from San Francisco. I'm going to be speaking on the management of gastric varices and I will argue in favor of an EUS guided approach. Let's start with a little history here. My mentor Nibs Sahendra at the University of Hamburg, University Hospital Eppendorf in Germany, was the first to treat large esophagogastric varices using the sinoacrylate bucrelate and that was back in 1986. This was reported in the journal Endoscopy as a new method. So sinoacrylate is a very unique substance. It is a liquid with a consistency similar to water and when added to a physiologic medium like this agar plate of blood, it rapidly polymerizes and forms a rock hard substance. And when we inject this into varices, gastric varices here in the fundus, typically in a retroflex view here, we can plug up the lumen of these varices replacing the blood for sinoacrylate. And what we get is a sinoacrylate glue plug. You can see when we mix the sinoacrylate with a contrast medium such as lopidol, we can visualize the convolute of communicating varices that are now filled with the rock hard sinoacrylate. And this is a histologic cross section showing the glue plugging up a varice lumen. And over a period of weeks, the glue is extruded as a cast into the lumen of the stomach and then it's passed spontaneously. So none of the glue stays in the body. It is expelled as a foreign body. The dreaded complication of glue, and this is known to all physicians who use glue in the many medical specialties, be it interventional radiology or vascular surgery and interventional endoscopy, of course, is systemic migration. I used to call this systemic embolization, but in fact, it's confusing because we're using the glue as an embolizing agent. The emboli, the migration of the glue, can go to virtually any part in the body, wherever the blood flows to. When we inject it into gastric varices, it will most commonly go to the lungs. You can see that here. Or to any of the neighboring abdominal organs. It can get into the portal vein here. It's actually occluding the branches of the portal vein into the spleen. It can cause splenic infarction. It can get into the kidneys here. You can see it in the renal vein. And if a patient has AV shunts or a patent for amylovali, it can even get to the brain and cause paralysis. This is a compilation of some of the reports that have been published in the literature. I don't think this is complete, but it's a nice table. And it summarizes 27 reports with 11 organ infarctions, 7 deaths. So you can see here pulmonary and splenic infarction. Here pulmonary infarction, myocardial infarction, cortical blindness. So this actually got through some AV shunt into the arterial circulation, splenic infarction, cerebral infarction, pulmonary infarction, DIC. Some of these patients became septic. Splenic infarction, pulmonary, splenic, splenic, cerebral infarction. So to say the least, this is very humbling to see severe complications sometimes fatal like this. So this is why we endoscopists who have been using glue and have all experienced complications of glue migration, that we have been thinking about how we can reduce this risk of migration. Ideally, we want to eliminate it, of course. And this is a report from the Seville group in Spain. And this is the first time that EUS is being used to guide the injection of sinoacrylate. And the authors are targeting the perforating feeder vein. So the rationale here, the goal, is to block the feeder to the gastric varix with a lesser, a lower amount of glue. Five patients, histoacryl was used, the MDU2 sinoacrylate, and total amount of glue used was 1.6 milliliters, which was much less than what the authors had historically been requiring for glue obliteration of varices. And they had great results. An alternative strategy using EUS for guidance and is required here would be to inject coils rather than the glue to obliterate the varices. And this is a case report for patients only, proof of concept, targeting again the perforator vein with decent results here and there's proof of concept. But they did require a large number of coils to obliterate the varices. So mean of nine, but up to 22 coils. To follow was a multi-center trial in Europe, six centers, so 30 patients. Now this is a retrospective analysis. It's not randomized, but it compares the use of glue injected under EUS guidance, targeting the perforator vein, replacing the coil, again requiring EUS guidance, into the perforating vein. And for glue they used a mean of 1.5 milliliters, similar to what you just heard, but they got a CT scan in all the patients and they noted lung emboli in nearly half of the patients. Now fortunately these patients were asymptomatic at the time. In a third of the patients who got coils, analyzed patients, again targeting the feeder, a mean of six coils per patient, up to 13 coils. But 18% of these patients still needed glue to achieve complete obliteration of the varices. And what's interesting is that the CT scan in these patients, and they all got CT scans, showed no lung emboli. And so that raises the question whether placing a coil before injecting the glue may reduce the risk of migration, or perhaps even eliminate it. So these coils, you can see that it's a metal alloy with these wooly fibers attached to it, and this fills the lumen of the varix and it reduces flow in the varix. But more importantly, these wool strands serve as a scaffold to retain the glue at the site of injection. So this is here a ex vivo demonstration of how this works. We've injected the glue after placing the coil, and you can see that the glue is adherent to the coil, and there was no residual glue left in the container of heparinized blood. So this suggests that we can avoid the risk of migration. Here you see a video of how we use it. This is an impressive case, 45-year-old male with hematemesis. He's undergone multiple band ligations for esophageal varices. So you can see here in retroflexion view, these large IGV1 varices, so these are isolated bundle varices, and the Doppler confirms the blood flow, and we are now puncturing under US guidance using a 19-gauge needle, the varix lumen, and we are deploying the largest coil that we have, which is 20 millimeters. And you can see the coil now coiling up, deploying inside of the varix lumen, and then we follow that immediately with an injection of the glue. And the idea, again, is that this glue should adhere to the woolly fibers, and here you can see nine months later, the varices are completely obliterated. Now the glue and the coil are no longer seen on EUS inside of the varices. I'm not showing that here, but it is entirely extruded and passed. And we published our long-term results in 2016, EUS-guided treatment of gastric bundle varices with combined injection of coil and glue, and we had great technical success, just one failure. We had long mean follow-up, 436 days. Our re-bleeding rate was only 8% for gastric varices, but altogether was 16%, so that's quite good compared to what's reported otherwise in the literature. And we had few adverse events. We had four patients that had pain that self-resolved. We had some bleeding. Now this is not late bleeding, this is bleeding that was occurring after the injection, so we required additional injections. And then we had one patient who did develop a pulmonary embolism, but it was a week after glue injection. The patient had already been discharged home, so there was no evidence for any migration, glue migration, prior to discharge. So we're just not sure this is truly due to the glue, but certainly counted as an adverse event. So this is a trial now that for the first time is looking at the outcomes of EUS-guided versus conventional. So we want to compare how we were doing it previously with an EUS-guided approach, in this case, using coil and glue. And it was what I would call a pilot randomized controlled trial. Only 32 patients randomized. The rate of thrombosis was similar in both groups. The rate of pulmonary embolism was higher in those patients undergoing the conventional approach, just endoscopy guided, 50%, and that's similar to what was found in the European study. In those undergoing the EUS-guided approach, coil and glue, 25%. So 25%, nonetheless, but it is half, although this did not reach statistical significance, so it could be a type two error. This is a nice meta-analysis that is fresh off the press, comparing an EUS-guided therapy versus direct endoscopic glue injection therapy. So we have 23 studies that qualified for EUS-guided approach. So all modalities, whether it's glue alone, coil alone, or hybrid, and then the conventional endoscopic, 28 studies, almost 3,500 patients. There was a significant benefit for the EUS-guided approach in terms of recurrence of gastric avarice. So this was statistically significant, 9% versus 18%. But otherwise, similar results. This study from Indiana University randomized patients to either EUS using just glue versus the direct endoscopic approach. So I like this study because it's really a better apples to apples comparison. So we're just injecting glue, but we are differentiating EUS guidance versus non-EUS guidance. And what they found that EUS guidance was superior in that they required less glue to achieve obliteration. That was statistically significant. There was a lower rate of gastrovariceal rebleed, 9% versus 24%. And there was a lower rate of non-gastric variceal rebleed, 11% versus 28%. So this is quite significant. The rate of adverse events, that would of course include migration, was similar. So this is the meta-analysis to go back to this, just comparing the different EUS-guided modalities, glue versus coil versus the hybrid coil and glue. And this found that the recurrence of gastric varices was significantly lower when using the combined hybrid approach of coil and glue. And also the rate of late rebleeding was significantly lower, basically half. Here it's actually a third of the patients have recurrence of gastric varices using the hybrid approach. So suggesting that this may be a superior approach using coil and glue versus just glue alone. I would like to conclude by reviewing the AASLD practice guidelines, first from 2007. Recommendations in patients who bleed from gastric fundovarices, endoscopic variceal obturation using tissue adhesives, such as cytoalkylate, is preferred. Clear statement. Tips should be considered in patients in whom hemorrhage from fundovarices cannot be controlled, or in whom bleeding recurs despite combined pharmacological or endoscopic therapy. So tips is backup. Now we go to 2016, the updated guidelines, and now it's stated tips is the treatment of choice in the control of bleeding from cardiofundovarices, GOV2, and IGV1. And the glue is an option for cases in which tips is not technically feasible. So now the glue is the backup and the tips is the procedure of choice. And that's a bit baffling. But the Bovino 6 consensus guidelines from 2015 clearly state that glue is the procedure or the modality of choice. Endoscopic therapy with tissue adhesive is recommended for acute bleeding from isolated gastric varices and those gastroesophageal varices type 2. These are the GOV2 that extend to the fundus, extend beyond the cardia. EVL or tissue adhesive can be used in bleeding from gastroesophageal varices type 1. So these are the more ropey varices that extend into the stomach along the lesser curvature. So there's some discrepancy here, and this is clearly an area that continues to be debated, including whether we should be using EOS guidance versus endoscopic, direct endoscopic. And if you're using EOS guidance, should we be using glue alone, coil alone, or a hybrid procedure? And our choice right now is the hybrid procedure. Thank you very much. I would like to thank the AASLD as well as the ASGE and also Drs. Levy and Telford for inviting me to give this debate presentation on management of gastric varices. I've been charged with talking about standard endoscopy, otherwise known as non-EOS endoscopy, specifically for treating bleeding gastric varices. My name is Richard Wong. I'm a professor of medicine at Case Western Reserve University, University Hospitals Cleveland Medical Center in Cleveland, Ohio. My one disclosure is that I do receive academic research grant support through Case Western Reserve University. I'm also a consultant for Vascular Technology Incorporated in Nashville, New Hampshire. This talk will be divided into five sections. The first one will be a short video showing our technique. The second one will be a description of the different types of glues available. The third one will be a description of our non-EOS technique of glue injection. And our fourth section is the debate as to why standard endoscopy is better than EOS. And the fifth and final session is conclusions. Let me show you a quick video of our non-EOS standard endoscopy glue injection technique that we use. This is a patient who came in with major severe upper GI bleeding found to be from gastric varices. He is located in the medical intensive care unit. You can see here that at standard endoscopy, we see that he has a gastric varix with a fibrin clot in the cardiac. We use the Doppler probe to determine subsurface blood flow. Listen. The continuous low pitch hum is indicative of active subsurface blood flow. It is a gastric varix. We then inject a glue. As you can see, we get some immediate back bleeding. The white material is glue. Do not suck that back into the endoscope as that might clog up the endoscope. We inject glue, 0.5 mLs, in two different areas here. There is some self-limited back bleeding, which stops. We then use the Doppler probe post-therapy, post-glue injection. The Doppler ultrasound signal is now absent, indicative of absence of subsurface blood flow. In other words, the varicose has been thrombosed and we don't need to inject anymore. Let me talk about the different types of cyanoacrylate glues. This is actually very important for endoscopists to know. So anybody doing this should know the different types available and choose the right type. So not all cyanoacrylate glues are the same. This may seem surprising, but they are quite different. Theoretically, a faster glue polymerization rate may reduce the risk of systemic glue embolization. Systemic glue embolization is the most feared complication of glue injection. And you want to choose a glue that polymerizes quickly. In other words, you want to choose a glue that polymerizes in the gastrobarics quickly without having time to circulate systemically and cause embolization elsewhere. Therefore, it's important for the endoscopist to understand that the polymerization rate of glue is affected by many factors, including the specific isomer of cyanoacrylate that is used, by certain additives that may be used, by the technique of injection, and by the patient's plate count. This has been nicely studied by Stephen Caldwell and his colleagues in this article here. Medically available cyanoacrylate glues include the four versus eight carbon groups. The four carbon glue is the N-butyl 2-cyanoacrylate versus the eight carbon or the 2-octyl cyanoacrylate. The N-butyl one polymerizes in five seconds in the undiluted form, and the eight carbon or the 2-octyl polymerizes in 16 seconds in the undiluted form. Therefore, you have much faster polymerization with the four carbon or the N-butyl 2-cyanoacrylate group than the 2-octyl cyanoacrylate group. So it's important to choose one that is faster polymerization. Some endoscopists add certain oils to the glue in a one-to-one ratio, such as lipodyl or ethiodol. It is important to know that if you add these oils, you dramatically slow down the polymerization rate of the glue. In addition, severe thrombocytic penea, if a patient's patent counts under 20,000, it also slows down polymerization of the glue. Let me describe now our technique for non-EUS standard endoscopy. This is the equipment needed for non-EUS glue injection. We use a specific glue consent form. It is important to know that glue is not FDA approved for use in the US. We use a diagnostic gastroscope. We use large caliber injection catheters, 21 to 23 gauge. We need to know the precise volume of the dead space inside the injection catheter because we use this information to decide on the volume of chaser fluid. I use the fast polymerization N-butyl glue, the medical grade undiluted. Again, this one polymerizes in five seconds or less inside the varix. I have sterile five ml syringes. I use olive oil, single use sachets. I have eye protection available for the endoscopist assistance and for the patient in case of glue splatter. An optional approach and optional adjunct that we use at my institution is the non-EUS through the scope endoscopic Doppler ultrasound probe. Our technique, we first of all perform an EGD to identify the gastric varices as a source of bleeding. We then identify the culprit gastric varix for glue injection based on stigmata of recent hemorrhage. Our aim is to treat the culprit bleeding gastric varix at the first session. And generally this is a single bleeding varix. We use the optional adjunct of the Doppler probe to demonstrate venous blood flow, but this is optional. We get everything ready and syringes labeled and draw up glue at the last moment. This is important because glue can polymerize in air as well. We flush 10 to 15 mLs of culinary olive oil through the accessory channel of the endoscope into the duodenum. The belief here is that the oil coating the channel will decrease the risk of glue sticking to the channel. We then decide on the volume of glue injectate based on the size of the