Computer generated transcript

Warning!
The following transcript was generated automatically from the content and has not been checked or corrected manually.

Uh This is uh uh I would be your moderator for uh for today's webinar. Uh This is uh the last webinar of 2022. Uh It is a joint uh initiative of the I PBS and the OC the is a European network of uh trainees in obstetrics and gynecology. And uh so, uh we have uh with us uh three distinguished uh speakers uh from uh Sarah Feldman from the USA er from Austria and Nicholas Entin from uh the USA uh doc, Doctor Feldman uh is uh a professor uh of uh obstetrics and gynecology. Uh She's uh at the Harvard Medical School and the Brigham and Women's uh Hospital. Uh She oversees the Colposcopy and ambulatory gynecology oncology clinics at this uh hospital. And she has been involved in numerous research projects uh and also has sat on multiple national and international committees at the SG O the A S CCP, the IP DS, of course, the American Cancer Society and the others. Uh Doctor uh Vinsen uh is uh a senior investigator and Deputy Chief of the Clinical Genetics Branch and head of the Clinical Epidemiology Unit in the division of cancer epidemiology and Genetics at the National Cancer Institute. Uh Doctor Vinson is an uh international expert in molecular and clinical epidemiology of gynecological cancers with a focus on cervical disease detection and the identification and validation of novel biomarkers for a translation into new prevention methods. Uh He regularly serves as an expert on national and international committees focused on gynecological and other uh cancers. He's Coprin investigator of the NCI Moonshot project to accelerate cervical cancer control. Uh Doctor Ya is a professor uh of uh gynecology in uh the Medical University of Vienna in Austria. Uh He, he graduated from uh university of uh grad uh subsequently worked at the Department of pathological uh anatomy. Uh And uh he has, he has focused on the gynecology oncology, uh received numerous scholarships uh worked in uh Auckland, New Zealand. And uh currently is the head of the Clinic for Colposcopy and Vulvar Disease. And the consultant for pelvic surgery, being involved in the development of different surgical techniques since 2001. He has been an investigator for the HPV uh vaccine uh trials and his past president of the European College for the study of vulvovaginal disease. Uh So, uh the the topics uh are very interesting and uh are related to uh cervical uh precancer. Uh Professor Feldmann will talk about uh the C two surveillance to treat or not to treat. Uh Professor will talk about the HPV, vaccination for women scheduled for cervical colonization and uh Doctor Vinson will talk about the Artificial Intelligence applications for cervical screening and the diagnosis. So we can uh move on to the first uh speaker who is a professor uh Sarah Feldman, uh Doctor Feldman. Uh you may, you may start sharing your screen. Uh The I PBS. The International PPI Virus Society is very proud to host you. And uh let's hope that you will have a, a great uh webinar and uh a motive for young trainees who have, who are watching us to become members of uh this uh very, very wonderful uh society, the HPV society. Thank you. And uh you may start. Thank you so much. Thank you so much to IV I PVC and to NT for inviting us. Um It's exciting to speak to you today. Oh, my slides are not moving. Um I am a coinvestigator on several um grants to um funded by the NCI in the United States to screen and prevent cervical cancer. I do also want to uh to acknowledge today that this is a complicated topic when you're trying to approach it globally. Because not every place in the, in the, in the world is going to be able to give pathology, telling about CCI N two versus C in three and not every place in the world is going to be able to treat with lobe or leep, which is what we're gonna focus on in this conversation. So I apologize in advance that this is going to apply to those environments where, where you have the pa uh capacity uh with pathology to distinguish C in two versus C in three and also the ability to treat um with an excisional procedure. So, before um I talk about that, I wanted to start by uh just um talking about the history of the nomenclature. So years ago, cervical clear uh uh dysplasia was thought to be um basically you moved from mild to moderate to severe with the progression from one level to the next. And um initially, all dysplasia, including all the low grade abnormalities were treated very aggressively, uh uh way back. Um you know, 30 or 40 years ago with colon or hysterectomy. But then over time, um it was understood that low grade abnormalities were found to resolve the majority of time and to rarely evolve into cancer. And so there was a feeling that for low grade abnormalities or c in one, we would just observe most of them and that they really weren't precancerous lesions. Um And uh c in two slash three were both thought to be precancerous and were combined eventually into one category under the last nomenclature called HSIL. And the idea was that um those would be treated similarly. So there was just a dichotomy. The mild abnormalities were now thought to be something we follow and the moderate to severe were thought to be something that we treat. Um And uh just to um, remind you again that, um, these abnormalities, uh, it is, it sort of has a variety of, of information in the, when you get a biopsy and in, um, the normal ones, um, and the low grade ones, there's still a majority of normal cells and in the moderate to severe ones, there's, uh, a differing amount of a normal cells and abnormal cells. But it was felt that if you treat all of the moderate or severe, all of the um the uh pa pathology diagnoses that encompass some of these um uh abnormal cells, then you would, you know, prevent cervical cancer. And that, so that was the thought and the question really is, should c in two and C in three be treated the same. And that depends on a lot of factors. First, are they, do they really act differently? Are they equally precancerous? But also, is there a con to treatment, is there some reason we wouldn't want to treat every single person? And so that's what we're gonna sort of talk about today. So the first thing is um looking at, you know, are they really uh you know, likely to progress, progress to cancer and are they similar in their likelihood of progressing to cancer? So, this was a um a meta analysis that looked at 36 studies with about 3000 women and it looked at what was the, what were the rates of aggression, persistence and progression? Um In these studies and they found that, you know, about 50% of patients with c in two documented lesion regressed, about 32% persisted and about 18% progressed on to c in three or cancer. And in the studies in aggregate, the rate of noncompliance with followup and prospective studies was about 10%. Now, I tried to look at all these studies and I can tell you that I'm not even sure uh exactly how, what the compliance or noncompliance means, meaning I'm not sure what they considered appropriate follow up. Uh They obviously wasn't the same in all the studies. And so it's a little bit hard to interpret. Well, what did they do to determine if the patients progressed and if they regressed? And so here is a study. Um Oh, I, I'm sorry, I should just say before we get to that, I just wanted to emphasize it in the women younger than 30. In this study, the rate of regression was even higher