Thank you, Dr. Nam. We will start the symposium with a lecture presentation by Dr. Alessandro Repici and the title of his lecture is AI in colonoscopy, useful or just cool? Dr. Repici, you're live. First of all, good morning everybody and welcome to this ASGE meeting. I'm so glad I want to thank ASGE and the course directors for inviting me and for giving me this opportunity. So as you can see here, the topic of my presentation will be artificial intelligence in colonoscopy, it's useful or just cool? These are my disclosures and some of them may be relevant for the topic I'm discussing later on. I will start with a couple of numbers. So colorectal cancer facts. This is the forecast for the next 30 years. As you can see, colorectal cancer is expected to remain stable as a third cause of cancer in the Western world and the second cause of death from cancer at the global level. There are a couple of numbers that are very difficult to interpret because if you look in detail in the last about 20 years, we have seen a decrease of incidence of colorectal cancer by at least 1% each year. But indeed, this phenomenon is mostly related to the older population. It's masking the rising incidence of colorectal cancer in younger population. If you look to the data about younger population, younger than 55, so in the last 10 years, they had an increase of incidence of colorectal cancer by 2%. And if you look between 50 and 64, it's 1% increase for the incidence of the cancer. So this is a controversial discussion because in the meantime, we are doing a lot of colonoscopies. So if I remember well, the numbers of colonoscopies in US are around 18 or 19 million per year. And based on this nice paper from UK, if you see, they looked at the main changes in endoscopic procedure across four years' time. They were able to demonstrate that most of the procedures that are endoscopically performed in UK, they are around colonoscopy of exploration of the distal rectus in more. So what is wrong? So we have rising incidence of colorectal cancer, especially in the younger population, in population until 50 or 60 years old. And we are using a lot of resources, we are doing a lot of colonoscopies. So there must be something which is not going in the right direction. Probably we have to do more procedures. Probably we have to make a better selection of patients because sometimes we do too much follow-up procedures and we are not including in our screening program the younger population they should be. And also there is an issue with the inefficiency of colonoscopy. So it's been well reported in the last 15 years, the issues of adenoma detection rate variability and adenoma miss rate. Indeed, colonoscopy should be a very perfect diagnostic tool because the impact on in the prevention of colorectal cancer is so great. So it's a shame that still we are discussing about adenoma variability and the adenoma miss rate. And this is going to be a present discussion, it's not just the past. This is a very nice systematic radio meta-analysis published in Clinical Gastroenterology reporting the association between endoscopy and colonoscopy outcome and endoscopy specialty. But if you look to this data, about one-third of all procedures reported in this systematic review, there are no gastroenterologists. If you can see here, most of the studies, they come from North America, US and Canada. And what is astonishing that when they analyze the data, the main findings were that when colonoscopies are performed by surgeons, there was a lower adenoma detection rate, a lower secal intubation rate. But even worse, when the colonoscopy were performed by non-gastroenterologists and non-surgeons, so the endoscopy, the colonoscopy outcome was terrible with even lower adenoma detection rate, very high perforation rate. And they stated, the authors stated that it was high rate of post-colonoscopy colorectal cancer. And we are talking about one-third of the entire colonoscopy reported in all studies summarizing this nice systematic review. So it's important that we understand why we are missing polyps, why adenoma detection rate is so valuable. So I invite all of you to read this very nice practical paper by Dr. Rexis, Impressing GIs, Top Tips for Maximum Detection During Colonoscopy Withdrawal. So there are two main reasons that we have to consider when we're missing polyps. One is incomplete mucosal exposure, and the other one is failure recognition of exposed lesions. So for the first point, it's clear that we have a number of different devices that can be used to improve the exposure of the mucosa, especially in certain anatomical locations. There are a lot of papers, there's been a lot of papers, a randomized trial about these devices with very promising results. But the latest publication this month in clinical gastro and the pathology, they are reporting very disappointing data. So this is a multi-center and U.S. study led by the group of Pradeep Sharma. It's just, I think, Kansas, Indiana, and Cleveland. They reported in a randomized trial that ADR did not change when they compared two different devices to improve the mucosal exposure, a standard colonoscopy. So they looked at the adenoma detection rate, even adenomas per colonoscopy didn't change. Same for advanced adenomas, say radial lesion, and Y-colon adenoma detection rate. So what does it mean? Probably the device helps in exposing more mucosa, but still we're missing polyps. So what about AI? How can AI help in failing the recognition of polyploid lesion or those flat, they are more difficult to be detected. So if we consider the AGI disease are mostly diagnosis, assessed, and monitored endoscopically, this is the right place for AI computer vision, and more specifically for colonoscopy. Colonoscopy is an invasive, operator-dependent, real-time examination that you cannot repeat. And again, this is a special place where computer vision AI can make a difference. So a few slides about history of artificial intelligence for colonoscopy. About 20 years ago, the first paper just