So, so thank you everyone for coming here for for lunch and listening to me talk a little bit about my experience with GI genius. We'll get started here, would like to leave a little time for questions at the end and and things of that nature but though before I start, um, who has GI genius, you know, in the room and their endoscopy unit. Does anyone. All right, in the back. Okay. Alrighty. Right. These are my disclosures. Electrical risk. Okay, so I'm just real brief kind of outline I'll give kind of a quick personal background sort of terms of who I am and where I practice and the setting of practice and talk a little bit about you know artificial intelligence and why now, and the rationale that you know behind implementation at least rationale that I use for bringing it to my center. We'll go over some kind of cornerstone studies with GI genius, and then I'll just review some videos and case studies, and then go over some just kind of basic AI misconceptions. I work in a tertiary care center in North Philadelphia, and anyone who's familiar with North Philadelphia. It's certainly, you know, there's has its fair share of healthcare inequities and healthcare disparities. So I'll bring that to a to a point of in terms of my rationale for for GI genius in my community here in a second. I'm also a fellowship program director. It's kind of a new role for me, taking it on this academic year, and certainly a passion of mine, and something I've been enjoying enjoying doing so far. So just to kind of give an overview of what our unit looks like for your for everyone. We have four rooms, one is dedicated to advanced endoscopy advanced procedures and then three more kind of general GI rooms. We now have in our general rooms, all GI genius. I got my first unit in November of last year, second unit in March we've been kind of slowly rolling out actually I just got my third unit on Thursday, installed. So, and that was through the health equity assistance program through ASGE Medtronic and AWS. So certainly excited to have full AI capabilities. So, you know, why now artificial intelligence and GI in particular. So, you know, I think everyone in the room is familiar that, you know, we, the care we give for our patients are complex, both procedurally and certainly, you know, are on the endoscopy side of things we rely heavily on visual data, and a lot of it. We also have a growing tool set growing patient population more patients to care for. And we're also outcome driven too and these are all kind of qualities of the field that really make it a target rich environment for AI. There's also been a lot of technological advancement over the years and, you know, without kind of going too much into the history of it. You know, in the last, you know, 10 years on the research side of things for machine learning deep learning, but also on the hardware side with implementation of the GPUs. I mean, you know, for an average colonoscopy taking 30 minutes about 270 gigabytes of data, that's a lot of data so it's going to require a lot of tech to handle all that in real time. And we're now at a place where we can do that and do a good job of it. So, in terms of implementation of AI in our endoscopy unit, there's kind of three rationales I used when I was bringing that number one of course is, you know, colorectal cancer prevention that's something that we all do and a very near and dear to our heart. You guys know all this but it's, you know, a third leading cause of colon cancer in the United States, second most common cause of death and, you know, mortality is rising people under the age of 55, and hence the fairly recent decrease in age of starting screening at 45. We see racial disparity as well in terms of the incidence and mortality rates. And, you know, of course colonoscopies prevent colorectal cancer by removing polyps. But we know that kind of baseline miss rate for adenomas is approximately 25%, so we're not perfect at it nobody is. But it's certainly kind of a role that hopefully that AI can kind of fill, or a gap I should say. And of course polyp detection is crucial so for every 1% increase in ADR, you know, we can drop interval colorectal cancer risk development by 3%. So, hence the focus on ADR. The second kind of rationale is colonoscopy is highly, you know, operator dependent certainly experience and expertise being an important factor but there's other things that are totally out of our control, you know, we can control withdrawal time, certainly, yes. But there's things like fatigue, number of cases that you're doing day we know later in the day cases tend to our ADR tends to drop. And there's multiple reasons for that besides just fatigue there's focus there's physical pain in some instances if you're doing a lot of scopes that day or having a difficult day. There's technique, you know, looking around folds, you know, making sure you're covering the mucosa as best as you can. Things like team dynamics I mentioned time of day there's a little bit of a polyp fatigue phenomenon too as we take out more polyps in a case, we tend to, you know, our ADR, or the number of adenomas we tended to make drop or increase miss rate. From the patient centric side there's about preparation quality BMI demographics medical comorbidities that can play a role intra procedurally and how good we are at detecting adenomas. So that's our second rationale and more for my community and the community I serve, where there's a large number of minority patients that I see, we know, racial disparity is well known in colorectal cancer. Black patients have a 20% higher incidence rates of colorectal cancer compared to our white counterparts. Black patients are also tend to be diagnosed at a younger age and a later stage, and have a 40% higher morbidity rate versus white patients too. We also see kind of similar sort of disparity within sexes male versus female. So, you know, in my case, you know, every colonoscopy counts for all of our patients but particularly in my population this may be the only time I have a chance to do a colonoscopy, hopefully not, but I want to make it count as much as I can. And there's some various social drivers in my community that drive all that. And as someone listed here, you know, food insecurity, health literacy, education, income rate, annual income, things of that nature. So, just