at 60% and this meta analysis was even um higher at 60% persistence, 23% and progression was lower at 11%. So that was sort of reassuring. But as I said, um didn't give us that much information about exactly how our patients followed, how long they followed for, you know, what ultimately was a standard. So, um this study is looking um among uh Danish women using registry data and it looked before and after there was a new guideline for management of women with c in two among women ages 18 to 44. And what they found is that um the, the guideline was implemented in 2 2012 and it continued to recommend treatment for women who no longer desire fertility. So I think for everybody who is see to and fertility is not an issue at all. I think pretty much every guideline is saying, treat those patients that would include postmenopausal women and women who actively tell you I don't care or I don't wanna have more Children or whatever. But for women who state that they wanna have um more pregnancies, they offer the option of conservative management for the ci and two patients with repeat biopsy at six months and every six months for up to two years. And so looking um from before they made that guideline, which allowed for that close observation to after they made that guideline, which um then, you know, these patients, some of them could be surveyed. They found that um you know, they went from about 30% to 50% of the patients actually got surveillance. So that allowed more patients to be followed. And in this, in the post guideline period, they found that um they had aggression by 10 months. So they only evaluated essentially once in this study at 10 months that almost half of them had already regressed by 10 months. Um 35.5% had persisted and about 17% had um progressed. So the second question is, suppose that we decide we wanna follow the C in twos. Um How confident are we that we actually have ACI N two? Now, this is important because C in two is the least reproduce, reproducible of all the diagnoses. Uh Meaning compared with C in one, which is pretty clear to more pathologists I should say, I mean, or yeah, pathologist than um C in two and the same with C in three of note, um C in one versus normal is often confused as well. But we're, but the management distinction comes at that c in two level. So that's what we're focusing on. Um And so, um this is a study that um looked at the reproducibility of C in um diagnosis among different pathologists. Now, I wanna say there's been many, many, many studies about this as there have been about every topic I'm discussing today. So, II also wanna indicate that I'm just putting representative studies up here, but they are representative studies that I think show a lot of information. So this was um this study and the new technologies for cervical cancer studies. And they are reviewed of all cases of C in two and they had and um every C in diagnosis blindly reviewed by two independent pathologists. So there were nine centers for this study, each in each of those centers, they had pathologists who made the initial diagnosis and then they had panels. So two additional uh pathologists for each um, case that confirmed the diagnosis and they found that c in two had the lowest class specific agreement with fewer than 50% of the cases confirmed by the panel member. Which means if you get ACI N two diagnosis, you're not sure, it's not ACI N three and you're not sure it's not ACI N one. And so if treatment is based on that diagnosis, you could be making a mistake. And um I looked them to see well, is this the, you know, this is similar in other studies and of course, this has uh been a controversy that's been going on for many, many years. So there are many, many studies that look at this and the studies uh have widely differentiate uh um confirmation rates of C in two, but basically many of them have even higher rates of discordance um for a diagnosis of C in two. So based on that, um um in um 2012, um the uh uh the last nomenclature was published and the last nomenclature is basically nomenclature which is um combines high grade dysplasia. So, um ma moderate and severe dysplasia into one category which is called H cell. And in that um nomenclature, the last nomenclature, um the, the part of the argument for it was that they really felt that people um couldn't really distinguish well between CN two and C in three reliably. And that the treatment decision started at the C in two. And so everybody should be treated the same way. They did talk about using H PVI mean not sorry. H PVP 16 positivity in lesions where the pathologist thought that the um appearance of the cells was consistent with a high grade abnormality. Um They could use P 16 to confirm that um there was HPV oncogene overexpression and transformation. And so they in, in the last guidelines, um they emphasized that a positive P 16 immunostain supports the diagnosis of an HC meaning ACI N two slash C in three. If the morphologic assessment, in other words, if the pathologist also thinks that it looks like Aci N two. So in that particular, in last nomenclature, they did not feel you could use P 16 to diagnose C in two. But that you could, if the slides itself looked like C in two or three, you could use the P 16 for confirm confirmation. In other words to say it wasn't P uh C in one. So now um we're gonna talk about treatment um in many countries. Um A leep is the standard for treatment and that's what we're gonna focus on in, in sort of the conversation today because many of the pros and cons of treatment are more salient for treatment with leep as opposed to um cryotherapy or ablation. So, uh that is sort of what I'm focusing on today. Um That does not mean that you cannot use um cryo um therapy in appropriate cases as needed. But we'll focus on the leep today. The pros of the leep um are that it's more effective at um clearing the C in two. It gives diagnostic information and it can be adjusted for size. But the cons of a leep are that um it can result in bleeding, infection, cervical stenosis and the biggest problem that people are cons that most um clinicians are worried about in patients is possible adverse pregnancy. Oops. So that's what we're gonna talk a little bit about. Um Before we get to the pros and cons of treatment or not, I tried to find guidelines, international guidelines for management of C in two. I know that in Denmark they have them and I was able to find some for several other countries. Wh O to my knowledge does not yet have a guideline for distinguishing C in two. So I'm putting up here the A S CCP, the American Society of Colposcopy and Cervical Pathology Guidelines. Um that was a um is basically management consensus guidelines that we use in the United States, I think um is the basis for a lot of other um uh environments as well. These are evidence based guidelines and in these guidelines, they do offer the op option of patients with a diagnosis of a histologic C in two who um is worried about pregnancy outweigh their concerns about cancer, then either observation or treatment is acceptable, provided you have an adequate exam, you're able to visualize the way of colum junction and there's nothing in the endocervix uh in the end the cervical canal and that there's no concern for a higher grade lesion lesion. So that's important if um the histologic A so cannot be specified as C in two regardless of fertility preferences. Um Treatment is preferred, but your it's observation is acceptable. Now, I actually mentioned this