gastric varix. If the gastric varix is small, we inject 0.5 mLs, medium size, one mL, large varix, two mLs. We then decide on which chaser fluid to use to push the glue into the varix. We can use more glue versus sterile water for injection versus normal saline. Volume of the chaser fluid equals the dead space volume of the injection catheter. The speaker prefers to use glue as a chaser fluid. For example, if the dead space injection catheter volume is one mL and you want to inject one mL of glue into the varix, then you would inject a total of two mLs of glue with one mL of glue remaining inside the catheter. The endoscopist expresses one drop of glue at the needle tip then quickly punctures the gastric varix with a needle and inject glue by rapid intravaricyl bolus followed quickly by rapid bolus of the chaser fluid to push the glue into the varix. Do not suction. This is very important because if you suction, you might suck the glue into the channel and ruin the scope. Then you retract the needle. You can expect some immediate back bleeding, but this is self-limited. I wait two minutes, then determine whether or not the varix is thrombosed by, there are two methods. The conventional method is that the endoscopist probes the varix with a blunt tip of the injection catheter. If it's soft, that means there's still active blood flow. If it's hard, that means it's thrombosed. The optional method, which I use, is to use the Doppler probe looking for subsurface active venous flow versus diminished or absent blood. If the varix is still soft, or if there's still strongly active venous blood flow by the Doppler probe, consider injecting additional glue into an adjacent area. The maximum total volume of actual glue injected into a gastric varix per session is three mLs. I use a brand new injection needle for each injection because these needles clog up very quickly with the glue. Continue IV octreotide infusion for two to five days, then initiate non-selective beta blockers. Continue standard dose, one steady PPI. First surveillance EGDs at one week, then every two weeks until the culprit leading gastric varix is thrombosed. Subsequent surveillance EGD intervals, I follow the regimen of post-band ligation for esophageal varices. During the surveillance program, it is important to treat other medium to large gastric varices, but there is no need to treat small gastric varices. The endoscopist should expect a glue cast to form at the sites of glue injection, and the glue cast will eventually be extruded into the stomach and through the GI tract. This extrusion can expose a scar or ulcer that can sometimes bleed, but this bleeding is mild. It is important for the endoscopist to monitor esophageal varices during surveillance because esophageal varices can increase in size after glue injection of gastric varices, and the esophageal varices themselves may need their own treatment like band ligation. For refractory gastric variceal bleeding, I use a temporary balloon tamponade, resuscitate the patient, then talk to IR about TIPS versus BRTO. Debate. Why stent endoscopy is better than EOS? Some people have said that EOS is superior to stent endoscopy. This is false. The superiority of one technique versus the other has not been proven. However, research studies are ongoing, which hopefully will answer this. There have been two proposed advantages of EOS. The first one is that when there's too much blood in the stomach, the EOS can still see varices through the blood, whereas traditional endoscopy cannot. This is true. However, there are many ways to clear the blood out from the stomach before you perform stent endoscopy. The second proposed advantage of EOS is that the EOS can see and treat subsurface vessels that feed the gastric varices. This is true. However, treating the subsurface feeder vessels has not been proven to be beneficial. Some people argue that stent endoscopy is too subjective and that endoscopic ultrasound is much more objective. Well, we know that injecting too much glue can theoretically increase the risk of systemic glue embolization, which is the most feared complication of glue injection. We know that stent endoscopy is probably too subjective in terms of how much glue to inject. This is true. In other words, traditionally, we probe the varices after glue injection and we look for consistency. If it's soft, that means there's still active blood flow and we can inject more glue, whereas if it's hard, there's no blood flow, the varices thrombose and we don't inject any more glue. But we also realize that consistency is a subjective measure. It depends on how hard you probe, how hard you push with the probe. In our institution, we use an optional adjunct to stent endoscopy. Again, this is optional, but we use the Doppler probe to more objectively determine subsurface blood flow before and after glue injection and to titrate the volume of glue injectate based on subsurface variceal blood flow. How about equipment? Stent endoscopy versus EUS. Patients with acute gastric variceal hemorrhage are in the ICU. Stent endoscopy is highly portable with a travel card to the ICU. However, the EUS equipment is much more cumbersome and less portable to the ICU. Complications. Do these occur? Yes, while ruining the scope. That's a complication that has been reported in several cases when the scope, the accessory channel is glued with the sideracrolate. Fortunately, this complication is uncommon, but it is still better and less expensive to ruin a standard diagnostic oscilloscope than to ruin a very expensive EUS linear echo endoscope. What endoscopist training is required? Well, surprise, surprise. The advantage of standard endoscopy is that you don't need to learn EUS. Endoscopists who are trained in general therapeutic hemostasis can readily learn the technique of standard endoscopy glue injection with or without using the Doppler probe. But perhaps the most important point of why standard endoscopy is better than EUS is to talk about the natural history and prevalence of varicose bleeding. We know that cirrhosis affects 3.6 out of a thousand adults in North America. In Asia and other parts of the world, this is even higher. We know that varicose bleeding occurs in 25 to 40% of patients with cirrhosis. So it's common. We know that the mortality is up to 20% for each episode of varicose bleeding. So mortality is high. In this landmark study from Sarin et al, published in Hepatology in 1992, Sarin and his co-workers looked at the natural history of gastric varices versus esophageal varices. Although they found that bleeding is less common with gastric varices than with esophageal varices, the group found that when gastric varices bleed, the bleeding is much more severe with higher blood transfusion requirements and higher mortality with gastric variceal bleeding than esophageal variceal bleeding. So workforce considerations. Why is this important? Well, practically speaking, even if EUS is shown to be superior to standard endoscopy, except for a few select centers of therapeutic EUS excellence, such as the center where Dr. Ken Binmiller is from, there just aren't enough expert endoscopists in the US or frankly in the world to perform EUS guided therapy for bleeding gastric varices. The key takeaways. Glue injection by standard endoscopy has many advantages over EUS techniques for most cases of gastric variceal bleeding. The major disadvantage of EUS is the limited number of expert therapeutic endoscopists available to treat acute and potentially life-threatening hemorrhage in this situation. Glue injection is not FDA approved for use in the US, so it needs to be used off-label in the US and with specific patient consent documentation. Many different methods of therapy are available to treat bleeding gastric varices that vary in availability, complexity, cost, and possibly in clinical patient outcome. Much more research is needed to better define the roles of non-EUS standard endoscopy glue injection versus EUS guided glue injection versus IR techniques such as TIPS and BRTO in treating patients with gastric variceal bleeding. Thank you for your attention. Hello, I'm Julie Heimbach, and today I'm gonna speak to you about a surgeon's do's and don'ts for the endoscopy. I am a professor of surgery and transplant at Mayo Clinic in Rochester, Minnesota. Nothing to disclose. And today I was asked to speak to you about the role of the EUS in tumor staging, particularly for cholangiocarcinoma, as well as the question of drainage in unresectable versus resectable tumors. And finally, speaking about the role of endoscopy in the management of post-transplant complications. First, speaking about tumor staging. So specifically, we'll start with the role of EUS in staging for periheliochangiocarcinoma, which is a tumor in which establishing the diagnosis can be quite challenging. The treatment is surgical resection in most cases. Unfortunately, many patients present with unresectable tumors, and overall, the prognosis for this difficult tumor remains quite poor. So first, in establishing the diagnosis of periheliochangiocarcinoma, there's an important role of imaging, and the utilization of CT can give us a good look at the vascular involvement, whereas MRCP is really the primary mode of identification of ductal involvement. MR is superior to CT overall, and ideally, this would be performed prior to the placement of stents, although this obviously is not always possible. EUS really has a primary role in detection of the nodal disease, and we would want to avoid the biopsy of the primary tumor through the peritoneal route for patients being considered for curative therapy, whereas ERCP is really the workhorse of establishing the diagnosis, because first, we're able to demonstrate a malignant-appearing structure, and then perform an endoscopic biopsy and brushing, and although this is not always diagnostic, it can be very important in establishing the diagnosis. The elevation of CN-99 can be helpful, but unfortunately, it can be absent in patients who are lewis antigen negative, and of course, if there's cholangitis, this can really make this confounding. We also want to differentiate this from IgG4 cholangiopathy. The primary treatment is resection, but unfortunately, there are some major technical challenges to resection, making resection possible only in about 70 to 80% of the cases where it is attempted, and many patients present with unresectable tumors, which are identified even before attempted resection. Why is this? Well, unfortunately, this is a very high-stakes real estate area, as you can see identified in the image, where the tumor is sort of centrally located right in the area where the ducts are entering into the liver, and we also have the key vessels of the portal vein and hepatic artery going right through the same exact area. Outcomes for resection really depend on the ability to obtain a complete resection, as well as the nodal status, which is where EOS becomes so important, and pathological factors, such as perineural invasion. So this surgical series I chose to highlight only because it has careful long-term follow-up, and it just demonstrates that even out to five years, we can see patients still experiencing recurrence after a complete resection, with a median survival at 3.9 years and a 43% survival at five years, which is really excellent and higher than is achieved by most series, but still, even in this very careful series, we can see that continued evidence of recurrence. Another treatment strategy which has evolved for patients who are not eligible for resection is the use of neoadjuvant chemoradiotherapy in liver transplant, and this series highlights the experience of 12 centers with 214 patients, and you can really see in this patient population with no other curative option because their tumor is unresectable, this series is able to establish a 65% five-year survival. Led to the OPTN making this a standard indication for liver transplant, with the diagnosis being established by the appearance of a malignant-appearing stricture on ERCP, as well as at least one of the following, which would be malignant cytology or histology, and this would be obtained through the endoscopic route, elevated CA-99 in the absence of cholangitis, if the patient has demonstration of polysomia on fish, and if there is a mass, this should be less than three centimeters on cross-sectional imaging. The patient would then be treated with neoadjuvant chemoradiotherapy. The tumor should also be un-resectable either because of anatomic considerations or because it is arising in the setting of PSC. Patients are excluded from the possibility of transplantation if they have had a prior violation of the tumor plane either in the setting of attempted resection or if they have had a transperitoneal biopsy which could be either through the skin or through the stomach or duodenum as in the setting of EUS. If there is a mass, it needs to be less than 3 cm in the radial margin. The extension up and down the duct like the length of the stricture is not a contraindication if it is greater than 3 cm. The patient would also be excluded if they have had evidence of metastasis which can be demonstrated on EUS or other factors that would exclude eligibility for the treatment. This is a small series but it highlights the importance of not performing the transperitoneal biopsy. This is a series of patients treated at our center where those that had no biopsy performed at the primary tumor through the peritoneum either percutaneous or endoscopically had an 8% incidence of peritoneal metastasis found at staging whereas those that had a transperitoneal biopsy where tissue was obtained that was demonstrating malignancy had an 83% incidence of peritoneal metastasis demonstrating seeding essentially right along that needle track that we could see at staging. Those that had an attempted transperitoneal biopsy that was negative so they got no positive tumor cells had no evidence of peritoneal metastasis. This series is really one of the seminal papers on the use of EUS in staging for cholangiocarcinoma not just Tyler actually but they also looked at other subtypes intraepatic and distal and they had 157 sequential patients treated at a single center between 2014 and 2018 and the use of EUS was able to identify regional lymph nodes in a higher percentage of patients compared to those that were seen on cross-sectional imaging as you can see in this image and those are the lymph node beds that are typically sampled. EUS was able to identify 87% of those with metastatic lymph nodes and thus concluding that EUS with FNA biopsy is effective in identifying metastatic regional lymph nodes in patients with cholangiocarcinoma of all subtypes not just Tyler but intraepatic and distal. They also highlighted in this study that almost all of the nodes were actually sampled as they had previously demonstrated that simply the appearance of nodes was not satisfactory to rule out the evidence of metastases. The metastases were found in 17% of patients and then in the Kappelmayr curve that shows that that is associated with worse survival so patients that had malignant nodes are in the solid line demonstrating poor survival. Those that were able to have no nodes demonstrated on EUS and went on to curative resection or transplant have the most optimal survival as you can see demonstrated there in the Kappelmayr curve and therefore identification of these metastases avoids application of neoadjuvant chemoradiotherapy which is a you know essentially non-therapeutic and has morbidity as well as the morbidity of a staging laparoscopy or laparotomy that can be avoided if the patient has the positive lymph nodes identified with EUS. So to summarize the endoscopic role in the patients with cholangiocarcinoma particularly those receiving neoadjuvant chemoradiotherapy are that the EUS staging should be performed prior to neoadjuvant therapy or surgical staging because it can avoid the morbidity of these treatments in cases where it would not be helpful. We would want to stress the importance of avoiding biopsy of the primary tumor through the transperitoneal route not through the endoscopic route through the bile duct but across the peritoneum if a curative therapy is being considered. Recurrent cholangitis is common in this patient population as demonstrated in that explanted liver where you can see extensive inflammation in the biliary tract showing why recurrent cholangitis would be common and why the patients would need frequent ERCPs while they are awaiting therapy. We would want to avoid the non-removable self-expanding metal stents in this population if you are considering curative therapy as they are very difficult to extract out of the duodenum when you are removing the liver and the bile duct. Covered metal stents can also be a problem in this population because they can obstruct the cystic duct leading to cholecystitis and potentially gallbladder perforation. Obviously this can happen even in the absence of these covered stents and in this group we have found that the endoscopic placement of a cystic duct tube can also be helpful. Moving on to biliary drainage for resectable and unresectable patients. First we'll start with the easy group which is the patient with unresectable chondrocarcinoma and really has been established, this is one of the series that helped to establish that really the use of the uncovered metal stents is superior to the plastic stents because it provides a reduced frequency of the stent changing which allows the patient to have fewer interventions in their terminal phase of life which is a