. And I think we're likely to see the same with AI as we see those green boxes that maybe we didn't realize that was a polyp, the student example or training example that we saw from Raj. So I think in the end, if we use these tools as not just replacement, but really as teachers, it can ultimately teach us to do a better job at detecting and classifying lesions. But the risk is that we do the very rapid colonoscopy pullback. We say, I'm going to just use the AI tool to replace me, or I'm going to use the AI tool to teach me. And so fundamentally, I think if we take the second approach to use AI as our teacher to improve our skill set, we benefit. If we use AI to replace us, well, then we get replaced. So I want to ask Raj that question. Somebody who spends a lot of time thinking about how we're teaching fellows, how do we teach our fellows and even teach ourselves as experienced endoscopists to use these technologies to make us better versus to make us maybe more lazy? Yeah, obviously some of it's, you put it more eloquently than I could, Mike. It's a challenge, but I think that to extend off what you said, I think of AI and our fellows when they've used AI in their training, have treated AI like a great endoscopy teacher. Like you said, finding things and telling you things that you didn't know existed and you didn't know you should know about. And so I think it's a little bit intrinsic, right, of your motivation to why you're doing endoscopy and wanting to care about finding these subtle lesions and helping patient care. I think that the flip side of it is, I think AI will help us identify people that are just doing a bad job in general and solely relying on the detection part because AI can help us with the metrics part too, right? So it's not just, oh, AI did it for me and I only needed a one minute withdrawal. It's going to make it a lot easier for us to pull out these quality metrics that too many of us are still not doing, still not abstracting. So I think that we have AI both as a potential problem by making us overly reliant, but then also being a backstop and saying, hey, you're not doing a good job. Actually, I can tell because I just reviewed the last 100 colonoscopies you did and I know what your pulp detection rate is, I know what your withdrawal time is, and I know that you're not doing a good job. So I'm sure you think about it the same way, Mike, that you can't just look at it in a silo of reliance on AI. Yeah, I agree. I think also measuring quality, right? This is the percent of bowel you looked at with good quality images. That's going to make a big difference and that'll also improve your overall quality. I see Dr. Seth Gross is on. Seth, do you have any questions for the panelists? Yeah, I actually have a question for Raj. First off, it was a really great session. I enjoyed everyone's talks. You highlighted something really important about that notion of a first year fellow not being able to identify that polyp. And when I think about it, it's been a long time since I was a first year fellow, but when there's so many things going through their head, not so much even finding the polyp, it's getting to the cecum, it's being straight and not looped. And how do you think AI could potentially help the technical challenges of colonoscopy? Because when you're in practice or when you're further along in your fellowship, all that's sort of automatic. You're thinking of loops, you're taking them out, you're moving forward, you're not thinking about the technical aspect of colonoscopy, you're thinking about looking for lesions. And I think that's really challenging when you're in training. It's a great question and it just shows how many areas that we can impact. And I think of a little bit about one specific example set that comes up in training now. And I know when the fellows scope with AI and they find a polyp and they need to remove it, when they're early on in their second year, they're extremely afraid to basically reduce the loop or put the polyp at six o'clock to remove the polyp in the appropriate manner, to actually put the polyp in the correct location, get a snare on because they're afraid they're going to lose the polyp. And so an unintended benefit of AI has been that people can actually perform high quality polypectomy because they're not afraid they're going to lose the polyp, that they can find it. The question is, how many of these sort of algorithms can you build in into a system? And I guess that's where the industry partnership will come up when we talk about it later. Can you build a force meter on your colonoscope and say, is this too much force or not enough force based on the patient characteristics? Can AI tell you how to reduce the loop based on using some sort of idea of the confirmation of the scope? The question is where the biggest bang for your buck is. So what do you think, Seth? No, listen, I think that we can't forget that we still have to learn the art of endoscopy and we can't necessarily rely on machine. We have to rely on ourselves and the people that are training us. You know, there are things out there, right? If you think of something like ScopeGuide that's been around forever, you know, if you really want to train a fellow to see what a loop looks like or just being cognizant of how much scope you have in. I think that for artificial intelligence, all the things that are mentioned, quality metrics in terms of surface area exposure, I think that's critical. And you'll never maximize AI if you don't have an endoscopist that can offer optimal surface area exposure, whether it's on a naked scope or using an enhancement like a mechanical device, whether it's a cap or, you know, a permanently fixed balloon to really maximize, again, surface area exposure. Thank you. I can think of six more questions, but we do have to get to break. So, if it's all right with everyone, we'll take a really quick break and we'll come back for the third session at 11.30. Thank you all.

artificial intelligence

physicians

patient care

quality metrics

training