is actually a very important um caveat in my hospital, which is a large um hospital that uh uh is focused on women's health actually. And so it's a whole gynecologic pathology division. Um They were uh coauthors of the last guidelines. And so they strongly believe that you can't tell ACI N two. So we do not get information on C in two. So in my environment, I am only gonna be told it's an H cell, that's the last nomenclature. And so I had emphasized at the beginning, you have to be able to have a pathologist willing to tell you c in two versus not in our environment. We, we don't know about that. So we are gonna get an HC and we're gonna have to make a clinical decision to the best of our knowledge they do in the um A S CCP guidelines emphasize that um for women ages um 25 and older, the observation is basically a colposcopy and a HB HPV TE uh based test at six month intervals for up to two years in patients saying that are younger than 25. Uh we see two observations preferred. Um but treatments acceptable. Let's talk about that for a minute, that observation. So here I have a patient in my office and I tell her you could have a loop today or, you know, whatever I soon and then we'll happily just follow you with an HPV based test, which is relatively noninvasive in six months and a year, or you could come back every six months for a biopsy and colposcopy for the next two years. It's not a completely straightforward decision even in people who desire fertility because there's a lot of burden on the patient in that option. I just wanna emphasize that cause sometimes guidelines don't mention how what the patient burden is in, in an option like that. Um And in this guideline, if you find that the um C in two is, you know, the same um and or better, basically after two successive um occasions, um you're able to maybe go to once a year, but basically, um it does require quite a few more colposcopies. And then if you have C in two for a two-year period, that might argue that this is one of those um lesions that isn't going away and that's when you treat. Um this is just a visual format of the same guidelines. Um So let's talk a little bit about the pros and cons. So the pros of treatment um with leep are that there's a decreased risk of progression, you treat them, then you um there's a decrease um need for f uh frequent surveillance, um which has a significant burden to patients, but also increases your risk for loss to follow up. And it decreases uncertainty both for the provider and the patient, the patient is taken care of. We know what they have. We're not worried about them. The cons um of treatment with loop of course, are the concern about increased risk of adverse pregnancy outcomes and we'll talk a little bit more about that. Um And uh this risk of bleeding and infection and the in contrast, the pros of surveillance um would be fewer loop procedures. Um except of course, you still need a lot of aggressive surveillance. Um but potentially lower risk of adverse pregnancy outcomes. The kinds of surrounds again, are not small patient anxiety, increased cost and disruption to patients requiring every six-month biopsies and potential increased risk of progression um to cancer, risk of loss to followup and increased rate of recurrent dysplasia. So, you know, there real are real pros and cons to either option. So how do you decide, how do you decide? So I sort of list up here things that have been proven in the literature but also how I think about it. So one is the age of the patient. What HPV genotype did they have, if you know it, if you, um, if they are 16 or 18 positive, um, or HPV positive at all, that's important. Have they had HPV for several years as documented by prior testing? Have they had high grade dysplasia in the past either C in 23 or another, you know, or a AIS or some other serious problem. Does the P 16 staining? Um, is it, is it aggressive and strong, consistent with um active viral um activity? Are they immunosuppressed? Um for example HIV, positive women, according to the WH O website have a six times six time risk of cervical cancer relative to HIV negative women in our practice. We see a lot of people um rheuma with rheumatologic disease and, and organ transplants. These are also people at very high risk. So, of course, um we're gonna have different concerns for them. And then an important thing is, um you know, informed decision making with the patients. What is the patient's um pre uh preference regarding future pregnancy risk coming back for colposcopy and ability to comply with frequent surveillance visits? Um I have my residents ask every patient um Several questions on the first is we make no assumptions of either about young women or older women about whether they wanna have more Children. Um And we ask them, do you, are you planning to have more Children? We are sometimes surprised by 48 year olds who are planning to have more Children and by, um, 24 year olds who insist they will never want a child ever. So we, we try not to make that assumption and to ask them, um, we ask them how easy it would be for them to come back for colposcopy. Um, and to comply with coming back, we uh ask them how far they live from the clinic and other financial barriers or whatever to coming. We also will at that point, um involve a patient Navigator who's like a community health worker to help us see if there's ways we can help them if surveillance would be their preference, but that it might be difficult for them. So let's ask the first question. Does age affect the cervical and the can risk of um, cervical cancer? Well, we all know that it does, but, you know, it's important and this is from the wh O website to just remind yourselves that the highest rate of precancer, which is the purple line is gonna be in that 25 to 30 year old age group, which is also a group that is, you know, highly likely to be thinking about pregnancy or being pregnant or et cetera. The highest rates of cancer are gonna be up uh lot rising in the late thirties to forties. So being under 30 means that you have a reasonable chance of having a high grade dysplasia, but not that high a chance I'm not saying it's zero but not as high a chance of actually having cancer. So 30 is a reasonable um point to start. Um thinking I'm um in the guidelines that we have in the United States, they use 25 as an absolute cutoff of where we make a difference between recommending observation versus surveillance. But even if you look at this and you look at the rates of cancer, et cetera, you can see that um 30 is pretty safe for rates of cancer. Um And um but definitely by the end of the thirties, it becomes less safe. Um And what about genotype? So this is a slide that looks at um a patient with either a normal or ascus pap at time zero and, and looks at their risk of C in three or greater based on their baseline HPV type. So in the dark pink, we have um HPV 16 in the orange, we have HPV 18, the blue is all non those non 1618 types and the white would be, you know, you don't have HPV at all. And one thing that's really important and the uh uh across the um uh horizontal is the month in TI time in months and look at the rate of increased risk of C in three if you are HPV 16 positive at baseline. So you can see that by 4.5 months already, the rates are going up. That's not as true yet. For HPV 18, which takes a bit longer for the rates to go up, the rates do go up with HPV 18 eventually, but not quite as, as early and as aggressively so being, you know, HPV 16 