describing a computer vision system to try to identify polyp and characterize the shapes of the polyp. And afterwards, the first clinical paper came in endoscopy 2016 and was from a Spanish group. Afterwards, there has been a booming of publication with very nice paper, a lot of well-done studies that have been reported, especially in the last two years. So what are the available systems right now? This is the AI universe you see for colonoscopy. There are so many companies that have invested money, resources, and studies, and technology to develop systems. Some of them are already FDA approved. I want to mention that Medtronic was the first with AI genes. And afterwards, the AI vision has been approved. Recently, I can predict that very soon a couple of other systems will be approved. So most of them right now, they have CADHE for polyp detection, but three of them, GI Genius, CADI from Fujifilm, and the NEC system, they also have already approved at least in Europe the CADHE system for polyp characterization. Just a single video to show you what can be the impact. So on the right side is the video with artificial intelligence. I was copying these patients with no symptoms, a screening colonoscopy, primary screening colonoscopy. So the system is alerting me there is something wrong in this area. So I was focusing myself, and I was able to recognize this was quite large serratus style lesion into the right colon. What is nice is that when you are familiar with artificial intelligence, it works under different conditions. So even though you are using chromoendoscopy, any kind of chromoendoscopy, artificial chromoendoscopy, or a standard conventional chromoendoscopy, still the system is very good, is working well, helping you in detecting lesions. If we should give you a definition based on my about three years experience adopting AI for colonoscopy in my practice, I would say that AI is at least for detection is precise, reliable, consistent, limitless, works all the time, it's a very friendly user. So from a clinical standpoint, all these items, they translate to mean increased ADR, increased adenoma per colonoscopy, reduced means rate. In general, we are not prolonging the procedure too much. So very equal, very similar operational time when we compare with standard colonoscopy, a very minimal false positive rate. There has been a lot of randomized trials in this field. I think that we look to the colonoscopy field, AI has been evaluated very carefully, with several randomized trials. Here I presented the data, I would say the four or five studies that the most important have been published in the last two years. So two papers from our group, one with expert operators, the other one with non-experts, a paper from a Chinese group published on oncology with a double blind operator, another paper lead by Mike Wallace with the tandem modality, and finally a very recent paper on gastro by the group of Hadma Shaukat, again with tandem modality. So all of them were very positive, all of them, they basically demonstrated a significant increase of ADR, significant reduction of means rate, a significant increase of adenoma per colonoscopy. So in all of them, so I could not see a randomized trial with a negative impact of AI system for detection of polyp, whatever is the methodology that's being used to make the trial. There's been two meta-analyses, the first one by the group of Michael de Tauer and Yuichi Mori, published two years ago in endoscopy, they demonstrated there was an increased detection of adenoma, but no impact on advanced adenoma, and also our group meta-analysis in GI endoscopy last year, we're working on a third meta-analysis because there are more data coming and our feeling is that AI increases also detection of serrated lesions as well in the right colon. So what about artificial intelligence plus endocap? So as I said before, endocap is expected to increase mucosal exposure and AI is expected to reduce the failure of lesions recognition. So we just completed, these are very, very preliminary data from our group, we are making very detailed analysis, but just to anticipate the results of the study, in our multicenter randomized trial with more than 1,200 patients, it's a multicenter, so academic and non-academic center, we were able to demonstrate when you put endocap on top of AI, endocap is exposing better and AI is helping the endoscopist in the recognition of all lesions, whatever is the size of the morphology. So AI plus endocap was superior, either in terms of adenoma detection rate and adenoma per colonoscopy. So very final slides, personal consideration, I think what makes AI different from other technologies that we have adopted in endoscopy, this is going to bring humans and machines working all together, this is very new for all of us. And also I think that very frequently AI is framed as a machine replacing humans. I think this is wrong, it's not machines replacing humans, but it's machines supporting humans, augmenting our power, our capability, our diagnostic precision, so we can make more diagnosis, we can be more consistent, more precise. Human and machines, they have different strengths and weaknesses, and I think that just putting the combination of machines and humans working together, this will allow us to democratize colonoscopy outcome. It's rather unacceptable right now to have low detectors and high detectors with some patients being exposed to an increased risk of colorectal cancer as compared to those who have been discovered by high detectors. So in conclusion, we have colleagues, useful for sure, but also very cool. Artificial intelligence definitely improves colonoscopy efficiency by detecting more adenoma and reducing adenoma misread. Future direction includes automatic assessment of histology, polyps size, polyborders, evaluation of bone prep, anatomical landmark, and also personalized risk stratification. Thank you all for your attention.

artificial intelligence

colonoscopy

colorectal cancer

polyp detection

adenoma detection rate