benefit. However if we are in a very small subgroup able to consider the candidate for the possibility of chemoradiotherapy and liver transplant we would not want to use these uncovered metal stents and instead use plastic stents. So what about resectable patients and the role of endoscopic biliary drainage? Well this is actually controversial. This shows a large series, 11 centers with more than 350 patients who are undergoing resection and half of them had biliary drainage. And in this retrospective series, large series, they showed that patients who had biliary drainage had a decreased mortality rate in the setting of a right hepatectomy as shown in the bar graph, however there was no difference overall in mortality with or without biliary drainage. So in the subgroup where there is a right hepatectomy it seemed to be beneficial. A randomized, multi-center randomized controlled trial also from Europe with 54 patients enrolled. The chosen criteria would be patients with persistent jaundice, so bilirubin greater than 5 or a dilated duct in the future liver remnant. Unfortunately this series demonstrated a 41% mortality in those having a PTC versus 3 out of 27 in those undergoing ERCP. Overall the complication rate in both groups were similar but the mortality was significantly higher in the PTC group leading to the stoppage of this randomized controlled trial prematurely due to the higher all cause mortality in the PTC group. So this was a small sample size and really does require further study and overall the authors concluded that they would want to have a potential reassessment of the role of preoperative drainage in general for patients with resectable chondrocarcinoma and whether this should be selected based on right lobe versus left lobe being resected and of course now with neoadjuvant therapy being considered in some patients who are going for resection drainage would likely be more often to be needed so really there's a lot of uncertainty around this subject at the present time. This is a seminal paper really in the role of surgical treatment for perihyalcarcinoma. This is from Nagino in Japan who's really one of the very most well-known surgeons who are treating this problem and showing his evolution over time of 574 patients and this shows time with the top Kaplan-Meier curve being the more recent series 2001 to 2010 showing improved results in both time eras lymph node metastases importantly was still the strongest predictor of outcome and this particular series actually has a switch from percutaneous biliary drainage for all patients switching in 2002 actually to nasobiliary drainage for all patients so further confounding clarity on the issue of biliary drainage. So moving on to the endoscopic management of post liver transplant biliary complications and this is a nice review in endoscopy clinics of North America from 2019 really talking about biliary structures which of course could be managed endoscopically in most cases and they can be anastomotic the risk factors would be if it's a living donor liver transplant certainly much more risk for developing a biliary structure and non-anastomotic whether they have a DCD transplant if it's an ABO incompatible or if there's an arterial complication that's where we would see the non-anastomotic structures importantly and good reminder that ultrasound may not show dilated ducts so MRCP can be useful in providing a road map in this population. ERCP is the standard treatment you may need to consider rendezvous for not able to transverse the structure and resolution it can be seen in a high percentage of these patients although it does take time and repeat procedures. For leak you can see this here and this is very often to be early in fact in the first seven days in most cases typically these are technical issues and they're higher in living donor liver transplant or in a DCD and that has to do with the ischemia that can be established with this type of donation and in some cases we would need to use operative management but some of these are very easily managed with endoscopy where the stent is placed across papilla and the leak and is left for about six to eight weeks and then removed. This is a study that well highlights this problem of ischemic cholangiopathy which you can see in the image here and then in this cast that is removed following the endoscopy and really this series I chose to highlight because it gives the frequency of biliary intervention which is needed whether it is ERCP in this series patients had a range of one to ten biliary procedures per year leading to the interest in that a patient to have a PTC or even be considered for retransplant. This is a very difficult population to treat and often requires many interventions. So specifically coming back to this in living donor liver transplant it's important to highlight that this occurs later than in deceased donor transplant and we typically will see this leak identified with abdominal drains and it can be either from the cut surface or from the anastomosis. Stenting across the leak is able to control it with deep compression of the biliary system and oftentimes that drain will manage the leak over time although then the patient will be at risk for the development of subsequent stenosis. In stricture we can see this in upwards of 35% of the patients of having living donor liver transplant. This again is typically managed with stents and it's very important when you're stenting these patients you want to be prepared for the presence of multiple anastomosis, multiple ducts and the presence of the colodocal jejunostomy. The use of MRCP can provide a roadmap and importantly I've learned this from Dr. Tephazian at Mayo Clinic that undrained ducts can contribute to the non-resolution of the stricture so it's important to be sure that you identify all those segments. This paper from a large series from Chalk et al demonstrate endoscopic treatment for biliary strictures in living donor liver transplant where they look at it by the morphology of the stricture which you can see here they describe it as pouched, intermediately pouched or triangular referring essentially to the shape of the stricture on the distal end and really they found that overall these can be endoscopically managed successfully 73% of the time but the risk for failure related to that shape with the pouched shape being more at risk for failing and recommending that if you do have a pouched type after and a failure after that first intervention that you would want to move to PTC or even surgical revision earlier in that subgroup. So this is just highlighting the role of the double balloon so when a patient does have a roux anastomosis you're still able to approach those endoscopically with a relatively good success rate as you can see highlighted in this series of 20 patients who had a colodocal j adenostomy they were able to get to the anastomosis in 85% of those patients so recommending that not only using a balloon but a balloon and stent because this had a higher success rate. Also we can see papillary stenosis in this population it's quite uncommon sphincterotomy is quite successful and biliary stones are also a possibility. ERCP again can be used and generally as successful although importantly they can recur. The key takeaway that EOS provides highly effective management of biliary, highly effective staging in biliary malignancy but we would want to avoid biopsy of the primary tumor if the patient is being considered for curative therapy. For unresectable hyaluronic gland geocarcinoma we want to consider whether liver transplant or palliation is being considered. I want to avoid permanent metal stents in a potential transplant candidate. For resectable chondrocarcinoma drainage is controversial and evolving and most importantly endoscopic management of post-transplant biliary complications is really key to a successful liver transplant program. Thank you very much for the opportunity to speak to you today. Hello my name is Ian Gann. I'll be speaking on salvage gallbladder and biliary drainage taking the standpoint of EOS as the primary treatment modality. I'd like to thank the AASLD for inviting me to speak on this interesting topic as well as the course directors. This is a short biography that you can refer to. These are my disclosures. EOS guided therapeutic endoscopy has made significant advances in the last two decades. In particular we now have the capacity to provide biliary decompression both the gallbladder and the bile ducts themselves either as salvage procedures such as EOS rendezvous with the bile ducts or by direct transmural stenting of the biliary tree via hepatic gastrostomy or colodoco duodenostomy. We will begin by discussing EOS guided gallbladder drainage. Acute cholecystitis is a common condition representing the most frequent reason for admission to surgical wards. While laparoscopic cholecystectomy is a very safe procedure, high rates of acute cholecystitis in the elderly are marked by high rates of morbidity and mortality. In addition other comorbidities such as cirrhosis can substantially increase surgical risks. It was for this reason that in 2013 the Tokyo guidelines for acute cholecystitis suggested that in moderate and severe cholecystitis there should be heavy consideration of percutaneous decompression of the gallbladder and delay of surgery until it could be performed electively. Indications for percutaneous cholecystosomy tubes included severe acute cholecystitis, elderly patients or those patients with significant comorbidities or those with contraindications to GA. The issue however is that percutaneous cholecystosomy tubes are not without their drawbacks and carry with them a high rate of morbidity of 5 to 40 percent including bile duct injuries and leaks, hemorrhage, pneumothorax and displacement. They are invariably a cause of some discomfort for patients and furthermore they are contraindicated in patients with ascites. The Tokyo guidelines were updated in 2018 to reflect new evidence for endoscopic means of decompression but more needs to be done to promote these as mainstream interventions. Two forms of endoscopic gallbladder drainage are available. The first is endoscopic transpapillary stenting. This was first described some 25 to 30 years ago by Dick Kozarik. They're often not performed due to perceived complexity but are relatively actually simple to do. In these procedures the cystic duct is selectively cannulated during the time of ERCP followed by guide wire insertion. Then over the guide wire a plastic double pigtail stent is inserted leaving the proximal end in the duodenum. Second, a newer technique has been revolutionized by lumen opposing dumbbell shaped metal stents that can easily be placed directly under EOS guidance from the stomach or the duodenum into the gallbladder. The advantages of endoscopic therapy over percutaneous therapy have been well established. Here we see a meta-analysis of endoscopic therapy versus percutaneous drainage with similar technical success rates, although some favoring of endoscopic therapy. The adverse event rates, however, highly favor endoscopic drainage, as you can see here. Advantages were also seen with much lower rates of recurrent cholecystitis in this study. Again, 22% recurrent cholecystitis in the percutaneous drain group versus 3 or 4% in the endoscopic group. Also a significant reduction in length of stay, 16 days in the percutaneous group versus 7 days in the endoscopic group. Here another study, this time a retrospective study showing lower pain scores, hospital stays, and re-intervention rates, and adverse event rates with EUS-guided gallbladder drainage over percutaneous. The re-intervention rate was almost a tenfold difference. The adverse event rate was 18% in the EUS group versus 31% in the percutaneous drain group. Here's an example of EUS-guided gallbladder stent placement. The gallbladder, the dark structure, is easily visualized from the duodenum. The stent is introduced with an electrocautery tip. The distal phalange is then deployed. Well, as I stated before, endoscopic transpapillary stenting does enjoy a relatively high technical success rate of 81% to 84%, with a relatively low adverse event rate at 3%. There is, however, a significant variation in cystic duct anatomy, and this, in addition to the fact that ERCP in itself has a 5% to 10% failure rate, can lead to several technical failures. Technical failures associated with transpapillary gallbladder stents include just unsuccessful barrier cannulation, the inability to find or cannulate the cystic duct because of a highly-angulated take-off or aberrant anatomy. You can have impacted stones that prevent either cannulation or the ability to pass a guide wire. In addition, you may have a pre-existing duodenal metal stent or a biliary stent. And finally, there can be inability to advance the plastic stent into the gallbladder. Technical failures of EUS guided therapy, however, are just usually due to stent misdeployment or perhaps malpositioning of the gallbladder relative to the EUS scope. Complications are generally the same, including perforation, bleeding, recurrent cholecystitis, and stent migration. In addition to the sedation risks, EUS guided gallbladder drainage has increased risks of stent maldeployment, while complications of transpapillary gallbladder drainage include those of pancreatitis. There is unfortunately a paucity of literature comparing endoscopic transpapillary gallbladder stenting versus EUS guided gallbladder stenting. This is a three-way study looking at percutaneous versus transpapillary versus EUS guided. With over 372 patients overall enrolled in the study, we can focus in on the transpapillary group and the EUS group. As you can see here, there is significantly improved technical success with the EUS lumen opposing stent group versus the transpapillary group. There is a trend to more adverse events in the EUS group, but this is not statistically significant. In this retrospective study of 71 patients, technical success rates of EUS guided drainage were higher than with transpapillary drainage at 97% versus 84%. There were no significant differences in length of stay, pain scores, or adverse events. There should at least be some mention of costs, as the cost of lumen opposing metal stents is literally thousands of dollars more than a plastic stent and may alter the cost effectiveness of this strategy. More studies are required. In conclusion, endoscopic drainage of the gallbladder is safer than percutaneous cholecystostomy. Early evidence shows trends towards improved success with EUS guided gallbladder drainage over transpapillary stenting. We now turn to EUS guided biliary drainage. Biliary obstruction has traditionally been treated with ERCP with failure rates from 5% to 7% due to malignant luminal obstruction, indwelling duodenal stents, ampullary pathology, or variant anatomy. Failures have usually been treated with percutaneous transapatic biliary drains. These are limited in patients with ascites and also have a very high morbidity rate of up to 33% due to catheter dislodgement, recurrent infections, cholangitis, pneumothorax, and patient discomfort. EUS guided therapy has recently played an important role in salvage therapy for biliary obstruction. Several techniques have been described. Biliary rendezvous involves advancing a wire into the bile duct, either in the clomid portion as seen here from the duodenum or from the left hepatic duct through the stomach, guiding the guide wire through the ampulla, and allowing for wire access with a regular ERCP scope. Anterograde stent placement can be performed with needle access through the stomach into the left lobe of the liver, wire placement beyond the papilla, followed by a transpapillary stent. Similarly, EUS guided hepaticogastrosomy uses the same approach with the left hepatic duct being accessed directly from the lesser curve of the stomach using a needle. A guide wire is then placed, and the gastric wall is then dilated, and a stent is placed directly from the stomach into the left hepatic duct. Finally, choledochoduodenostomy uses luminoposing metal stents placed from the duodenum directly into the proximal common bile duct. Potential advantages to EUS guided biliary drainage include stenting away from the site of malignancy or obstruction, and reduced risk of recurrence from tumor ingrowth. Both hepaticogastrosomy and choledochoduodenostomy do not require intubation of the second portion of the duodenum, which can allow biliary decompression in patients with duodenal and gastric outlet obstructions. Finally, there is a reduced risk of post-procedural pancreatitis with no need for ampullary manipulation. This meta-analysis reviewed nine studies with 483 patients comparing percutaneous biliary drainage with EUS guided therapy. There were no differences in technical or clinical successes, but there were significantly less adverse events, re-interventions, and reduced hospital stays. So while there does appear to be a significant advantage of EUS guided bile duct drainage over percutaneous drainage when ERCP has failed, this starts to beg the question as to if EUS guided bile duct drainage can be used as a primary means of decompression over and above ERCP. While studies are starting to show this to be the case, this is a Korean study of 125 patients, and this showed a higher technical success rate in patients with duodenal involvement, as you might expect, as the EUS guided therapy does not require duodenal intubation versus ERCP. So 94% versus 90% with ERCP. The clinical success rate was high in both methods, and