Good afternoon. I hope everybody had a good lunch, and we'll get started with the afternoon session. Now I'm going to talk about diagnosis of battered esophagus getting to the next level. This would involve describing the various advanced imaging modalities used to improve detection of dysplasia within the battered segment and also evaluate the underlying mechanisms and answer if they should be incorporated in routine clinical practice. Going over some of the basics, the current definition of battered esophagus is the extension of the salmon-colored mucosa at least one centimeter or more proximal to the gastroesophageal junction plus the second criteria of intestinal metaplasia. One thing to know is that the intestinal metaplasia is patchy, and that's the reason for the recommendation of doing at least eight biopsies as Dr. Iyer has alluded to earlier. Does it make any difference whether we do it with a large capacity forceps or the jumbo biopsy forceps? The answer is no. I know most endoscopy practices use the large capacity biopsy forceps, so those are completely adequate for doing biopsies in battered esophagus. Here is the PROC criteria. Just to recap again, the first thing is to identify the diaphragmatic impression and then figure out the hiatal hernia, so that's the distance between the diaphragmatic impression and then the upper end of the gastric folds. The circumferential extent of the battered is extending from the upper end of the gastric folds to the circumferential extent, and then the maximal extent is from the upper end of the gastric folds to the maximal extent of the battered esophagus, not from here to there, but from here to there is the maximal extent. I see some endoscopists reporting this as a regular z-line in a patient with long segment battered esophagus if there's not much of a difference between the circumferential and the maximal. That should not be used. Irregular z-line is when the variation at the gastroesophageal junction is somewhere between five millimeters to one centimeter or less. One thing to remember, too, is that the PROC classification does not take into account the islets, which are discrete areas of salmon mucosa away from the circumferential or the maximal extent. So PATIS classification is widely used for describing the visible lesions within the battered esophagus, and usually the ones which are amenable for endoscopic therapy are the type 0. These are the superficial, polypart, flat, depressed, or excavated tumors. So how do we make a difference between this and this? This is more than 2.5 millimeters, and this is less than 2.5 millimeters. 01 polypart ones include the polypart and then the sessile, and then 02A, slightly elevated, flat, and then the depressed. So these are the visible lesions which are amenable for endoscopic therapy. And then the excavated is when it is extending beyond the mucosa into the submucosa. So recently there was a study, which is an advance over the PROC classification, about measuring the battered surface area. So it is not just a circumferential and then the maximal extent, it is actually to describe the area. So it is a deep learning technique. So what is a deep learning technique? It can think like a human brain and come up. It is fed a lot of input, a lot of data, and then can come up with its own predictions or recognize patterns which the human brain is not able to. So in this study, they measured the circumferential and the maximal extent using a depth estimator network to predict the distance from the castric pores, and then quantified the area of the battered epithelium as well as the islands too, and then used a three-dimensional reconstruction of a phantom esophagus, and then used this system to see if it is able to validate the area of the battered esophagus. And then, I'm sorry, this phantom esophagus is used to train this deep learning algorithm. And then it is validated on high definition videos from 131 patients and compared to expert endoscopies for the CNM values. So here is the picture. So this is the endoscopic image. And then this is the predicted depth. And then a color gradient is used to measure the length along the esophagus. And then here are the three-dimensional images of the reconstructed esophagus in these patients. So one advantage of this, eventually it is an advancement over the CNM, and this would help in describing more accurately the remaining battered esophagus, and it would be of particular value when we are doing the ablation patients. So we see large areas of islands sometimes in these patients with batters, but they're not in the CNM criteria will give you a spuriously low value about the extent of batters in these patients. Coming on to the different advanced imaging techniques, there are several chromoendoscopy, either using the electronic, which is an out-of-band imaging, or using the colored dyes, acetic acid, methylene blue, and then the magnification endoscopy, then the autofluorescence imaging, and confocal endomicroscopy, OCT, high-resolution microendoscopy, multispectral scanning, and molecular imaging. The goals of the advanced endoscopy techniques, ultimately, is to improve the visualization, better visualization, ability to recognize the neoplasia, and improve the patient outcomes, and then reduce the cost. So with those as the end goals, the different criteria for an advanced imaging technique is to improve the effectiveness of the screening and then the surveillance, and then be able to reduce the number of biopsies we have to do to figure out. So for now, doing the Seattle protocol in, say, for example, a 10-centimeter segment of batters or so, if you're doing every one centimeter, four segments, so you're talking about 44 specimens at 11 levels, and then guide the endoscopic eradication therapies too. Sometimes we have this varied neoplasia with a neo-squamous epithelium, so this would help in detecting the neoplasia, but it's sub-squamous neoplasia, and direct the endoscopic therapies to these areas. So coming on to the narrowband imaging, the electronic chroma endoscopy, the advantage is there is no need for the dye spraying, and then it can visualize the mucosa using the light filters and then the computer processing technology. So what are available, the narrowband imaging with Olympus, and then FICE