positive, as we know in our American guidelines becomes right away, a marker for a, a more aggressive um infection and a more aggressive course. Um And then of course, if the patient has, let's say C in two and for some reason, the HPV negative, um this do, of course, this chart doesn't combine, there's C in two and HPV, but the point is HPV, negativity is somewhat reassuring, quite a bit reassuring. So, on the other hand, how bad is it to have a loop? Like we have some information about what are risk factors for progression. But let's also think when we're counseling a patient, you know, isn't really a terrible thing to have a loop. So the first thing I wanna say is not all loops are the same and this is uh before I even get into the analyses. Um I've had the good fortune of being able to work in different parts of the world and set up um clinics there. And what I have learned is that the loop sizes are quite different in different parts of the world. So the, the loop, the wires that we use for loops in our clinic in the United States are quite small and shallow compared to what we are seen in some other places in the world. I think actually most other places in the world. So when we look at these analysis, um the depth of excised tissue or conversely how much residual cervix remains is really important when you look at the outcomes. So multiple meta analysis have suggested that excisional procedures are are associated with an increased risk of adverse pregnancy outcome, that includes preterm, rupture, the membrane preterm birth as well as um second trimester abortion. However, uh more recently and this is a really great article, if you wanna look at a summary of this, um it turns out that the comparison group in many of these studies might not have been the right comparison group. So, in general, they were comparing patients who had a loop versus those who had no cervical dysplasia and didn't have a loop. However, it turns out that in studies that compare people who have dysplasia who are surveyed versus those who have dysplasia treated, um It the the difference is not that great. It turns out that um the actual fact that you have cervical dysplasia may also put you at risk for preterm delivery and then some of the and preterm rupture of the memories and some of the adverse outcomes as well as some of the co factors about those patients. You know, whether they have multiple sexual partners, other sexually transmitted diseases. Do they smoke et cetera? There may be other factors So when the comparison group is other people with dysplasia with patients, um who either and so between patients who did or did not have loop, there was virtually no just difference in the obstetrical out um outcome. In addition, more recent studies that have looked at the depths of excision have found that the excision less than 10 millimeters um show no increased risk. So again, when we're thinking about um treatment with a leper loop, not all leper loops are the same. What I do counsel younger patients when we're doing a loop is we can do a smaller loop canal. Um And see, you know, if that solves the problem by checking you in six months with an HPV based test. Um And in some percentage of the cases, we'll have to do a second loop and that's usually about 10%. But that means 90% of people actually get away with their C in two s with um a fairly small shallow leap which may, we're not sure how this works, by the way, we usually have excised lesion, but also it may increase the local immune system at that point to clear the HPV. We're not really sure totally how this all works. But the bottom line is those patients do well as well, but they also in the studies have less risk of an adverse obstetrical outcome. So in summary, C in two is a mixed diagnosis with a lot of interobserver variability about 50% of patients in these in the studies where they're able to tell us about C in um C in 2 may resolve between one or two years. However, patients who have a history of high grade dysplasia HPV, 16 or 18 immunosuppression or persistent HPV are less likely to reduce the um CI two. Um the standard of care for women in the United States with the C in two who do not desire fertility is treatment. When I looked on the wh O website, um it states that the standard of care using the last guidelines. L ast as I said before would be treatment. So I do not yet see in the wh O guidelines, a distinction between how we would treat CIN two and, and three. So, but in the United States uh um and it sounds like in Denmark and several of the other countries that we looked at um the treatment is still for anyone who uh does not care about fertility as treatment, but it is acceptable to um watch patients who are concerned about fertility and who don't have other concerning ri um risk factors. Um A note um it would be OK. Uh As long as you meet the cryotherapy uh guideline requirements to treat these patients with cryo as well. Um And I would emphasize that in reproductive age women who opt for treatment, smaller loop procedures should be considered. Um And now um we're gonna play a game. So, hold on one second. We're gonna do um, some cases and see how, um, people feel. Um So the first, the first case is 44 year old with no partner has C in two, she's had HPV 16 for three years prior history of C in two slash three. Who says I want to wait for treatment until I find a partner to have Children. She has previously refused treatment for C in three with another provider. What would you recommend? And I'm just waiting for response and I will tell you this is a real patient. Actually, I want everyone to know this is a real patient. All right, we have now about 100 and 50 responses. So I am going to end the poll and share the results and you can see that 60% of people said immediate leep, 28% said surveillance and 12% said it's up to her. Well, we'll start by saying it's always up to the patient. So that answer is always correct. It's up to her. But what would I recommend? This is 100% a patient who needs to have immediate excisional treatment. She needs to have an immediate leap in this particular patient. Um She ended up with stage four cervical cancer. So she has every single risk factor. She has HPV 16, at least documented for three years. She could have had it before we know she had high grade dysplasia in the past not treated. Um And she didn't have treatment. Um And she's waiting indefinitely to find a partner at an age when her fertility is also not as high. This is a woman who is going to get cervical cancer. She is not a person for um surveillance. Now, of course, it's up to her. I just wanna make that clear. But in my uh setting I explained to this woman, you have basically, you have are going to get cervical cancer. And unfortunately, she had stage four, which was preventable if she had been treated the first time, she had high grade dysplasia. And she still, at that point, if she had been treated, then would have been able to have Children because she would have kept her uterus. So that's also the part that's important. Um, the next case and I'm just gonna say that we're beyond time. So I'm going to do maybe one more case and then chemo and you'll tell me if that's enough. So the second patient is, um, here we go. 23 year old has had one child. I want to be treated for my ci two. She's been HPV, vaccinated. First abnormal pap showed an ach and a biopsy showing C in two. What would you recommend? All right, in the interest of time. I'm gonna end the poll now and share the