the overall survival was approximately the same. The difference, however, was the six-month stent patency rate. You can see here there's a large difference in stent patency rate at six months, with 85% patency rate in the EUS group versus just 49% in the ERCP group. This has borne itself out in meta-analysis as well. This is a recently performed meta-analysis looking at EUS guided biliary drainage versus ERCP for primary drainage and the management of malignant biliary obstruction. Again, EUS guided biliary drainage had significantly lower rates of re-intervention secondary to tumor in-growth. So it is possible that EUS guided biliary drainage will become the modality of choice for primary drainage in a palliative setting for malignant biliary obstruction. In summary, EUS guided gallbladder drainage is safe and provides excellent amelioration of patients who are non-surgical candidates. In addition, EUS guided biliary drainage is proving to be an ideal form of palliation of biliary obstruction, both as a salvage procedure in failed ERCPs and potentially a primary means of biliary decompression. Takeaway points, endoscopic gallbladder decompression is safer than percutaneous drainage. EUS guided gallbladder drainage is more successful than trans-papillary. EUS guided biliary drainage is safer than PTBD. And finally, primary EUS bile duct drainage may be more durable than ERCP, but more studies are needed in the future. Thank you very much for your attention. I again thank the AASL. Hello everyone. Thank you for joining. My name is Praveen Chauhal. I'm an advanced endoscopist at Cleveland Clinic. Today we are going to talk about post-transplant ERCP. Should we use colonoscope, balloon endoscope, or endoscopic ultrasound? I don't have any disclosures, and this is the outline for my talk. We'll start with introduction. We'll discuss the anatomy. We'll talk about the tools needed for completing the procedure successfully in surgically altered anatomy in post-transplant patients. We'll talk about endoscopic ultrasound, single balloon endoscopy, double balloon endoscopy, what the future holds, and the take-home slide. So for traditional anatomy, ERCP is successful in close to 98% to 99% of the patients. However, in patients who undergo ruined viability construction, the success rate of ERCP for biliary indications go down. Ruined viability construction is usually performed in patients with underlying TSE disease, if they have small bile duct, or if there is discrepancy between donor and recipient duct, patients with biliary atresia, and sometimes ruined viability construction is done in patients who need retransplantation. Some of the complications that can occur after ruined viability reconstruction are leak, fractured tubes, strictures, or stone formation. These are also the common indications for getting a call for ERCP. So this is a diagram depicting what a ruined viability colonoscopy looks like. It's an end-to-side construction of the donor bile duct with the recipient jejunum. So when you insert an endoscope, there's an intact stomach, intact duodenum, and a little bit of a jejunum, and then you encounter the jejunal-jejunal anastomosis. And the rule is typically about 25 to 30 centimeters long before you reach the colito-cojejunal anastomosis, which is about 10 centimeters or so proximal to the blind end. So when you are embarking on performing ERCP in these patients, it's important that you check your toolbox to make sure you have all the tools necessary to complete the procedure successfully. The most common endoscope that we use for these patients is pediatric colonoscope. And the compatible accessories for colonoscope are seven-fringe stent. You may use straight or double-peaked tail, and you'll need long wires for this. You can use standard ERCP accessories. They are long enough to go through the colonoscope. However, once the accessory, like an occlusion balloon or sphincter tube, exits the colonoscope biopsy channel, there's a little bit of a length left, which is usually not an issue with colito-cojejunal anastomosis because the common bile duct is often short before you hit the bifurcation in these patients. The advantage of adult colonoscope is it's a stiffer, especially in somebody with longer length roux limb, it may facilitate advancement to the anastomosis, perhaps a little bit better than the pediatric colonoscope. Another advantage is with this therapeutic channel, you can deploy fully covered self-expandable metal stents through the adult colonoscope. If the patient needs single balloon endoscopy, you will need longer length wires, typically 400 centimeter long. You have to have the accessories designed for balloon endoscopy, including special occlusion balloons and sphincter tubes. If you are using double balloon endoscopy, again, you will need extra long wires, which are usually 600 centimeter long, and accessories which are designed specifically for double balloon endoscopy. For endoscopic ultrasound guided access, you are using a therapeutic channel curvilinear echo endoscope, and you should have a slew of fully covered metal stents of variety of sizes. Most common size is eight centimeter long, either eight millimeter or 10 millimeter in diameter. You may have to use lumen opposing metal stent if you are getting connection to the afferent limb, a variety of FNN needles, including excess needle, both 19 gauge and 22 gauge, and the corresponding wires that go through those needles. Usually we start patient in a left lateral position or left oblique, but you may have to move patient to either supine or prone, just like we do in colonoscopy to facilitate passage of endoscope to the end of the afferent limb or to the anastomosis. Most of these procedures are performed under general anesthesia, and we use CO2 for insufflation. So some of the tricks to facilitate biliary access in patients with Roux-en-Y anatomy, it just so happens I had a couple of these patients on my schedule yesterday, and fortunately they were pretty simple, straightforward, and they were successful. But over the years, you encounter patients with different challenges, and some of the things you can do is if you're starting with pediatric colonoscope, you may want to switch to adult colonoscope, depending upon the challenges you encounter, or a single balloon endoscope. You can change the patient position, as I mentioned earlier, from left lateral to prone. You can use abdominal pressure, and also be mindful of the utility of air cholangiogram. I think oftentimes when we are closer to the anastomosis, we are in the afferent limb. We are not close to the anastomosis yet, but you can see air cholangiogram, which is reassuring that you are heading in the right direction, and that the anastomosis is within reach. You can change the accessories, moving from straight to angled or tapered, depending upon the angles you are dealing with. You can use secretin provocation to delineate bile flow. Sometimes you can try putting the patient in a Trelandenburg position, and you can flood the afferent limb with contrast to see if there is filling of the biliary tree via open anastomosis. The anastomosis has to be open for this. So this is a case that I did yesterday. Again, the endoscope position doesn't show there are a couple of loops before we were able to reach the anastomosis. You see a beautiful air cholangiogram, a decent-sized common bile duct with left and right, and then the corresponding cholangiogram was just a mirror image of the air cholangiogram. This patient did not have any anastomotic stricture. As I mentioned earlier and I showed in the picture, when you are embarking on endoscopy in these patients and you reach the jejunal-jejunal anastomosis, then some of the things that you can use to find the Roux limb or the afferent limb, oftentimes it's the more tricky one to enter into. It's more angulated. Usually with the efferent limb, the peristalsis is going downstream, and with the afferent, that's not the case. And once you enter into the afferent, as I mentioned earlier, presence of air cholangiogram and also the use of fluoroscopy, which shows that the endoscope is going in an upward direction towards the right upper quadrant is reassuring that you're headed in the right direction. And once you have reached anastomosis, don't forget to tattoo the afferent limb takeoff in case this patient needs future procedure. So it's easy for you to identify the limb and you're not going into efferent and afferent back and forth. So I mentioned earlier that we use a colonoscope. So how good is the colonoscope in successfully completing the procedure? This is the study we did when I was a fellow at Mayo, first study of its kind, looking at 132 orthotopic liver transplant with Roux-en-Y reconstruction of these 31 patients under ERCP. And ERCP was successful in 71% of the patients using colonoscope. Pediatric colonoscope was used in the majority. And the causes for failure was either the long limb or tight angulation or scope looping. I think one point I would like to make before moving on to other technologies is ability to internalize BTC. Not uncommonly, we get patients who refer to us who have had a BTC after the Roux-en-Y orthotopic liver transplantation with an assumption that the anastomosis cannot be reached. And unfortunately, these patients, they carry this BTC around for months and sometimes a year for treatment of anastomotic stricture. It's very easy to internalize the BTC provided you are able to reach the anastomosis. So make sure first that you are able to reach the anastomosis. Once you are at the anastomosis, then all you need to do is to cut the PTC from the hub and it just unravels or unlocks the lower loop and you can internalize by putting a stent across anastomosis. Be mindful that the PTC at least six weeks old, it should not be internalized or manipulated if it's a fresh PTC less than six weeks old. Again, some of the other things that can be accomplished provided you are able to reach the anastomosis is make sure sometimes patients they have both the right and left anastomosis separately where you have to look for two separate anastomosis. Most of the times they are pretty close to each other, but sometimes they may be a few centimeter apart. And that's where reviewing the surgical note and reviewing previous imaging is helpful. And this is a patient who had inadvertent closure of the right hepatic duct. And as you see the cholangiogram obtained from the existing PTC, there was a complete cutoff of the right hepatic duct and the patient was condemned to external percutaneous drainage. So when we were fortunately able to reach the anastomosis, again, another cholangiogram demonstrating an abrupt cutoff of the right system. And you can see the PTC in place and those go pretty close to what looks like an anastomosis. We put a wire percutaneously and just with the extrinsic impression, you can delineate where the right duct is located and you can pass the wire percutaneously and grab it internally with the grasping device and finish the procedure by placing a fully covered self-expandable metal stent through this newly created choledocojejunal anastomosis for the right system. Again, patient did well and we were able to remove the PTC in the same session. So very important to know the anatomy of the patient before you embark on these procedures. When standard tools fail, you can consider balloon endoscopy or U.S. guided biliary access. Both these procedures, they need experience. Balloon endoscopy is about moderate difficulty and U.S. guided biliary access requires significant technical expertise and can be moderate to significantly technically difficult. Both could be time consuming. And for balloon endoscopy, you have to have your tool kit stocked with the compatible accessories, which are oftentimes limited. The reported study showed full success rate of about 80% with adverse event ranging up to 0 to 6%, mainly being abdominal pain perforation. One study that I would like to mention more about is the study that also came from Mayo. This was a follow-up study from the previous study that I mentioned earlier. In this study, they compared single balloon endoscopy to colonoscope and they noticed that patient who had failed ERCP with the colonoscopy, the pediatric or adult, they were able to successfully finish the procedure with SVE. The technical success rate was similar for initial procedures. However, for follow-up, single balloon endoscopy outperformed in terms of technical success. There were no adverse events in these groups. So they recommended proceeding with single balloon endoscopy as the procedure of choice in patients with Rheumatoid anatomy post-transplant. So moving on to the U.S. guided biliary access, the success rate is variable and complication rate are reported up to 20% based on the type of U.S. guided access being used, either rendezvous, anti-grid, hepatic gastrostomy, or U.S. guided access of afferent limb via lumen opposing metal stent. Some of the technical considerations for U.S. guided access, especially if you're performing rendezvous, you have to be able to reach the biliary anastomosis in these patients. So make sure you drive down and you reach the end of the afferent limb. You have to use long wires. Usually we start with a 19 gauge, but if the intrahepatics are not significantly dilated, by that I mean minimum of about five millimeter or so, you can use 22 gauge. And with 22 gauge, oftentimes you have to use either 018 or 021 wire. Be careful of the wire sharing. And as everybody know, we are using the segment two, rarely three, to access the bile duct and dilated intrahepatics. For anti-grid stenting, one thing you have to be very careful about is the biling at the access site. And you keep your dilation of the access site, the bare minimum that's necessary to pass the stent. Oftentimes it's about four to six millimeter. And you are using fluoroscopy and maintaining a stable scope position with the scope really butted against the lesser curvature, sometimes closer to the cardia. For hepatic gastrostomy or on the side of using a longer length stent, usually eight centimeter, they are fully covered or partially covered stent. And be mindful of the liver and stomach movement while the patient is breathing, because that can lead to stent misdeployment. Oftentimes we leave double pig tail stent through the fully covered metal stent to prevent migration. Again, an x-ray imaging showing rather a straight and stable scope position, decently dilated intrahepatics with a needle in the dilated intrahepatics of segment two, and a good cholangiogram wire traversing down and out. And this is the case where you can either do a rendezvous or you can finish the procedure in an anti-grit fashion. As I mentioned earlier, if you're performing hepatic gastrostomy, you are choosing a really longer length stent to make sure at least three centimeter or so is hanging into the stomach and finish the procedure off by placing a pig tail stent to prevent migration. So I think the future direction, we need innovation in both endoscopes and the accessories for the endoscope, you know, modified therapeutic channel scopes, forward viewing echoendoscopes, multichannel, and perhaps robotic platform. And accessories that are sorely needed are the closure devices, anchoring devices, access devices, and stents, which are intended and made for biliary and liver work and tools to help with the complications like perforation, bleeding. So ladies and gentlemen, this is my final take-home slide. If you have a patient with Roux-en-Y post-transplant, start with reviewing the surgical report so that you know what's the length of the Roux-en-Y and obtain relevant imaging in like MRCP in somebody with stones or obstruction. Hide out if leak is suspected. We start with colonoscope. Choose a cap if you are dealing with a stenosed choledo-jejunal anastomosis. Otherwise, I don't find any utility of using cap in these patients if the anastomosis is wide open. If the colonoscope fails, you can move to a same session, single balloon endoscope. However, if that fails, pending your local expertise, you can choose to go with an EUS-guided access or the good old PTC. And with that, thank you so much for your attention. Hello. I wanted to introduce the next session related to the evaluation of biliary strictures. Advanced endoscopic and cytologic techniques for bile duct strictures. Does this provide increased clarity or more uncertainty? And I'll be speaking on these aspects of increasing clarity. These are my disclosures. There are several techniques to diagnose cholangiocarcinoma, and we'll be focusing primarily on biliary strictures and cholangiocarcinoma, given the setting of the liver meeting. But it's probably applicable to other ideologies of biliary strictures as well. But techniques include things like ERCP with brush cytology, intraductal biopsy, cholangioscopy. EUSFNA is performed. If you're dealing with distal bile duct lesions, then sampling of the primary lesion is usually acceptable. But in the setting of higher strictures, we tend not to do sampling transduodenally due to the potential risk for seeding. And this is especially pertinent in patients who are surgical candidates for resection or liver transplantation. Percutaneous biliary brush cytology and biopsy, I won't be speaking about. There's been some concerns about tumor tract seeding related to percutaneous interventions. This EUS image here to the right just shows a higher mass with surrounding vasculature. With the yield and sensitivities of ERCP sampling of strictures, it does depend on whether you're