to kind of shift focus more towards GIG and the system overview, you know, it was first to market globally, came here in the United States in April 2021, and validated using over 13 million images, and this data that set that was collected globally as well across 20 different centers. There's a little bit, or there is flexibility too, and certainly evolution is possible as data set expands and additional training for the AI system is obtained and certainly can be upgraded in the future. So, the system is very sensitive. You can see close to 100% sensitivity, less than 1% false activations, and it's fast too. It's about 82% faster at detecting polyps than the endoscopist. I think even with the newer version, it's more like 87%, version two, which is impressive. I know at times, personally, it's definitely beaten me for sure. And this is just kind of what the system looks like and how it kind of plugs into, you know, the tower. It's compatible with really all endoscopic processors, Olympus, Fujifilm, Pentax. It's kind of about the size of a, aging myself here, a VCR. It's about that big. I mentioned that to my fellows and they kind of looked at me a little weird when I first said that, but it's very small. It's very low profile. It doesn't take up much room in our towers. And it's, you know, we've seen plenty of pictures today, but it's basically an overlay over your image with that green bounding box highlighting, you know, polyps and things to focus your attention to. So going into just kind of the sort of the cornerstone literature review, there's three kind of main ones that I want to sort of highlight. Dr. Rappicci having the first two here and sort of laying down the initial sort of groundwork for things here. You know, we're seeing a high ADRs compared to control that's for CAD-E versus another control beam, you know, without it. There's certainly importantly no difference in withdrawal times between CAD-E as well as, you know, the control. You can see they're exactly the same, you know, really shine. The system really shines with, you know, polyps less than a centimeter in size. And I know those are kind of the polyps that keep me up at night and probably keep a lot of you guys up at night. It's not the real big ones, it's those small ones. And I'll go into a little bit in terms of location in the colon where those polyps are detected. So second paper here by him and his team kind of exploring physician experience and influence on CAD-E effectiveness with GI Genius, kind of defining, kind of separating expert and non-expert based on number of colonoscopies at what the cutoff being at 2000. And, you know, again, we're seeing, you know, high ADRs compared to control, but we're seeing that both in non-expert endoscopies as well as in expert endoscopies as well. And that's what I have personally seen. I've seen my ADR with CAD-E with GI Genius increase about 11%, which is sort of similar to what his group found here. And again, these were kind of polyp, you know, characteristics being really less than a centimeter in size, non-polyploid, and no impact on withdrawal time, which is great. Dr. Wallace here back in March released his tandem colonoscopy design study, kind of highlighting or looking at adenoma misrates with a reduction of approximately 15% using GI Genius compared to control. And seeing kind of seeing that both in proximal colon polyps as well as distal. Again, adenomas less than 10 millimeters in size, non-polyploid, that's where things shine. There was a little bit of underpowering in a per polyp analysis to really show any kind of difference on per patient ADR to highlight APC, but there was definitely a trend to increase adenoma per colonoscopy, or increase adenoma detection per colonoscopy. So I just wanted to kind of show you guys a few cases here just to kind of go over some videos. You know what, I'm just gonna, let me just, there we go. Okay. So this first one is kind of the one that sort of most impressed me when I was kind of going through some of these videos. So on the left is going to be with the GI Genius system on, on your right, no CADI. And this polyp is detected pretty quickly under fluid. I still kind of have a little bit of difficulty catching, finding it until it's right up at the very end here. So very sensitive, very impressive example of just how GI Genius shines. Let me go to the next one here. I don't want to go back. Okay. It's a very subtle lesion that gets picked up here, flat, less than 10 millimeters for certain. Again, kind of very sensitive, very quick to pick it up to really kind of highlighting how fast the system is. And then one last example I wanted to show you guys, this is just how the system is able to detect multiple polyps at a time. You know, I've seen it, you know, once kind of counted upwards of like 10 of it at a time in the rectum, which I thought was kind of impressive. There's no, there was no slow down. There was no bogging. I haven't encountered any of that at all. Okay. So just kind of in terms of my personal experience in implementing it within the, within the R suite, installation is very simple. You know, this week, I think it took maybe 10 minutes to install it. It's like kind of just plugging in video feed, plugging it out, you know, running a wire out and then hooking into pattern. I know on our, you know, tech side, the learning curve is minimal. I think there's really only two buttons, you know, you need to be aware of. There's, you know, the power on button. And then there's kind of like a pause button where you can kind of set the system to like sleep mode that you can kind of turn off and on during the actual procedure, not the whole machine. There's a little bit of a boot up sequence when you do turn it on. But that pause button is instantaneous. I haven't run into any like technical what I'm calling hiccups with the system. I haven't noticed like, as I mentioned, slow down, freezes, crashes, anything along those lines. I mean, it goes, we have it on all day and have not run into any issues. There are a couple of other buttons on the system that are more, you know, there's kind of some sub menus. You can turn the volume on whenever it detects a polyp and there's various, you know, I mean, sound on when it detects polyp. You can increase and decrease the volume as you choose. I have my volume kind