results. And um in this case, 18% said immediate leap, 78% said surveillance and 4% said it's up to her again. It's always up to the patient. So that answer is always correct. But in this case, I would very certainly recommend surveillance myself. This is a very young woman, 23 years old. Um, how many Children she already has is not really relevant because sh she can certainly want to have more Children. Um, I understand she's anxious about being treated. Um, I mean, about having the C in two, but the truth is, um if 50% of the time or even 60% of the time, this is gonna go away on its own, then she'd be better not being treated. Um She's been HPV vaccinated. She doesn't have 16 or 18. I mean, we haven't tested her for it cause she's only 23 but she doesn't have 16 or 18. She was vaccinated on time and appropriately and she's had one abnormal PAP. So from my perspective, this is a person that I would strongly encourage to consider surveillance. Um And now I see chemos on the screen so I'm gonna stop sharing and move forward. Thank you so much to everyone and I appreciate your time. Thank you. Uh Sarah, I personally, I was very happy to be your student for, for just a half an hour. And I think that uh all of our vibrant audience uh uh thinks the same feels the same and the, the poll uh was very uh interesting too. Uh we don't have much time to uh to address all the questions asked by our uh uh colleagues here in the Q and A uh box. However, I would like to uh summarize a bit of these questions. And uh uh if you uh if you could tell us your opinion about the application of uh vaginal uh creams that have been uh circulating in the market uh on the uh surveillance of uh maybe CN two or CN three lesions. There is a question, a question about it. OK. So, uh again, I'm gonna summarize in a couple of things. So the first thing is there are many different um I uh vaginal gels that are out there as well as vaginal creams. And I have to distinguish that they have different ways of working. So there are um vaginal gels that are going to affect the vaginal microbiome and help clear the HPV by normalizing the vaginal microbiome. There are other gels that are gonna work by um stabilizing the membranes. There are other um creams or um treatments such as Efudex and Aldara that are gonna work in different ways in increasing uh like the uh like immuno immuno uh immune modulation, I'm not good in the vocabulary um or in this fedex at um slumping off abnormal cells. So, basically, um none of these are approved in a standard way. And my understanding of specifically the vaginal gels that have been studied um related to C in one mostly and HPV clearance is, they're mostly looking at and are effective for early infections and HPV clearance and CI um C in one, they're not as far as I know, been proven or shown to uh reliably work for higher grade lesions and certainly are not the standard of care. So there is no current guideline that I'm aware of that would indicate that you should use one of these creams as opposed to following a standard manage management, either treatment with Aleve for patients 25 or over who are not designed fertility. Um That being said uh uh oh, I should actually say that one other thing, the patients don't love these creams. I have to say. So one and I have to say a disclosure that I currently have a grant looking at patient acceptability of these creams. So we're looking at patients about whether they wanna use them, how they feel about them. And we're looking at that among people who are HPV positive and um of uh the uh diverse patient population who of poor women. So that's who we're looking at women who um we wanna see, would they be willing to use this? And, and the results are not out yet about that. But my feeling is there's also discomfort with some of these creams. They involve a lot of work for the patient. They haven't necessarily been proven, but they probably don't have much harm if the patient wants to use them. So I think it's reasonable to offer them if the patient is interested. That's fine. Um But I wouldn't be saying to them this is the standard of care and I wouldn't be uniformly offering them to patients. Thank you. Thank you, Professor Feldman. And now we can move on to professor uh Yaar. Uh we'll talk ab we'll talk about um the uh HPV uh vaccination around uh uh around the time of uh a leep procedure. So, Elmar uh thank you, Hemon. Uh Thank you for this uh invitation to be part of this great faculty. Uh The topic I'm talking about is still a bit controversial, although we have quite a lot of evidence uh by now. So we all agree and we have uh excellent data that early vaccination is best. And uh when you are vaccinated before the age of 17, uh most of the cases of invasive cancer can be prevented and it's less effective when you're vaccinated later. However, we know that uh women with a history of um cervical high grade dysplasia and undergoing colonization, they have a lifelong risk of or developing invasive cervical cancer even when they are under surveillance. And we also know that uh these women and these data from uh the Netherlands are at a lifelong uh quite uh tremendously increased risk for other HPV related cancers of the vulva, uh the vagina, the and the oropharynx and also for the precursor lesions So these women are really at risk. And so it's important to look into the question if they might benefit from the vaccination, what is also important to know if you're vaccinated with one type and most, uh, women are just, uh, infected, uh, with one type, then you immediately get the same level of protection for the other type. So it has no therapeutic effect. But when you have infection with one type and you are vaccinated against nine types, you have the effect on the eight remaining types. And when you're seropositive, so you don't have an active infection, but your immune system has uh already uh signs of antibodies that you have had an infection. You also get the same efficacy as if you were HPV cleared. An infection is uh puts you at zero. We also have data on efficacy up to women up to the age of 45. And with the nine valent vaccine, we've also done a study of immunogenicity uh with this vaccine up to 45. So the vaccine is still working and highly immunogenic. It's now 10 years ago that we first brought up the question, looking into the phase three trials of the quadrivalent vaccine and looking what happened to the women once they have reached an endpoint because in the primary analysis they have been censored. And so we looked at them after the treatment. So the study protocol included also HPV, positive women so that uh we're not allowed to have a history of HPV related disease. But when they were positive, they were included and more than 1000 needed a treatment soon after the randomization. And to our, at that time, big surprise, we saw that there was a significant reduction of subsequent disease at the same site. So after conversation at the cervix, but also at the vulva and also the other way around. So when you had HPV related disease of the vulva and you got vaccinated in the future, you also had less disease at the cervix. So these are the numbers of women uh having been treated for uh CRN uh vulva vaginal disease, they had subsequently uh a 65% reduction of CN two plus and even more than 70% of CN three plus. And the overall reduction of any HPV related disease was 46% just looking at the four vaccine types of the quadrivalent vaccine. Uh The level of reduction was in 80%. Meanwhile, uh this uh topic, um the little of interest and many papers have been published and a great me on published um some time ago and you see all these studies are dealing with uh the question, do women benefit? Uh after conversation. Meanwhile, there in more studies and they are providing the same result. So in total, our results have been confirmed, there's about a 60% reduction of recurrent high grade disease of the cervix. Uh in this meta analysis. It was also broken down by the age under the age of 26 and over the age of 26. And the results were pretty much the same. And also if the women were in these studies vaccinated before colonization or after colonization, the results uh held up and there was uh more than 50% reduction of subsequent high grade disease. Another meta analysis um which uh has some overlap but included also newer studies uh was published uh last year. Uh by these Italian team, they also check for the quality of the various studies and most of the studies have at least a decent quality. And uh they found very similar results. 