dealing with cholangiocarcinoma or pancreatic cancer. And with cytology, you can see a range that's even been lower than 40 percent to as really as high as 80 percent, but that's fairly old literature. Forceps biopsies for cholangiocarcinoma, anywhere from 33 to 50 percent sensitivity. Often we implement FISH, fluorescence in situ hybridization. And these are four fluorescently labeled probes that hybridize to the pericentric regions noted here. And really, I think the most helpful aspect of FISH is polysomy or tetrasomy. Polysomy is the most literature behind and most predictive potentially for malignant. These are five epithelial cells with three of these labeled signals, or excuse me, three signals in these chromosomal abnormalities and at least two of the probes that I consider the 5-3-2 rule. And when utilizing FISH in the patients with primary sclerosing cholangitis, you can see sensitivities that approach 50 percent, which is comparable to several series in cytology. And specificity is 88 percent if you combine a dominant stricture plus FISH polysomy. And the patient outcomes seem similar to those PSC patients with cholangiocarcinoma if they have polysomy. However, it should be noted that really it's the serial polysomy that may be more indicative or predictive of malignancy. And a single isolated polysomy that is subsequently followed up with either trisomy or something that's without any aneuploidy has very low rates of association with cholangiocarcinoma. So something to keep in mind when evaluating patients such as this. Endoscopic ultrasound is often used for the evaluation of ability strictures. The sensitivity is about 80 percent. And that, for higher strictures, tends to be the more difficult area to evaluate. Partly it's related to the limited sampling and sometimes it can be related to difficulty with visualization. The specificity, as with most tissue sampling techniques, is really quite high. But the hilum does remain to be a challenge. We then consider intraductal sampling with cholangioscopy. And at our center, we published the first consecutive series of cholangioscopic directed biopsy more than 10 years ago, looking at a fibro-optic mother-daughter system. And our operating characters are really quite robust. This was not specific to PSC. This was all comers for biliary and pancreatic pathology. But you can see a sensitivity approaching 90 percent and specificity of 96 percent, both of which, or at least the sensitivity, significantly higher than compared to conventional tissue sampling. Videocholangioscopy. These are some examples. And you can see some tortuous dilated vessels here on high definition white light. And then on narrowband imaging, this is a vegetative mass with the infiltrated stricture noted below with some neovascularization, or what I would consider tumor vessels as well. And this has really helped to further delineate the characteristics of lesions within the duct that can help and predict malignancy beyond just cytology alone. And we published our experience utilizing this technology. And we looked at the subgroups of characteristic lesions that had a higher propensity toward malignancy and sensitivities ranging from 48 to 60 percent with a single lesion noted here, infiltrated stricture, as well as tortuous dilated vessels. And what we decided is that, you know, what if you had more than one of these lesions? You had, say, an infiltrated stricture and a tumor vessels, for example, then what would be the sensitivity? Well, it's about 50 percent here. But specificity, this is without tissue sampling, just visualization alone, specificity becomes quite high. So perhaps it's not just one clangoscopic feature, but if you have two, potentially even more, that can help your confidence in determining whether malignancy is present or not, even if sampling were non-diagnostic. Digital clangoscopy has been introduced over the last five years. And this is the data taken from publication from a few years ago. Myself and several of my colleagues around the country have put together their early experience with digital clangoscopy. And you start seeing in these digital images, these tumor vessels and infiltrated stricture, villus mass, it's really getting close to a video image. And we're really in a different era with clangoscopy and digital imaging that allows us to be a little more confident in what we're able to visualize and help to predict whether malignancy is present or not. And in this group, in this study, we had a sensitivity of about 97 percent, specificity 93 percent, and directed biopsy, as you can imagine, yields the specificity close to 100 percent. So it's really quite encouraging early experience utilizing digital clangoscopy. This is not specific to PSC, but just all common biliary strictures, as well as pancreatic pathology as well. So this is a 57-year-old male presented with painless jaundice. It was referred to the University of Colorado Hospital for a high-level biliary stricture and bilateral inter-hepatic dilatation. As is often the case, the ERCP breast cytology was negative. A 10-front stent resolved the jaundice. And this is an MRI showing diffuse dilatation of the right and left inter-hepatic ducts. A hyler stricture is noted here, maybe abysmuth 2. We perform endoscopic ultrasound because of our concern related to malignancy, given the presentation. And you can actually identify a hyler mass that's seen here. And we did not do sampling transduodenally because of the concerns related to seeding. And this is the clangogram, fairly smooth, but some irregular stricture. And I have to admit, I mean, it's not terribly concerning for malignancy, except for the length. But there's a little bit of regularity, but not too impressive for malignancy, in my honest opinion there. So then we do clangioscopy. And actually, it doesn't look all that bad. You know, it's fairly smooth, concentric stenosis. On narrowband imaging, there's a suggestion of some tumor vessels. But compared to the earlier examples, not terribly impressive. But contact friability was noted in East Asian populations and literature from that part of the world. They often discuss and describe contact friability. And that was very evident on this case. And biopsies did, in fact, by clangioscopy, diagnose adenoparsinoma. And some excellent work by colleagues of ours from Ecuador, looking at, trying to see if we can classify intraductal lesions with clangioscopy. And then even training those that are clangioscopy users, but also non-clangioscopy users. And the inter-observer agreement for these operating care, for lesions, I should say, that predicted malignancy was 94%. That's really quite impressive. I look forward to seeing some of this data and then findings being replicated at other centers to see how we can incorporate this into our practice. A meta-analysis has been recently published, looking at digital single-operator clangioscopy. Six studies, pulled sensitivity and specificity are really quite robust. There are some limitations. I'm sure my colleague, James, will be speaking on this, but I just wanted to mention this series of 80 patients from Northern Europe. The visual impression was only 64%. Targeted biopsy sensitivity, only 15%. But it should be noted, breast cytology, about 50%. There's 40% of this patient population was PSC. So confocal microscopy is a technology. It's a three French laser probe that looks at not the epithelial, but the sub-epithelial surface. And to determine if there may be malignant or atypical cells present. Miami criteria have been established that have looked to suggest malignancy. And these are some of those examples noted on this slide. There was a prospective series, multi-center study with the central PI, Dr. Slipka from Pittsburgh, looking at the yield of confocal microscopy compared to tissue sampling alone, which is in orange. You can see sensitivity of 56% and specificity of 100%. But when you look at clangiography plus confocal, that sensitivity really goes up as well as the specificity, of course, goes down. But that's something that should be considered helpful with respect to confocal microscopy, adding a little bit of advantage compared to tissue sampling alone with respect to sensitivity. So this is a PSC patient that we saw a few years back. And you can see maybe a dominant stricture here, maybe one here, maybe it's multifocal dominant stenosis. Breast cytology was atypical, but the confocal microscopy showed dark clumps. And this is a representative video showing epithelial structures. And you'll kind of see some dark clumps actually right there. And the video played a little bit slower, but there were some changes of epithelial structures here early on. You can see that there. And then on explant, the patient actually had dysplasia in that area where we were concerned by confocal microscopy. So that early experience prompted a multi-center study, and this is not currently published, but it's been submitted for peer review. And we looked at about 60 patients at multiple centers throughout the country, including our own. And out of those 60 patients, seven had confirmed malignancy, 12 thus far have gone on to liver transplantation. And these are the operating characteristics. Quite a busy slide, but I'll just point out the confocal microscopy had a sensitivity of 86%, low specificity compared to tissue sampling, of course. But look at the breast cytology, really quite disappointing. Of course, it's a small sample size, a small number of patients, despite a couple of years of recruitment of patients at the several sites. Fish sensitivity, about 60%, and fluoroscopic biopsy, 50%. So does combining ERCP-based imaging modalities and tissue sampling methods improve yield? So we looked at our experience of over 600 patients that undergone breast cytology over a seven-year period of time, and we looked at the criteria and multiple modalities. If you looked at breast cytology alone, the sensitivity was 40% and specificity was 99%. But when you start combining modalities, let's say you add cytology, fluoroscopic biopsy, and clangoscopic visual impression, fairly common combination, almost 98% sensitivity. This is all comers, a subset of our patients had PSC as well but by and large, it's really the combination of modalities that appear to be encouraging to help improve yield related to ERCP sampling. So the key takeaways are breast cytology is simple, but unfortunately has a low yield. Combining cytology with other advanced techniques will increase the ERCP yield, and EUS for distal biliary sampling, and if there's hyalur pathology, then sampling of the nodes alone. The more difficult the patient population, such as PSC or hyalur strictures, the more likely advanced techniques such as FISH, clangioscopy, confocal microscopy will be required for diagnostic evaluation. And consider advanced techniques upfront or at a minimum during a second endoscopic session if initial cytology is negative and clinical suspicion for neoplasia remains. Thank you very much. Hello everyone, my name is James Tabibian and I would like to start by thanking our program chairs, Drs. Michael Levy and Jennifer Telford for the opportunity to present on the topic of advanced endoscopic and cytologic techniques for bile duct strictures. Here are a few words about myself as requested as well as a picture since the video component wouldn't quite work. Also, if you have not noticed yet, there is an icon on the lower left of each slide whereby you can scroll through and adjust the audio narration. With regard to disclosures, I am a consultant for Olympus Corporation of the Americas. Additionally, I disclose that the position I've been tasked with defending for this debate presentation may or may not reflect my actual position. Therefore, please tune into the live Q&A session on Monday, November 16, to learn more in this regard and for an interactive experience. The objectives of my presentation are to provide a background on the heterogeneity of bile duct strictures, BDSs, and the approach to their diagnostic evaluation, present a critical appraisal of advanced endoscopic and cytologic techniques utilized in BDS evaluation, and perspectives as to them leaving us with more uncertainty, and to identify future directions regarding such techniques for BDS diagnosis. Let's start with some brief background on the heterogeneity of BDSs and the approach to their diagnostic evaluation. Of the many variables to consider, BDSs may be in various parts of the biliary tree, with the perihylar and distal bile ducts being the focus of today's presentation, as they are usually intraductally accessible. BDSs may be short or long, mild or severe. They may be de novo, associated with an underlying cholangiopathy, such as PSC, primary sclerosing cholangitis, chronic liver flu infection, or IgG4 sclerosing cholangitis, or they may coincide with a specific acute injury or insult, such as a post-operative bile duct stricture. BDSs may be incidentally found, or they may be symptomatic. They may occur with or without accompanying serum tumor marker, CA19-9 elevation, and of course elevation does not always indicate malignancy, neither does malignancy always trigger a CA19-9 elevation. Also, BDSs may be in their native state or already manipulated, for example, with a stentin C2 at the time of referral, and last but not least, BDSs may be benign or malignant, or somewhere in between. To say the least, this heterogeneity can pose considerable clinical challenges. You may be wondering, why is there this image in the right upper corner of the slide? Well, this is to illustrate that bile duct strictures can be quite different from each other, just as an almond is different from a cashew, and that determining their ideology can drive the relevant stakeholders nuts. The goal of a BDS workup is to efficiently generate data that confidently establish a diagnosis and identify the most appropriate treatment. Advanced endoscopic and cytologic techniques are of course a mainstay in the diagnosis of BDSs, the former also representing a major therapeutic modality. Despite this, I've been tasked with defending the position that these techniques result in more uncertainty. Well, how can that be? Well, as the audience knows, tissue is the gold standard datum for diagnosis. However, the clinical reality is that a tissue diagnosis is not always achievable by advanced endoscopic and cytologic techniques, even after multiple resource consuming attempts. This said, one can in fact argue that anything short of a tissue diagnosis may, depending on context, be a costly gamble that adds to uncertainty rather than clarity to guide treatment. Is that a risk that the care team and patient are prepared to take? Let's now discuss and appraise advanced endoscopic and cytologic techniques for BDSs. For the purposes of this presentation, there are several advanced endoscopic and cytologic techniques to consider, and these can be dichotomized into tissue diagnosis techniques and essentially everything else. With regard to the former, we have intraductal brushings and intraductal biopsies. Intraductal biopsies can be subdivided into those that are fluoroscopically guided and those that are cholangioscopically guided. Other diagnostic data include cholangioscopic appearance with or without NBI, probe-based visual techniques or imaging techniques, and fluorescence in situ hybridization or FISH. Let's first talk about intraductal brushings. This is a technique wherein a brush is passed through the scope over a wire or not over a wire and into the stricture lumen. The pros are that this is a technically simple undertaking. Positioning of the brush is fluoroscopically confirmed. It's inexpensive and has high specificity. Of course, specificity also depends on what is considered a positive. If both suspicious as well as carcinoma are considered positive or if only carcinoma is considered a positive. The cons are that oftentimes cellularity is scant. And as with many of the techniques discussed herein, the pros and cons are also impacted by the presence of an underlying cholangiopathy such as PSC, which typically makes performance characteristics worse. The sensitivity of brushings is about 30 to 40% in expert hands, perhaps somewhat lower in the average endoscopist. And specificity approaches 100%, but again, with the aforementioned caveat. Fluoroscopically guided biopsies are a technique wherein forceps are passed through the scope either over the wire or traditionally not and advanced to the distal end of the stricture lumen. The pros are that positioning is fluoroscopically confirmed as with brushings. It's inexpensive and specificity is high. The cons are that larger capacity forceps may be difficult to get into or elevate into the bile duct. And the cellularity is also scant all too often. Sensitivity approaches about 45% in expert hands. Specificity approaches nearly 100%, but again, with the same caveats as cytology. Cholangioscopically guided biopsies are a technique wherein a cholangioscope is passed through the duodenoscope into the bile duct over a guide wire. The stricture is visualized, the guide wire is removed and mini forceps or spi-bite are passed through the cholangioscope. The pros are that there is direct visualization of what is being biopsied and specificity is quite high. The cons are that the forceps capacity is minimal, cost is high, both for the cholangioscope and for the forceps, and there are adverse events such as acute cholangitis. The risk of acute pancreatitis does not seem to be higher than conventionally RCP, but that's oftentimes because cholangioscopy is