with a Fugenon, or the iSCAN with the Pentax. There is a classification system for the findings on the narrowband imaging. Essentially, two things that you're looking at, one is the mucosa, second one is the vascular pattern, and in each of them, you're looking whether it is regular or irregular, so this simplifies the classification system. So if you look at it in different ways, the recognition is very easy, so in picture A, you're looking at the RON pattern, and then figure B, you're looking at the RICHT or the Willis pattern, so both are regular mucosal patterns. And then here is an absent mucosal pattern, which is abnormal or considered as irregular. Now here, you're looking at the vessels going along with regular branching or so, so this is a regular vascular pattern. And here we are looking at an irregular vascular pattern, so there is distortion of the architecture, and then in the irregular vascular pattern too, you see the irregularity of the blood vessels. Now using these, so it is very easy to classify, it is kind of intuitive, and when you look at that, and look at how it does in terms of the sensitivity and the specificity, when they describe, when the experts who looked at these images, when they interpreted them or classified them as regular or irregular mucosal pattern and vascular pattern, and when they said that with high confidence, we have really good values up here, 92% accuracy and 92% sensitivity, 93% specificity, and so on. But however, recognize that these results are among the experts, and it requires external validation and evaluation among the trainees, and then non-experts. So this is a study on 65 patients with battered esophagus undergoing previously, evaluation for previously detected dysplasia. So first, a standard white light endoscopy was used to detect the visible lesions. Then a second gastroenterologist came into the room and then did the nanoband imaging to detect and biopsies the area suspicious for dysplasia. So the lesions initially detected by the standard endoscopy were then disclosed and biopsies. So based on this, they were able to do a comparison of how nanoband imaging detected compared to the white light or the random four quadrant biopsies. So with the nanoband imaging, it was able to detect about 57% of what was diagnosed on random biopsies. And then it was able to detect 12 of the 65 with higher grades of dysplasia. And it also helped with reducing the number of biopsies per patient. So the mean number of biopsies in a patient, NBI targeted biopsies were 4.7, whereas with random biopsies, it was 8.5. Then going on to the conventional chromoendoscopy, that is the spraying of the dyes to improve the characterization of the mucosa. And then the different dyes used are methylene blue, Lugol's iodine, indigo carbon, and then the acetic acid. One problem with the conventional chromoendoscopy is that you have to take the dye and then spray it. And then the dye uptake may not be uniform along the mucosa. And then it would take some time to do that. And also there is a lack of standardized classification system. And that is the reason with the availability of the nanoband imaging, I think using a regular dye endoscopy has become almost obsolete. So in a meta-analysis, looking at the chromoendoscopy, the increased diagnostic yield was by about one-third, 34%. And then there was no difference between the virtual chromoendoscopy and then the conventional chromoendoscopy. So going on to the next technique, the VLE. VLE is a volumetric laser endomicroscopy, is a second generation OCT or optical frequency domain imaging. And it provides high speed and high resolution surface and subsurface images of the entire battery segment. So you pass a balloon, it's a catheter with a balloon at the tip, and then you pass it. And then when it is inflated, we can get, similar to an EOS, so we can get the images of the surface mucosa as well as the submucosa for up to three millimeters depth or so. So based on that, when there is complete effacement, you can classify based on the findings, whether it is dysplastic or non-dysplastic. So here you're looking at the surface is darker than the subsurface. This is considered as dysplastic. And in here, the subsurface intensity is more than the surface, that is normal. And then with partial effacement of the balloon, you're looking for the submucosal glands. So presence of glands more than five is dysplastic, and then less than five is considered as non-dysplastic. So the sensitivity, specificity, and accuracy are in the high 80s ranges. One problem with the VLE though, is it requires a lot of time. It adds on the average about 20 minutes to each procedure or so. So for this reason, they came up with, if automated image analysis is useful, and then when you're looking at the area under the curve, with the experts it is 0.8, and with this computer-aided detection system, it is 0.9 or so. So this is one promising technique to look at to make things easier while using VLE. The next one is autofluorescence imaging. And then this is from the Netherlands group. Here is a white light image. You're seeing a visible abnormality there. And then this is what it looks like under autofluorescence imaging. It requires a special endoscope. And the advantage with that was in those 84 patients, 102 abnormalities were detected on the autofluorescence imaging, which were not detected by a regular white light endoscopy. But the false positive rate for neoplasia was 81%. So it's not doing us any favors. And then by adding an adoban imaging, the false positivity rate was reduced. So it would add more time, but not adding much in terms of the outcomes or so. So this is the ASG PV statement that is a preservation and the incorporation of the valuable endoscopic innovations. I mainly come up with this to establish thresholds at which the endoscopic techniques or technologies are designed to resolve the clinical questions. So as per the ASG PV statement for an advanced imaging technique, for it to be used in the advanced imaging technique, for it to be usable, the sensitivity has to be more than 90%. And then the negative predictive value has to be more than 98% for detecting high-grade dysplasia for early esophageal cancer. And then the specificity of 80% to allow for reduction in the number of