65% reduction for any CN two plus recurrence of colonization just looking at HPV 16 and 18, also 65% reduction. Uh They also analyzed for prospective and retrospective studies. And again, uh the reduction was very similar uh related to the study design. Um uh 64% and 68% reduction. And this uh is quite reassuring that uh the data to answer these questions are rather robust. Uh a population based study uh from Denmark with the excellent registries told us that the vaccination might be time sensitive. So they compared women vaccinated before or soon after the colonization, they saw a disease reduction in those being vaccinated after the conversation up to one year in these population based data, they saw no effect. So probably the time of vaccination might met up. And meanwhile, we also did an analysis with the nine valent vaccine. And we saw the same result for the four types of the quadrivalent vaccine. And looking at disease caused by the five additional types, uh there was an 86% reduction of subsequent C A NBI N. Uh but trying intra and condyloma, the numbers were too small to be at the numbers. Um The data are quite consistent over the various groups uh to round up the picture. Uh There are also data on vulva, intraepithelial neoplasia. Grade three, a study uh run in Italy and they found a reduced statistically significant reduction of recurrent disease of uh 32 versus 19%. So less reduction. Then after conversation but still clinically meaningful and also in males. Uh a study from the group of Steven Goldstone from New York men with high grade anal dysplasia when they were vaccinated the vaccine. They also had a significant reduction of subsequent disease. Some countries like Austria have started to fund uh these uh vaccinations around the colation. In Spain, it's funded in some regions Madrid for instance. And in Germany, uh you have to apply at the insurance company, but usually you get the funding for the vaccination before or after colonization since it's time sensitive. Uh We recommend the patients to start at the time of diagnosis and to get the second shot uh at the time of treatment. And the third shot they can get when they show up for follow up six months after the conversation uh for the HPV test. This is a perfect time to receive the third uh vaccine. So what is missing up to now? We have now plenty of data from uh post hoc analysis from the large randomized controlled trials for the quadrivalent bivalent. And meanwhile, nonvalent vaccine, we have retrospective data prospective but non randomized data. And we have these population based data from Denmark. But what's sadly missing uh to give the final answer is there is no randomized controlled trial away level. But uh two of these trials are now on their way in the United Kingdom. The novel trial run by Maria is uh addressing these questions and there's another ongoing trial in the Netherlands. So I hope uh within the next two or three years, we get the final answer and I'm happy to answer any questions. Uh And thank you very much again for this current. Thank you. Thank you for a wonderful talk. Uh We have a few uh a few questions. Uh Let me just uh uh briefly, there is a long uh question here in between the uh box. Uh But I will summarize it. Uh They are saying about uh the what, what happens in the post vaccination world and what would be the utility of uh HPV 1618 triage uh in the post uh vaccination era since uh these types might be replaced by non HPV, 1618 types. Could you comment on that on the probable alleged replacement of HPV 1618 as causes of cervical cancer? Um ok. I think that's an important question and since uh we started uh using these vaccines now, um 16 years ago, uh this question has always been looked at and so far we don't find any uh signal for a replacement. Of course, when you have a reduction of 1618, which is the majority of the lesions, uh the proportion of the other um types is increasing. But at the end of the day, you always have to prevent infections with those types causing cancer. And in most of the cases, it's 1618 plus 3345 maybe 31. And in some regions of the world 5258. And once we omit these cases, uh we really move towards the elimination, which certainly will take some time and we need globally good coverage of these vaccines. OK. And uh we have another question uh which uh uh came in uh prior to this meeting. And uh they're asking uh uh about uh why do vaccinated women still get HPV infections showing up on a pap smear. Uh Good question. Uh I often get this answer. So the vaccine basically prevents uh infections with the types which are included in the vaccine. So when you have um the bivalent vaccine, you get a robust uh protection against HPV 16 and 18 and maybe some uh crossprotection. But that's waning over the time against a few other types. Um when you are vaccinated with the nine valent vaccine, all the remaining types cause less than 1% of cancers. So a woman which is has been vaccinated early with the nine valent vaccine, she still can get infection with insignificant types. But the likelihood that this woman gets cancer is quite low. So you have to know what vaccine did someone receive? At what time was it early in childhood or when this woman was adult? And then uh doing the HPV typing, you'll get the answer, but it's not a complete uh protection against any HPV infection, but it's definitely an excellent protection against the development of cancer. OK. And uh can you give us an interpretation? I mean, uh why uh do we uh do we see that uh uh recurrence rates are lower in women uh vaccinated around the, the le why, why does this happen? Um So first, uh what's logic? Uh the vaccine prevents new infections? So uh new infections are blocked, that's one part of the effect. But when we looked in the studies into the details, the effect uh seems to be bigger than that and it might have some uh influence on the reactivation of latent infections. So we cannot prove it at the moment. But uh the clinical data show that the effect is wider than just preventing new infections. OK. Thank you very much so much for your uh contribution for an excellent talk as always. Uh And now we can move on to uh Doctor Nicholas Vinsen, uh who will uh talk about uh the uh Artificial Intelligence in uh cervical cancer screening and pathology. Doctor Vinson, you can start. Thank you very much um on uh Thank you very much for the invitation uh to this uh really exciting webinar. Uh Can you see my