performed in patients who have already had a sphincterotomy previously. Also for cholangioscopy, general endotracheal anesthesia is probably advisable, if not necessary, given the amount of irrigation that's performed intraductally and the associated aspiration risk. Sensitivity of cholangioscopic biopsies is about 50 to 80% in expert hands and specificity is quite high at 80 to 100%, but again, with the same caveats as before, especially in patients with PSC or other underlying cholangiopathies. Though I'm supposed to be arguing the con, I should mention that there are now larger cholangioscopic forceps available, spi-bite max, which have the potential for acquiring two times as much tissue per bite. Some specific key improvements are listed in this slide, as well as some representative photos on the right. However, real-world performance data are still needed and it's unknown if desmoplasia will set a ceiling on the diagnostic yield, regardless of bite size. As for the non-tissue diagnosis techniques, the next few slides will provide an overview, but regardless of their diagnostic performance, unless there's sufficient credence in the results from the relevant stakeholders to actually change management, their utility is questionable at best. Context determines their utility or the level of credence. So you must know in advance if the result of a non-tissue diagnosis technique will be considered acceptable or not for actuating management decisions. And for this, multidisciplinary evaluation is paramount. Consider, for instance, the example of a common clinical scenario of a perihylar BDS, which appears to be a surgically non-curative or non-curable cholangiocarcinoma. If the oncology team will not administer chemotherapy until there is a tissue diagnosis, then these non-tissue diagnosis techniques will simply be wasteful. The first of the non-tissue diagnosis techniques is the cholangioscopic appearance. So this technically is cholangioscopy by look only. The pros are that it can characterize mucosal and some morphological features of a stricture better than non-invasive imaging or ERCP alone. The cons are the cost and that you also may not be able to traverse the stricture and perhaps most importantly, being burned by cancer mimics. Looking at these three sample images from some recent cases, you could see that they all are bile duct cancer, or are they? In fact, none of these three are from patients with bile duct cancers, despite the features that might suggest that there is a malignant neoplasm present. And we know this because the patients have been followed up for over a year and have had resolution of their biochemical and imaging and symptomatic abnormalities. So overall assessment, I think this can be useful for resection planning, but otherwise the reliability is questionable depending on the context. The next technique to discuss is the NBI appearance. And this is a technique that utilizes video cholangioscopy with the CHS B290 scope, for example. The pros are the very wide field of view and the very high image quality with video cholangioscopy. Also the improved visibility of blood vessels and mucosal abnormalities, which one would expect with NBI and the fact that this is a reusable scope, which paradoxically some may see as a con, ostensibly. The major cons are really all those that white light cholangioscopy has, plus this video cholangioscope is limited to just parts of Europe and Asia, and there's actually a need for two towers and two operators. So overall assessment, I think this is useful for surgical planning and it's also better at mapping and suspicious lesion detection than white light cholangioscopy, but availability is very, very limited. And again, it's not gonna be a tissue diagnosis. The next technique is fluorescence in situ hybridization. FISH is an advanced molecular cytogenics technique, which employs the use of fluorescent probes to evaluate for chromosomal duplications in cellular brushings. The number of cells containing nuclei with abnormal fluorescent probe signals is then quantified as shown on the right. The pros is that there's no additional sampling required. FISH can be performed on brushings used for conventional cytology. It's an objective technique, it's relatively inexpensive, and it does improve sensitivity and specificity over just brushings or even just biopsies. The cons are that most labs don't perform FISH, so you actually have to send it out and hope it gets there and that it gets back, the result gets back in a good time, and not all FISH polysomy indicates malignancy. We and others have in fact shown that nearly 50% of patients with FISH polysomy may either not go on to develop cholangiocarcinoma over a period of multiple years, or they may have reversion of their FISH polysomy to a less abnormal result once an underlying reversible issue is addressed. So assessment, in the right context, it is true that FISH may be considered compatible, FISH polysomy may be compatible with carcinogenesis, extant or impending, but otherwise I think there still are some gaps in what FISH can and can't illustrate. Though beyond the scope of this presentation due to time limitations, I do want to mention through the scope probe-based techniques which include confocal laser endomicroscopy, volumetric laser endomicroscopy, optical coherence tomography, and introductal ultrasound, among some other and emerging techniques. Some of these are still areas of continued investigation and refinement, whereas others have largely fallen out of favor, but fundamentally all suffer from the issue of subjectivity and not providing a tissue diagnosis. And as with the other non-tissue diagnosis techniques, context is really what determines their clinical utility or lack thereof. As we approach the end of this presentation, let's wrap up with a summary and discuss current needs regarding advanced endoscopic and cytologic techniques vis-a-vis the diagnosis of BDSs. So the key takeaways here are that bile duct strictures, BDSs, are very heterogeneous, thereby complicating diagnostic efforts and certainty, and context is critical. There are important respective considerations for de novo versus PSC or other cholangiopathy associated BDSs. Tissue diagnosis remains the gold standard datum for BDSs, but is not achieved in a clinically significant proportion of cases. Therefore, better tissue sampling capability should remain a primary focus of future efforts. In limited scenarios, data from advanced endoscopic or cytologic techniques may suffice to guide treatment for BDSs, but these techniques have room to improve in terms of objectivity, accuracy, availability, and cost-effectiveness. And the better their diagnostic performance moving forward, the more multidisciplinary buy-in there will be, and we really need to have incorporation of these techniques into and guidance from clinical practice guidelines in order to make for a more uniform and evidence-based approach. Thank you all for your attention, and I hope you've enjoyed this session and that you enjoy the liver meeting digital experience. Hello, everybody. At the outset, I would like to thank Dr. Levy and the organizers for giving me the opportunity to debate this topic of donor restricture should be worked up for IgG4 sclerosing cholangiopathy at all costs. I believe that the diagnosis of IgG4 sc is an art, and we have heard as medical interns and residents that when you hear hoofbeats behind you, don't expect to see a zebra. And indeed, IgG4 cholangiopathy is a rare disease. So how do you recognize a zebra among horses while at the same time not mistaking a rope for a snake? And also knowing that the eye doesn't see what the mind doesn't know. All these statements will make sense as we go along, but they are critical to understanding how we make the diagnosis of IgG4 sc without over-diagnosing it. The solution to all this is to get to know IgG4 sc, know when to suspect it, when not to suspect it, appreciate findings that overlap with more common diseases, and know how to confirm or not to confirm the diagnosis when it's not appropriate. So the cardinal features of IgG4 sclerosing cholangitis are in its imaging features, serology, other organ involvement, response to steroid therapy, and histology, knowing that only when you have histology of the whole organ do you actually have a diagnostic feature that can make a diagnosis independent of all other features. Everything else is like a piece of the puzzle, and one needs to understand the caveats that go with each of these features. Let's look at donor imaging in differential diagnosis. The disease can mimic pancreatic cancer almost identically when it causes a distal bile duct structure. It can mimic cholangic carcinoma when there is a proximal extra hepatic bile duct structure, and it can mimic an intra-hepatic cholangic carcinoma or a PSC-like picture when it is purely intra-hepatic and or intra- and extra-hepatic. So you can get the entire spectrum of very stricturing disease mimicked by IgG4 sc. And here you see a distal bile duct structure that mimics pancreatic cancer but is due to IgG4 sclerosing cholangitis. Similarly, you see a higher structure here which strongly resembles a cholangic carcinoma but was due to IgG4 sc. Finally, a disease picture identical to that of primary sclerosing cholangitis which is due to IgG4 sc. So are there distinguishing imaging features of biliru structures in IgG4 sc? We did a 17-physician intra-observer agreement study mixing up images of ERC images of primary sclerosing cholangitis, cholangic carcinoma, and IgG4 sc and found very poor intra-observer agreement. And this has been well-documented that you cannot make a diagnosis simply looking at images from the ERC. When you see a low-density mass on CT or MRI, one would strongly suspect that this is cancer, but IgG4 sc can also do it in about 15% of cases. So it overlaps other diseases radiographically quite a bit. So simply the imaging alone may not be sufficient to exclude or include IgG4 sc. So what about IgG4 in the differential diagnosis? Again, very commonly used but needs to be understood as for its categorization. Needs to be understood as for its caveats as well as its strengths. IgE4-SC is associated with elevated IgE4 in 75% of cases. So it is something that should be commonly seen in the setting of IgE4-SC. Unfortunately, 15% of mimickers, and sometimes 25% in some studies, of cholangiocarcinoma and PSC and even pancreatic cancer will have an elevated serum IgE4. And our general understanding is that twice the epitope normal is more likely to be IgE4-SC than mild elevations. But even that can be seen in these mimickers. And considering that IgE4-SC is quite uncommon compared to these mimickers, these numbers start to make sense in that they alone cannot make the diagnosis. So bottom line is that elevated IgE4 is helpful, but not diagnostic. What about other-order involvement? For me, this is a very important, but often overlooked clue to the diagnosis. There are combinations of bilirestrictors and other-order involvement that are simply not seen in other forms of bilirestrictioning. And so they should be looked at for, recognized and seen. The most common extra-pancreatic organ, extra-biliary organ involvement is the pancreas itself. And it's very difficult to diagnose IgE4-SC in the absence of pancreatic involvement and still remains the best predictor of the presence of IgE4-SC. Other organs involved include interstitial nephritis, retroperitoneal fibrosis, diffuse lymphadenopathy, lacrimal, parotid, and periorbital disease. I just saw a patient last week who had had the whole problem start with bilateral submandibular gland swelling. And one year later, nobody had still put the two together of a bilirestrictor and submandibular gland adenopathy. And the moment I saw the patient, I knew the diagnosis. But among all the physicians who had seen her in the last year, nobody had put the two together. So this is an important clue to the diagnosis of IgE4-SC. And here you see fairly dramatic retroperitoneal fibrosis. This is a biopsy of the submandibular gland showing almost all the changes you would see in the pancreas or the bile duct also seen in the submandibular gland. So these can be important clues to the diagnosis. You can see these kidney lesions that can also be diagnostic. And so one should be aware of these while they're looking at the bile duct. Response to steroids is invariable when given full doses. However, the time to response is variable. And some will relapse early, leading to a perception and not a response. This is another problem I see is patients are given a rapid taper of steroids over a period of a few weeks. And by the time the steroids are tapered, the disease has relapsed. And so you re-imagine the patient and you see, oh, there's no response. So that is a poorly done steroid trial and that can often be problematic in interpretation. So this is a classic resolution of bile restrictor using steroids. And you can see the significant improvement in just a few weeks. Same thing with this high restrictor. It's completely resolved with six weeks of steroids. So what are the problem with this? Now I'm starting to see surgeons also giving steroid trial and saying, if it works, it must be IG4SC. Unfortunately, it has to be interpreted with caution because it can have both false positive and false negative. The feeling of wellness is very difficult to separate from response. The moment you tell the patient you have benign disease and stuff, incurable disease, they feel better. Steroids themselves make you feel better. And there's a nonspecific decrease in liver tests with steroids, but they're unlikely to normalize unless they're mildly abnormal. There is an associated pancreatitis in some patients that resolves spontaneously and it's falsely considered as a response. Similar to the false negative side, you can have a fibrotic stricture that remains unchanged. I see patients many years later with completely distorted biliary tree that's not gonna respond to steroids. The other problem is re-imaging for response could be too early or too late. And in both cases, you will see that there's hardly any change and you would think there is no response. So even when you do a steroid trial, understand how it is to be given and how it is to be interpreted. What about histology? Then there are more and more devices that are coming up which are giving more and more access to the bile and visualization of the mucosa and so on and so forth. In our experience, endoscopic biopsies rarely provide histological evidence and fine needle aspiration is not useful to diagnose IgE4SC, more to exclude carcinoma. And in the biopsy, you're looking for more than 10 IgE4 positive cells for high-powered field. I would not recommend this on an FNA, but on a biopsy, this is often used. But this also has problems. So if you look at the ampulla, what I see most often is that patients have four or five IgE4 positive cells, everybody's very excited about it, but it's not really consistent with the diagnosis. And the truth is that ampullary biopsy is rarely positive unless the ampulla itself is visibly inflamed. So in my experience, it's an obviously inflamed ampulla that gives you the diagnosis. When it looks normal, the biopsies don't help you either. The bile duct biopsies, we were initially very enthusiastic. In our initial papers, we talked about 85% positivity in that. But as we did more and more of the PSCs and cholangiocarcinomas, we realized that there is a huge overlap between PSC and cholangiocarcinoma and IgE4SC. 30% of these case mimickers also have IgE4 positive cells, and therefore, it might be more useful to exclude cancer than to diagnose IgE4SC. Same thing with liver biopsy. My initial experience was that, oh, let's do this biopsy and maybe we can diagnose, just like we diagnosed PSC, we can diagnose IgE4SC. It's been much harder to diagnose IgE4SC on a liver biopsy, and it still may diagnose more advanced PSC, but we haven't found it very useful to diagnose IgE4SC. So how do you diagnose IgE4SC and cholangiitis? It still remains an art. It still remains putting the pieces together. The question is, how do you know which pieces are fitting and which are not? And this requires you to have a good knowledge of the disease. One other variable that it did not add is that 75% of IgE4SC patients are going to be males. So the other day, I saw a patient who was a female, was IgE4 negative. There was just too many probabilities of this not, and she had a low-density mass in her pancreas. Everything seemed like it was not going to fit. And yet, the surgeon gave her a course of steroids and everything disappeared, including the biliary dilatation, the pancreas dilatation, the pancreas itself, and it was so bad, the liver just normalized. So here you have nothing else fitting, a female with negative IgE4 and a low-density mass, all of which are relatively rare in IgE4SC. And yet, the steroid trial seemed to suggest that this was the diagnosis. So there is an art and a science to this, unfortunately, and there's no single diagnostic test for the disease. But if you know how these pieces fit, you can make the diagnosis. So, in summary, the best clue is the presence of AIP. Biliary involvement in AIP or proven other organ involvement is basically IgE4SC. An isolated biliary involvement is very hard to diagnose. We have had this discussion with the Japanese as well, who have a large experience with AIP and with IgE4SC. They agree that if you have just bile duct involvement with nothing else going on, and you look hard for other clues, it can be very