biopsies. So with that, only two appear to meet the PV thresholds. And then one is an adoban imaging, and then second one is the acetic acid diet. So after the advanced imaging techniques, let's come down to the basics. And then Dr. Iyer has mentioned this already. The most important is careful inspection, taking your time to go through the esophageal segment from the proximal esophagus to the distal esophagus. That is what is most valuable. So these are the patients referred for possible endoscopic therapy. And then when more time is spent, so the detection of suspicious lesions increased from 13% to 54%. And then the high-grade dysplasia, the esophageal cancer detection rate went from 6% to 40% or so. However, I do have to mention that these are the patients referred for endoscopic therapy. So they did have some, either on the random biopsies or so, some degree of neoplasia. So the high-resolution endoscopy on the left-hand side is to the standard of white light. And then here is the high definition. It is not the same patient. It is a different patient. But you can see clearly the degree of mucosal detail that you're able to see with the high definition endoscopy. Now here is a patient referred. And does anybody see the subtle abnormality? It is to the right, one o'clock. And then here is with the narrowband imaging. Now it is much more prominent. And then here it is with partial deflation or so. So you're able to see it with greater definition. And then here is after the EMR, and the patient turned out to have intramucosal cancer. Another one here is taking time to recognize those subtle abnormalities. That is the most important part. So here is a patient going through the baris endoscopic eradication therapy. You don't see much on the white light endoscopy. With the narrowband imaging, this part, this becomes much more prominent. The other thing to do in this patient is doing deep retroflexion into the hernial sac. So this is the regular retroflexion. So when the scope is passed into the hernial sac, this is an area of abnormality. And then here it is. And then after doing an EMR, this turned out to be high grade dysplasia. So after recognizing this, I came back unretroflexed and was coming back into the esophagus. This is what I saw. Now the location of the early esophageal cancer. So these are the different case series. And then most of the cancers are located in the right upper corner. So if you're looking at the right hemisphere or so, this is where to focus on. It is, in fact, a web-based teaching tool developed for improving the detection and delineation of the baris esophagus-related neoplasia. And these are the, this is a free teaching tool useful for going through that site and to go through those images and see if you're able to detect this subtle abnormalities. The main thing which is needed is an annoying eye. So it is like the eyes cannot see what the mind does not know. The main thing is to recognize, being able to recognize these visible abnormalities. So using this tool, the median detection rates increased by about 30% or so after the training. So this is the baris esophagus-related neoplasia project by the Netherlands group. So in phase one, all these videos have been looked at by the expert endoscopists experienced in evaluation of baris esophagus and baris eradication therapy. And then they marked the sites of the abnormalities as well as the margins or so. So in the phase one of the project, 80 videos were read by 68 endoscopists, and then feedback was given to them. And in phase two, after refining that in phase two, 121 new endoscopists read about five videos, and then they went through the training sessions of 20 videos and then took this test again post video. So with that, there was a 46% increase in the detection rate and 130% increase in the delineation rate of the lesion. So as you know, artificial intelligence is permeating into all aspects of our life, and it has come up in detection of the baris neoplasia too. So this is again reported by the Netherlands group, very elegant study. So a very elegant, very meticulously done. So here is initially the deep learning algorithm. It was trained on images from the esophagus stomach and different parts of the digestive system, about 500,000 images or so. And then again, it was trained on, so this is the input. Again, it was trained on 1,250 images of early baris neoplasia and then non-displastic baris. And if the neoplastic lesion was delineated by baris experts, a validation, internal validation was done on data sets of about 300 images of baris neoplasia and then the non-displastic baris or so. And then they tested the data set four, and then the data set five. So interesting, look at this algorithm. The accuracy for data set four with this deep learning algorithm is about 89%, sensitivity and specificity about 100%. And then the delineation of the area was about 100% or so. So the data set five was read by both the algorithm as well as by endoscopists, other regular non-expert endoscopists. So in fact, the deep learning algorithm performed better than the endoscopist. Look at this number. So in the terms of the accuracy, it's about 88% versus 73% of the human endoscopist. So in summary, the main tips to improve the dysplasia, improve the dysplasia detection rates in baris esophagus. One, you don't want the patient retching and being restless. You want to do it under adequate sedation. Second one, use a high definition scope. And if there is any mucus or so, you can use mucamins to get rid of the mucus, use water, take your time inspecting and obtaining multiple images, and then carefully scrutinize the right side of the baris segment. And at the GA junction, especially to separate the folds, you can use a distal esophageal cap and then document the landmarks, the proc classification, and then the visible lesions for the baris classification. And first biopsy the visible lesions, followed by the Seattle protocol. So Seattle protocol in non-known patients with dysplasia is four quadrant biopsies every two centimeters. In a patient with non-dysplasia or suspected dysplasia, it is four quadrant biopsies every one centimeter. Thank you.

dysplasia

battered esophagus

intestinal metaplasia

advanced imaging techniques

careful inspection

artificial intelligence