slides? Yes. OK. Great, great. So, um the II think this is a, a great selection of, of presentations. We, we're all addressing a similar topic but from very different angles. And I'm gonna sh show you some of our data on artificial intelligence applications for cervical screening and diagnosis. And I'm gonna talk about some things that are current and some things that may happen in the future. I have no conflict of interest. Um So I'm I, I'm talking about um artificial intelligence applications that we are using on uh medical images, particularly on pathology and cytology images. And uh there is um a great uh enthusiasm and a great hype uh in this area at the moment. Um There's really like any think about um application um that is being addressed at the moment. So there's, there's a lot going on. There are very few applications that have made it in the clinic. But at the same time, this is a topic that has been around for quite some time. So the terms of artificial intelligence and machine learning were actually coined in the 19 fifties. And then um deep learning came in the 19 eighties, convolutional neural networks later in the eighties. And and then even in 1990 whole slide scanners were introduced. So but but a lot of these technologies were not accessible, not affordable to a wider audience. And just um in the last 10 years, these have become um widely available. And now there are a lot of great opportunities to evaluate artificial uh intelligence approaches to, to pathology and cytology. So um the first FDA approved um um device uh came out in 2018 and there's a lot more in the pipeline now. So we will see a new applications coming down the pipe. But there's um there's a lot of pipe, there's a lot going on, but there, there, there are a lot of lessons that we're learning and I'm gonna show you some examples, but I'm also gonna talk more generally about some of these lessons and what they mean for future applications. So what is the promise of artificial intelligence? And, and I think the um of very obvious thing is that we, if we develop like a computer algorithm to generate the diagnosis or to classify uh a pathology slide, we we will reduce subjectivity. So that that's kind of like a a given. Uh the, the hope is that we increase throughput because um the automated evaluation can be done faster or it can be batched, it can be run continuously. So we may be able to do this faster than um a pathologist can do that. Um The hope is that we can improve accuracy and, and that's kind of a, that's, that's a really important area where the way to how we approach training and validation matters a lot because uh currently many artificial intelligence approaches have been developed to match the human evaluation. But if the human evaluation is flawed, then we will not be able to surpass that um uh that accuracy that, that we have from, from a um AAA general evaluation. So if we want to move beyond that, we have to use other endpoints, we have to use ground truth endpoints. And I'm gonna show you how we've done that in some of our studies. And another hope uh uh of really um using these new technologies is to improve accessibility of high quality uh screening triage and uh diagnostic um um performance. Um So, which currently is more restricted, requires training infrastructure and many other things. And the idea is that with these approaches, it is possible to basically collect a specimen um at a certain point and then have a central laboratory that that can process uh these specimens can generate the slides, can do whole slide scanning and then apply these algorithms and uh distribute them to more remote locations. This is something that is only starting and there is kind of a lot in the development and we haven't seen like um many examples of that but, but it's certainly a great hope that this will benefit um uh uh many users beyond like the high tech centers that are currently developing these applications. So here's some really important um uh general principles for the development and validation of A I imaging algorithms. And one of the most important uh points is really that we need large and well annotated data sets uh both for training and then for validation. And it's very critical that the ground truth for these training data sets and also for the validation, uh training sets is uh validation sets is, is established and, and that we really have excellent annotations that requires, can require a lot of work uh depending on what we are dealing with. And, and so this is really um the, the input really in the end determines what we can achieve. And it's, it's absolutely critical to conduct independent validation um in really probably multiple steps along the way um to avoid overfitting, which I will be addressing uh in more detail in a little bit. The other important principle when we use artificial intelligence and when we generate um evaluations of these pathology measures is that biologic data underlying these evaluations and the outputs that we're getting. So we usually get some sort of score. Um they're typically continuous and uh so at, at the bottom. This is a very simplistic um graph showing the probability of having no disease at zero and definitely having disease at 100. And then really the whole range of these scores where we're moving from one 0 to 100 at the extremes at the bottom and at the top end, it's very clear. The classification is very, very obvious. This is around zero, pretty much every case that we have with that score is not gonna have to see it around 100. Um That that's gonna be pretty much all of them have to see. But then there's this gray zone in the middle. And if we use a single threshold and if we say, OK, anything above 50 we call disease, anything below 50 we call normal. There will be misclassification, there will be some cases that we miss and there will be some non cases that we call positive. So there, there is no uh simple solution to that. Um because that gray zone is, is uh gonna be there in, in all settings. The question is, what is the dis the underlying distribution of these scores? If most of the are uh at the extremes and we have like few in the center, it's less of a problem, but that's obviously typically not what we see. So one way to deal with that is to have more than one threshold and to really say anything above threshold number two, we call disease and there will be very few false positives here and anything below the threshold, one we call normal and there will be few false um negatives here. But then we have to deal with this gray zone and do something else with that. But, and, and that, that's kind of one principle that uh we're using in some of our approaches. So here are some really critical terms when we uh develop um new algorithms. And one and, and I think there is often a lot of excitement based on the very first um uh developed algorithms, we see amazing performance to classify outcomes with very high A ucs and very high um scores on all levels. But the question is really, does this hold up? Does it is, is this a repeatable result? And that's kind of that's the first step that we need to show and that that is something that is not widely um done at the moment. So the, the, the question for about rep repeatability is does the algorithm as it is developed provide the same result when it's run on the same sample or image again? And surprisingly that that can be challenging for some algorithms and small changes of of the image. If we're doing the same scan again, if