difficult to diagnose. And that applies even to autoimmune pancreatitis, and I would say that you can't win every time, but with a good knowledge, you can avoid surgery or major interventions in the majority of the patients with this disease. My concluding remark is that if you're gonna go after it at all costs, know the disease well, and understand how to use and interpret the diagnostic tests. Otherwise, you will either under-diagnose it or over-diagnose it. And before you know it, the patient has had numerous expensive tests, and the data will be difficult to interpret. So, my plea would be for you to become familiar with the disease so that you can pick the zebra out of the horses. Thank you so much. Appreciate the opportunity. Good morning. Welcome to the ASLD endoscopy course. My name is Gideon Hirschfeld, and I'm a hepatologist in Toronto. Today's topic for discussion is around IgG4 cholangiopathies. Do you pursue at all costs, or for my component of the discussion, are you barking up the wrong tree? Now, of course, we have a lot of commonality, and really the debate is not as polarized, but I wanted to just give you some insights and thoughts as to how I think about this problem. These are my disclosures, none of which relate to IgG4-related diseases. So, are there four autoimmune hepatobiliary diseases, really? I know you're very comfortable with autoimmune hepatitis, primary biliary cholangitis, and primary sclerosing cholangitis. But in addition, of course, there is the rare IgG4-related diseases, which are important because they're highly treatment-responsive. Prednisone, Celsept, Azathioprine, Rituximab, and, of course, endoscopic therapy. And we also know that they have overlapping presentations with many processes, and when one considers autoimmune liver diseases, in particular with primary sclerosing cholangitis. This is a 29-year-old man, presents to your liver clinic. He's not jaundiced. That's a very important point. I don't know whether he has colitis, but he has an anechtoric cholestatic hepatitis. His IgG4 is elevated, not significantly, but definitely above average. And his MRCP shows a stricturing process. We need, therefore, to have a strategy for going forward in investigating such benign biliary strictures, and ensuring that the patients get the correct diagnosis. But that also means that rare presentations, such as specifically IgG4 sclerosing cholangitis, is not missed, and for those patients is treated. Now, benign biliary strictures have a variety of diseases, large duct PSC, relating to HIV infection, secondary sclerosing cholangitis, and now even related to COVID, recurrent pyogenic cholangitis, ischemia. I've seen a very significant sclerosing cholangitis from hereditary hemorrhagic telangiectasia, and as a consequence of pancreatitis. Now, IgG4 disease, at a high level, most often presents with pain and jaundice, not outpatient, has a number of different patterns of biliary strictures, but even those are not that specific, and hopefully is identified with a high IgG4, and very frequently, probably two thirds of the time, with other features of autoimmune pancreatitis. But through my PSC telescope, we have a lot of concern about missing IgG4 disease. When you think about the six C's of PSC, covert, cholangitis, cirrhosis, colitis, cancer, and the hope for cure, then our patients are very concerned that if they have an elevated IgG4, that we're not missing the opportunity to treat them very effectively with immunosuppression, which of course does not work in classical PSC-IBD, and more to the point, may carry adverse effects for patients who are unnecessarily immunosuppressed. So when we see someone with cholestatic liver chemistry, it's very important to exclude alternative etiologies, and to have an open mind that you're not missing a rare presentation of a rare disease, such as IgG4-related cirrhosis and cholangitis. And for that reason, we increasingly are measuring IgG4. Now, of course, when the patient has a more classical presentation with jaundice, this is definitely a little bit easier to follow through, and I don't think you'll over-diagnose so long as you're sensible. In patients with PSC, one needs to be extra cautious. When we looked at our clinic some years ago, we demonstrated as per evidence from the Mayo Clinic, that actually a significant number of patients with Crohn's and cholangitis, in particular PSC, will have elevations in their IgG4 without clearly having IgG4-related disease. Anywhere between 5% to 15% of the cohort will have an elevated IgG4. Quite what this means is not clear. Traditionally, it was associated with a more aggressive disease, but that data has not been replicated and hasn't been extended in very large, well-characterized cohorts. And when we think about the two diseases, a young person's disease, more commonly men, but a significant number of patients are women, very significant associated with inflammatory bowel disease and no response to steroids. And then IgG4-related disease, very significantly male-dominant, much older cohort usually, obstructive jaundice, long, smooth strictures, it raised elevations in IgG4, and importantly, quite infrequent inflammatory bowel disease. It appears to be so key in all of your decision-making, even more so with rare disease, to think about pretest probabilities, because they are key as to how far one does investigate and whether or not your patient needs to have endoscopic interventions. But of course, what we're talking about is an ulcerative colangitis. It's the same in PSC as it is in IgG4 disease, the main difference being the IgG4 presence histologically and the response to immunosuppression. Of course, with cholangioscopy, our patients do have the opportunity to get tissue biopsies, but nevertheless, we shouldn't be instrumenting the biliary tree and giving the patient risk of colangitis unless there's a real need. This is a small vignette. This gentleman fitted more for the concept of IgG4-related disease than PSC. He's in his 60s, he'd have a previous biliary stone, and then he's given the label of PSC, but clearly when you do his MRCP, what you see is hydronephrosis and pancreatic disease, and he has an elevated IgG4. He's treated, but over time, he gets interstitial lung disease, and then he's clearly got a systemic process that's immunosuppressive responsive. This is not classical PSC. This patient may not even need endoscopic intervention, but he needs effective treatment. And of course, when we're seeing these patients, we're very closely aligned to our radiologists, and the so-called sausage-shaped pancreas may be one of the features that will pick up the autoimmune pancreatitis, but additionally, other strictures and other disease outside of the hepatic biliary tree. And when you think about this fantastic disease, IgG4, it is so interesting that you have pancreatic involvement, lacrimal involvement, orbital involvement, lung involvement, thyroids, lymphadenopathy, the vasculature, the retroperitoneum for kidney physicians, kidneys themselves, the brain, and of course, the biliary tree, jaundice, weight loss, abdominal pain, closely mimicking PSC, but being distinct because it responds to steroids. And clearly, no one wants to miss that potentially reversible cholangiopathy. This amazing presentation shown in the 1999 Lancet paper of someone with devastating sclerosing cholangitis who responds fully to immunosuppression and then has a normal biliary tree. And you can classify the appearances of IgG4-related sclerosing cholangitis into four types. And within each of those types, you have a differential diagnosis that largely includes malignancy, chronic pancreatitis, PSCs we've talked about, and other malignancies such as including gallbladder cancer. Whether or not this definition is helpful for day-to-day clinical practice, I'm not sure. I think it's helpful for research and for communicating. And this slide is also helpful to remind you that you don't want to over-investigate. But if you do investigate, you know that the likelihood is you're going to have to do imaging, endoscopic ultrasound, ERCP plus or minus cholangioscopy, and you may need to do direct biopsies. When we look at what patients with IgG4 sclerosing cholangitis look like, I think this study from Asia is very, very helpful. First of all, the majority of patients did have a concurrent autoimmune pancreatitis. So that is, whilst not required, it is definitely present in, you know, maybe a third to half the patients. It may not help you for the other half. They do have significant elevations in their IgG4. And importantly, where they have appropriate histology, the IgG4 immunohistochemistry is usually very helpful. Steroid responsiveness remains a very important facet to look for in this disease. Cholangioscopy can be helpful. You can see dilated and tortuous vessels in IgG4-related sclerosing cholangitis. A careful look at the duodenal papilla can see a swollen duodenal papilla. And as I said, histology can be very, very key so that you can look for significant, not just a few, but significant IgG4-bearing plasma cells. But IgG4 measurement alone in the serum is not perfect. It's characteristic, but not diagnostic. You need to be seeing patients who have very significant elevations for it to increase the specificity for autoimmune pancreatitis, IgG4-related disease. And some patients with pancreatic cancer, cholangiocarcinoma, and even now proven IBD, will have elevated IgG4, just as we see in primary sclerosing cholangitis. So alone, you must not get carried away with a single IgG4 measurement, and you must put it into context, and you must put it into the pretest probabilities that you're working on. The ratio of IgG4 to IgG1 can be helpful. This is a paper from 2014 that's trying to differentiate PSE from essentially IgG4-related sclerosing cholangitis. And you can see that the ratio of IgG4 to IgG1 goes up, the more likely you are to have IgG4-related sclerosing cholangitis. So measuring IgG1 can be helpful, but again, it's not perfect. There was some significant research effort into looking for B-cell receptor clones, and in the original paper from 2016, it did appear that clones may help. Subsequent studies trying to extend and validate this do not seem to support this, but the concept that ultimately we'll have molecular approaches to diagnosing patients to avoid ERCP and endoscopic procedures is very exciting. So where are you left? First and foremost, please use pretest probabilities. Please don't go barking down the wrong door for something which is a very rare disease. Secondly, please have very close working relationships with your radiologists, because in all the diagnostic criteria which encompass IgG4 disease, the appearances by imaging is absolutely key. And we have one go-to radiologist in particular that we go and show all the imaging to because their experience grows over time. But remember essentially looking at imaging, laboratory, histology, other organ involvement, and response to steroids. Fundamentally, did your patient present with jaundice and duct dilatation? Our patient did not. When our patient had a colonoscopy, they had IBD. This IgG4 is nothing, therefore, to relate to IgG4-related disease, and I don't need to put this patient through an investigation to look for IgG4-related sclerosis and colonoscopy by endoscopy. But if you're jaundiced, if you've got duct dilatation, if you've got abnormal imaging beyond the biliary tree that shows pancreatic involvement, other extra biliary involvement, then of course you're gonna be doing more CTs, endoscopic ultrasounds, ERCP, and cholangioscopy. And you're gonna then put it together with the exclusion of other diagnoses, the elevations in the serum IgG4, and you may well then get some tissue diagnosis to see whether you can confirm it. And that's in that very small number of patients when you're going to start a trial of steroids. In this way, you're not gonna over-treat PSE. You're not gonna under-diagnose malignancy. But in the rare patient that presents to you, usually male, usually in their over 50s, 60s, and 70s, you'll make a very big difference. So I hope I've given my case as to why this is an important disease. And you should certainly consider it as a diagnosis at the right time, but not in the wrong place. So I hope that helps stimulate some discussion. OK, everybody, well, thank you for joining me. I'm excited to give you a talk today on artificial intelligence. My name is Neil Marriott, and today we'll be discussing AI. Is it improving intelligence or simply artificial? I'm an assistant clinical professor of medicine with training in advanced endoscopy and bariatric endoscopy. I have a particular research interest in artificial intelligence and its applications to advanced endoscopy, and I'm currently at the University of Massachusetts Medical School. These are my disclosures. Let's start with an overview of today's talk. We will begin with a discussion of the history of machine learning. We will then get into the different AI applications that have been put forward for both gastroenterology and hepatology. We will spend the majority of our time discussing a particular study that we did, looking at EUS as a platform for an AI model with a specific interest at addressing focal liver lesions. We will then end with some future directions of AI research for both hepatology and endoscopy. So let's start with our discussion on the history of machine learning. When we think about artificial intelligence and these complex neural networks, it's important to realize that these concepts were actually started back in the late 1940s by this gentleman, Donald Hebb. Now, Dr. Hebb was a Canadian neuropsychologist who had particular interest in how learned behaviors originate from cellular connections. And he put forth his original theories in this book, The Organization of Behavior. In his book, Dr. Hebb describes his theories and summarizes his observations by suggesting that when two neuronal cells that are connected to one another fire simultaneously, that axonal-somal connection between the two cells strengthens significantly. Alternatively, if you have two cells that are joined together but routinely do not fire at the same time, that connection between the two cells will weaken. The whole concept can be summarized by this notion, that neurons that wire together fire together. It is truly this simple concept that has been maintained as the basis of all neural network frameworks, even those that we use today. Now, moving into the 1950s, we come to this man, Arthur Samuel. Samuel worked for IBM at a time where corporations and government enterprises, like the Rand Corporation, had become particularly interested in game theory as a method to both model and potentially even settle conflicts. At IBM, Samuel was charged with developing a computer program for playing checkers. His design included a scoring function using the positions of the pieces on the board and associating them with values. The scoring function attempted to measure the chances of each side of winning. Now, Samuel also designed a number of mechanisms allowing his program to become better by learning. In what Samuel called rote learning, his program recorded and remembered all positions it had already seen and combined this with the associated positive or negative values of each position on the board. It was Samuel who actually first came up with the phrase machine learning in 1952. By the late 1950s, the United States government had actually become quite interested in the progress of neural networks. In 1958, the United States Navy provided a significant amount of funding to Frank Rosenblatt for what is considered to be the first attempt at creating an artificial intelligence capable of image analysis and perception. Developed at IBM, Rosenblatt created what he called the Mark I Perceptron. Now, the Mark I Perceptron is an example of a single layer network connecting inputs and outputs. There was initial promise that the Perceptron could be used to perform facial recognition and speech translation tasks. It proved incapable, however, during many trials of complex pattern recognition. It was unfortunately these failures that resulted in the concept of machine learning to languish for almost a decade until the development of multiple layer networks. Now, similar to the single layer Perceptron, multilayered networks also start with an input layer comprised of nodes. These nodes then interact with hidden layers. And within these hidden layers, multiple convolutions and processes occur that result in new weights and new connections that eventually output into an output layer that provides us with a final prediction. To better understand the concept of multilayered networks, let's take an example of a network that is trained to take images of different animals and provide a prediction as to whether or not the imaged animal is actually a dog. Now, in this network, you first see that the initial layer is simply looking at very sort of simple shapes like curves and edges. It is higher layers that are responsible for taking a look at much more complex features like the dog's tail or the dog's nose or the dog's foot. And then it's really the highest layer that takes the summation of all these different nodes from the previous lower layers to provide prediction as to the actual shape of the dog before actually providing a prediction that tells us what is the likelihood that the original image in that input layer was actually a dog and not a wolf. It has now been well demonstrated that multilayered networks are really required in order