of low. In fact, it's off whenever I use it because I like to hear, you know, pulse oxes and, you know, communication in the room and things like that. But that's certainly up to endoscopist discretion. Just in terms of feedback that I've gotten kind of by my colleagues and trainees, you know, and I would say there was maybe some initial hesitancy with my partners in using it, using the AI or GI Genius, but it has now morphed into daily usage. I mean, you know, they enjoy it. They like it. They insist on having it on. You know, I kind of mentioned there was sort of a slow rollout in our endoscopy unit only based off the number of units of Genius that we had. So people would kind of fight over the room a little bit that GI Genius sat in, but now we have it everywhere. So it's not a big deal anymore. You know, my colleagues have not expressed any withdrawal time concerns. They haven't noticed any difference about that. They did kind of mention initially there was some distraction with the green box, but they've almost kind of trained. They described to me that they sort of trained their eye to focus on what type of bounding boxes being displayed. You know, occasionally you may get kind of little bleeps, you know, here and there, but it's really the one that it kind of locks on to that they have drawn their attention to. So that once when it is locked on to a pod, that's probably worthy of your investigation. Of course, you know, at the end of the day, you know, you're still quarterbacking it. You know, the system is not telling you what to take or what not to take. It's certainly no substitute for good endoscopic skill set. So I think that kind of has made, you know, certainly myself, but my partners comfortable with using the system. There's no control, but they feel like they've given up on. And certainly they're curious about future applications. Now on the training side of things, you know, right off the bat, they've described like a lot of significant interest in AI and kind of where the future heads, which I think is very promising. And certainly, you know, they have expressed, you know, desire for deep understanding of AI, which is something, a curriculum that I'm currently putting together. They've sort of described a few things that I think, you know, Raj mentioned earlier, they kind of describe almost eye training. So when they're withdrawing, you know, the system kind of alerting them to various areas, they feel like they've kind of been using that alert to kind of train their eye when they're looking around during withdrawal, which is neat. And then also what I kind of call polyp training as well. So not only are they focusing on, this is what they tell me, focusing on the green box, but what the green box is highlighting and like what a polyp is. And especially for the early trainees, I think that's important for them to learn. So they're kind of learning polyp morphology, polyp type, things of that nature. And then I think Raj mentioned this earlier, polyp tracking during interventions. So I think, you know, for those who have fellows and they're about to do an intervention for like a polyp removal, let's say the polyps at 12 o'clock when they make that rotation, we're all kind of like, oh boy, here we go. We're about to lose the polyp. But with G.I. Genius, it makes it very easy. And I think it helps them kind of build skill sets and techniques for orienting the scope properly, not having so much to, you know, be concerned about losing the polyp at the same time, which has been kind of an added benefit that they've described to me. So just in terms of misconceptions, I just want to kind of talk about real briefly, you know, AI will not replace physicians at all. It's going to be a tool to augment our clinical judgment and also complement health care delivery, which has been, I think, highlighted throughout the day. You know, I think ultimately the goal is going to be to improve outcomes and value, decrease variance in care, data management, maybe even outsource treachery in the future, which is exciting. Misconception number two, AI will fix everything. No, that's not the case. Far from it. But, you know, I think it's important, you know, when you're engaging my colleagues that I've leveled expectations, that I, you know, address any sort of biases with AI where possible and things of that nature. I think the last misconception I want to bring up is AI is hard to understand and implement. I mean, with the GI genius system, that's far from the case. And I think that's where education plays a large role in it for providers. Kind of speaking about AI education real quick, you know, everyone's here today, so you guys are at the forefront of it. But certainly, you know, I think implementing societal resources, ASG leading the way, is an excellent step in AI education. There's also the American Board of AI and Medicine, which is not GI specific, just AI in general. They have a lot of teaching and learning resources for physicians and providers. And then, you know, certainly education at advocacy, I think is going to be important for AI rollout, both on the fellow level, as well as the patient level. You know, I've had discussions with my patients in the clinic prior to them getting a screening colonoscopy, what AI is. Some of them have expressed a lot of interest in it. Some of them a little bit of indifference, as long as they're safe and getting a good colonoscopy. But I certainly feel there's a lot of role for that, which is something I try to do as much as I can for patients. And then I do want to kind of just direct everyone to the Genius Academy. This is something offered through Medtronic, which really kind of goes over artificial intelligence, you know, basics, physician perspectives. As well as some really nice podcasts given by a few individuals in this room on delivering quality care and AI, embracing the future, societal implications, integrating AI into GI training. It makes for a nice listening and certainly good learning. So thank you very much. I'd be happy to take any questions from the audience about the system. All right. Well, enjoy the rest of the day. Thank you for listening. I'm going to leave this up for everyone. Thank you.

GI Genius

artificial intelligence

endoscopy units

colorectal cancer prevention

polyp detection

adenoma miss rates