to perform complex tasks such as image recognition. In addition to the development of the types of neural networks that have been able to be developed, improvements in technology have also been key to the rise of modern AI. This concept is well illustrated by Moore's law. Moore's law suggests that every two years, the number of transistors that can fit on a CPU chip will double. This improvement in technology has resulted in an increase in the computational speeds for artificial intelligence models. This is well demonstrated in the graph shown here. On the x-axis, you're seeing the year of introduction of different AI architecture models. And on the y-axis, you're seeing a unit of measurement referred to as a flop or a floating operation unit, which is a measure of computational speed. This combination of an improvement in technology and the types of neural networks that we are using has resulted in the evolution from artificial intelligence to machine learning to deep learning. As we discussed previously, artificial intelligence really started in the 1950s, but it's also the overarching concept that contains both machine learning and deep learning. The goal of artificial intelligence is to generate models and programs that are capable of mimicking human behavior. From within artificial intelligence comes machine learning, which is a concept where programs improve as they are exposed to additional data. And then finally, that brings us to deep learning, which is a subset of machine learning. In deep learning, models are learning using multilayered networks and really vast amounts of data that are available for training. Convolutional neural networks, or CNNs, are examples of deep learning algorithms that are specifically used for image data. Now that we have a good understanding of the history of artificial intelligence, let's move on to the different AI applications that have been developed for hepatology and gastroenterology. Similar to other specialties, there has been a big increase in the number of AI publications within GI and hepatology. And this seems only fitting, provided the vast databases that already exist, as well as the large amounts of endoscopic data that have been accrued over the last few decades. This is well demonstrated in this plot here. Now, there are several things to take away from this graph. But what I find most interesting is the big jump that occurred between 2015 and 2016. Between 2015 and 2016, the number of AI publications in GI and hepatology more than doubled. There are several factors that may account for this. First, the types of neural network architectures available for use significantly change in regards to complexity over the last decade. Perhaps the most commonly used deep neural network available on the market today is ResNet-50. And that was first introduced to us around 2015. Also, between 2015 and 2016, there was a significant jump in the types of graphics cards that were made available to the public for use. This was thought to be in response to the intensity of the Bitcoin mining operations that were done around the world and a need for high-powered GPUs. And that trend is only due to continue, as increasingly higher-powered GPUs are made available to the public at actually a reasonable cost. Again, this combination of complex neural networks and improved technology has allowed for the democratization of AI research. This figure here demonstrates the type of research that has been done in GI and hepatology so far. As we can see, the majority of studies that have been done have focused on malignant and pre-malignant intestinal lesions, such as colon polyps, Barrett's esophagus, and attempts at early detection of gastric cancer. While some of these studies use vast databases that are obtained through the medical record, the majority use endoscopic data. AI models have also been generated to help with the diagnosis of different inflammatory nonmalignant intestinal lesions, such as celiac disease and IBD. For GI bleeding, provided that there is massive amounts of image data for each capsule endoscopy exam, that has been a major focus of AI research. Additionally, medical record data Additionally, medical record data pertaining to morbidity and mortality, as well as readmission rates, has also been used to generate AI models. Within hepatology, medical record data and imaging data have both been used to help predict which patients suffering from viral hepatitis, as well as NASH, will develop advanced fibrosis. Additionally, medical record data and endoscopic data has been used to help identify which patients are most in need of variceal screening. Finally, the area that has actually received the least amount of attention in regards to AI development has been pancreatic disorders. Prior AI models have used contrast-enhancing endoscopic ultrasound and less advanced neural networks to differentiate PDAC from chronic pancreatitis. Now, when we look at what endoscopic platforms have been used to generate AI models, we see that video capsules have actually been the most commonly used platform. Now, the reasoning for this is relatively simple. Each capsule endoscopy examination is comprised of tens of thousands of images. Thus, within one capsule examination, you already have a treasure trove of data that can be used for model training. Next, we see that colonoscopy and upper endoscopy are the next most common platforms that have been used for AI development. This again makes sense, as both colonoscopy and upper endoscopy are the two most commonly performed endoscopic procedures around the world. The endoscopic platform that has received the least amount of attention, especially with modern AI models, has been EUS. This is likely due to the fact that not every center is capable of EUS, and also because of the lack of data that is available. Now, when our group became interested in trying to apply artificial intelligence to EUS as a platform, we began to think about what specific pathologies were of most interest to us. Certainly, the development of models that focused on the evaluation and differentiation of pancreatic disorders has been an initial focus of ours. We also recognized that no prior AI models had been developed using EUS as a platform for the evaluation of the liver. Thus, we felt that there was an interesting opportunity for us if we could use EUS to generate an AI model that could investigate liver pathology. So this leads into our project, where we sought to develop an EUS-based convolutional neural network model that would be able to successfully identify and classify focal liver lesions as being benign and malignant autonomously. The identification and classification of focal liver lesions, like those seen here, are key for optimizing patient outcomes. For example, patients with underlying malignancy who are found to have a focal liver lesion may no longer be a candidate for curative resection if that lesion is found to be a metastatic deposit. Now, this is relevant, as 17% to 30% of patients with an underlying malignancy are found to have an FLL on cross-sectional imaging. And of those patients where cross-sectional imaging is unable to determine whether or not the lesion is benign or malignant, 11% to 37% of those lesions end up being metastatic deposits. Certainly, cross-sectional imaging is the gold standard diagnostic tool available for the diagnosis of FLLs. With that being said, there is still a role for EUS in the evaluation of FLLs. EUS has been demonstrated, like in this example, to isolate FLLs with relatively good accuracy. In addition to visualization of these FLLs, EUS provides an opportunity to sample suspicious lesions. And studies have also demonstrated that in cases of underlying malignancy, EUS is capable of identifying distant metastases in 5% to 20% of cases. And finally, we have also demonstrated that when using expert-derived criteria, EUS is capable of 85% sensitivity and 82% specificity for the correct classification of FLLs as being benign or malignant. Now, while there is a significant benefit to EUS in the evaluation of FLLs, the role of EUS in this realm can still be optimized. In these examples, I show how EUS can identify FLLs, but how the appearance can be sometimes confusing. In this first case, we're seeing a hyper-echoic lesion consistent with a hemangioma. In this second case, you're seeing another sort of hypo-echoic lesion, which is more consistent with a malignant deposit. In this case, this was metastatic pancreatic cancer. But finally, in this case, right at the top there, you see another hypo-echoic lesion right at the border of the liver and the transducer. And while this looks similar to the second lesion that ended up being a metastatic deposit, this is actually a case of focal fat spread. This is actually a case of focal fat sparing. And while there may be some who suggest that sampling the lesion, if it's suspicious, is a good enough solution to any question as to whether or not lesion is malignant or benign, there are still significant risks associated with FNA or FMB of FLLs. Risks include bleeding, infection, and even transperitoneal seeding of the tumor. Thus, artificial intelligence and specifically AI-based image analysis tools could be highly useful in this realm. Previously, a group from France had a similar idea to use transabdominal ultrasound as a platform for an AI model. In this study, the authors used bounding boxes, which are the gold boxes seen in the top two panels, to focus the attention of the AI. Then, extracted data was put through the model as shown in the bottom frame to provide a classification score as to whether or not the lesion was a benign or malignant process. Using this technique, the authors generated a model that was 85% sensitive and 80% specific for FLL identification. And for classification of the lesion as being malignant or benign, the model was 100% sensitive and 60% specific. Now, for our study, we used a different process for training our AI. Unlike the prior study, we did not use bounding boxes or other types of expert-derived attention mechanisms to focus our AI. Instead, we used a technique called occlusion heat map analysis to allow the AI to autonomously identify lesions on its own. Once we had trained a model that was capable of successful identification of FLLs, we then turned our attention to FLL classification. As with any AI study, it was key to have a gold standard metric for establishing a true diagnosis. In our study, an FLL was only considered malignant if a histologic or cytologic specimen was available that confirmed malignancy, or if follow-up imaging after the EUS FNA clearly demonstrated a malignant process. Alternatively, an FLL was considered benign if a CT scan or MRI scan had been performed six months after the EUS that demonstrated that the leash was clearly benign, or if the FLL was no longer visualized. And finally, an FLL was also considered benign if the patient was still alive and had no clinical evidence of a malignant liver disease for at least two years following the index EUS. Our final database was comprised of 256 total patients. And from those patients, there were 631 still EUS image and 310 EUS video assets. With that data, we had a final database that comprised of over 200,000 unique EUS images. We then randomly assigned our patients to either being in the training and validation set or the test subset. As per the norm, the vast majority of the patient data was assigned to training and validation sets. We intentionally kept all the patients with diffusely benign livers in the training and validation subset, as the focus of this AI model was the identification and classification of FLLs. Of those with benign FLLs, the most common lesions were hemangiomas and areas of focal fat sparing. Now, for those who had malignant FLLs, the most common lesions in our database were metastatic. Specifically, the three most common FLLs in our database were metastatic PDAC, metastatic esophageal cancer, and metastatic pancreatic neuroendocrine tumors. To demonstrate that our CNN model was capable of autonomous FLL identification, we did not use expert-derived tented mechanisms like bounding boxes, as I had mentioned previously. Instead, we used a technique called occlusion heat map analysis. In occlusion heat map analysis, you sequentially apply different colored boxes over your index image, and then you different colored boxes over your index image, and quote, unquote, occlude that part of the image from the model. With that part of the image occluded, you then run your model and see what the prediction score is for the result of interest. In this case, we are interested in whether or not the FLL is present, or if it is not present. Thus, if we occlude a part of the image where the FLL is present, the score of the CNN model telling us that the FLL is present should drop considerably. You then continuously run different models or different predictions, where you are blocking different parts of the image, and you see how much your score drops. You then color code each occlusion block that you had placed over the image by the degree that the score decreased. In our case, if an area of the image becomes more purple, that implies that that was an area of importance to the model in detecting the FLL. When this is applied systematically, you obtain what's called an occlusion heat map, like shown here. When this is overlaid on top of the EUS image, you can see that the FLL was identified correctly by the CNN model. In the next series of examples, we demonstrate how our CNN model is capable of autonomous FLL identification. Overall, our model was 92% accurate in detecting FLLs. In regards to classification of an FLL's benign or malignant, this ROC curve demonstrates the C9 model accuracy. When exposed to all images, the model was 90% sensitive and 71% specific. This ROC curve demonstrates the classification accuracy for just videos. When just provided video data, the C9 model was 100% sensitive and 80% specific for the correct classification of an FLL's benign or malignant. While our study does have some impressive results, there are limitations. First this is a single center retrospective study, thus the generalizability of our results cannot be confirmed. Second, there are limitations of EUS as a modality for liver assessment, specifically as it relates to FLLs. Firstly, EUS is very good at evaluating the left lobe of the liver, but not necessarily the right lobe of the liver. Secondly, we know that cross-sectional imaging is the gold standard for FLL identification and classification, although it is important to understand that EUS is perhaps better than cross-sectional imaging when it comes to the evaluation of subcentimeter FLLs. We also did not perform a head-to-head comparison of the AI against the established human-derived criteria. And finally, as is a pitfall for almost all AIs, we do not know how this will actually work prospectively when it is implemented into clinical practice. These concerns were perhaps best demonstrated in a recent article by Winkler et al. in JAMA Dermatology. In their study, the authors used a commercialized AI that was designed to differentiate benign and malignant nevi. The authors demonstrated that when you use the commercialized AI on an unmarked benign nevus, it was able to correctly classify the lesion as benign, but when you simply put surgical markings around the benign nevus, all of a sudden the AI thought that it was a malignant nevus and jumped the score up to .99. And in their final experiment, when they simply just cropped the nevus, still the score was maintained high as malignant, despite it being a benign process. This is obviously concerning, as it suggests that the original model was focusing on non-clinically important criteria. In thinking about future directions of AI, I think there are many directions to go for research. First, we need to think about how we're going to prospectively implement our existing models so that we can validate them. Then we should be also thinking about other clinical dilemmas that require further or new methodologies where AI could be applied. Below are just some examples that I've been thinking about. Can we use endoscopic appearance of varices as a way to re-stratify them using an image-based AI like a convolutional neural network? Can EUS be used for the evaluation of NASH to predict those who are going to have accelerated fibrosis? Can we also use EUS images to prognosticate those who have underlying liver disease? Then finally, what is the role of other types of endoscopic platforms like ERCP and cholangioscopy? There's just simply a lot left to do. For some final takeaway points, with the new emergence of graphics cards and multi-layered neural networks, artificial intelligence research is here, and it's up to us to figure out how we're going to use it. Certainly, utilizing our large databases, especially those with images like endoscopic databases, are really a key resource. I believe that the intersection of endoscopy and hepatology represents a unique opportunity for all of us to collaborate and develop some important AI models that can directly impact patient care and outcomes. Thank you so much for your attention, and I hope you enjoyed today's talk. Thank you.

Endoscopic ultrasound-guided gallbladder drainage

Percutaneous cholecystostomies

Acute cholecystitis

Stenting

Recurrent cholecystitis

Bile duct strictures

Diagnostic methods

Artificial intelligence advancements

Machine learning models

Gastroenterology

Hepatology

Lesion detection

Capsule endoscopy

Liver assessments

Focal liver lesions