we're taking another image from the exact same um object um can introduce small changes and can really derail the algorithm. So that is one important part, then the next question is, is it portable? Can we use this algorithm uh in a different series in a different um set of, of um cases and controls that are on different slides that were scanned on a different platform that were taken with a different camera. So, so some technical variation but really not much else that is different and it should still perform the same way. But we often see and, and this is something that is done more commonly. And in many cases, we see that uh amazing algorithms that, that have shown amazing classification in the first step can break down quite a bit. And we need to do additional training to account for this uh heterogeneity. And a term that that is often used as overfitting, um overfitting uh refers to an algorithm that fits exactly the training data but has much worse performance in external data, which means that there, there are certain features of the training set that are just perfectly predicted, but um it doesn't work in an independent set and we have to deal with all all of these um aspects. And at the bottom, there's a little um graph just showing how this works in, in reality. So this is the, the first um box here shows an H and E slide. And then what is shown is that the, these slides are really cut up into very, very small um pieces and very small um uh uh like boxes with very few pixels in two dimensions. And then these very small boxes are of uh really run through the convoluted neural networks and that are used to classify what this is and, and but, but, but it's really, it's pretty uh amazing that all of these algorithms are based on, on very tiny pieces from that um whole slide. So the whole morphologic um appearance that we use in in pathology assessment is basically broken down and uh it's used in a very different way. So, um in the world of cervical cancer screening, there have been a couple of new A I applications and we've been, our group has been developing several of those. Um And they include dual stain cytology. So that's a cytology application using a biomarker and then quantifying uh the presence of these biomarker stained cells. P 16 histology. That's a very new development. We just started on that and I'm gonna show you some very early results. And then, so basically, we're moving from the cell to the um to the whole section and to the whole lesion and then here to the whole cervix. And that's the area of automated visual evaluation where like an image of the whole cervix is evaluated uh and whether there is a precancer or not. And so I'm gonna talk today, I'm gonna talk about the first two dual stains ology is gonna be my main focus and then briefly on P 60 histology which fits uh to uh what Sarah um said in the beginning about classifying C two. So automated dual stand cytology. Um Here's a little background. Um The first uh row shows histology images and the second row shows cytology images and they're matching um cytology and histology images in each column. So the first is uh um precancer uh with uh characteristic P 16 staining through all layers of the epithelium. The second uh in the second row, we see um individuals cells that have a dual stain. So it's a bronze stain for P 16 and a red stain for P 67. And these are characteristic, it's a characteristic staining pattern for uh that, that is, that indicates early HPV transformation in contrast. Um They're a handful of P 16 stained cells and a normal epithelium, but they don't have the dual stain pattern and can be clearly distinguished from the dual stained cells. We have evaluated uh manual dual stain quite extensively and uh a number of studies, many of them conducted with Kaiser Permanente, Northern California. Uh These are five year data uh that uh we published a few years ago and where we demonstrated that D stain here in the red curves compared to the blue curves for cytology. So, dul stain has um uh among the positives indicates a higher risk of C three or greater compared to cytology. And among the negatives, it gives better reassurance. The risk is lower when you dual stay negative compared to cytology, negative. And that really persists over five years. So this gives you a better uh discrimination for having or not having precancer compared to cytology. Uh with long term reassurance. This is based on manual evaluation. And in parallel, we uh developed uh the algorithm to really do an automated evaluation of these slides. And I'm gonna show you in a moment how the performance looks in this automated evaluation. So we used uh three different populations. We, we, we, we started off going broad with different cytology platforms. And we included both uh slides from uh cervical samples, but also anal samples from an anal screening study. We used fim prep samples and short pass samples. So we really went for a lot of heterogeneity from the starting point and then we uh did validation in a very large set. And um as you see, we uh used um 100s of slides for training, but what we actually used were smaller tiles on the whole slide. So basically our training set consisted of uh over 20,000 tiles which are pieces of these slides that were individually evaluated for having d standard positive uh cells. And that was our training set. And then we evaluated on the whole slide level. And this is what it looks like. Again, for short path in the first row, for thin prep and the second row, what we're doing is we're, we're, we're like basically slicing up these images into small tiles. And this is a single tile for each of these um um systems. And then we're evaluating on each tile, whether we see a dual, same positive cell. And then we count how many tiles are positive. And then we get a score for the whole slide. Um These are the uh different networks that we used. We used a simpler network first that we uh kept for uh ThinPrep slides. And then we uh developed uh developed that further into a much more uh complex network shown here um for the sure path slides. And we are currently building one model that, that we really can use for all the different slide types. And this is the more complex way of really generating these algorithms uh that include several circles within the training set to improve classification before a algorithm is blocked and then applied on the validation sets. And then we basically for each slide, we get the number of dual same positive cells and then we can develop a threshold and classify each slide as being positive or negative. And this is the uh results um And in comparison to the manual dual stain and to pap cytology, which at the time was the clinical standard. And you see here that um the sensitivity was similar across those I mean dual stain has higher sensitivity than cytology. But what is quite impressive is the increase in specificity compared to the cytology in the setting So e the manual dual stain had a 10% higher specificity compared to pep cytology. And then the automated further increase the specificity to uh 61% dramatically reducing the colposcopy referral while detecting the same uh number of cases. So this is um uh a really nice um algorithm that we're currently evaluating on a large scale uh clinical series. There's also a software tool developed by Neil Srb and Lamon from the University of Heidelberg. Um that really um can use this algorithm and use it to display individual events on a slide and it allows us to zoom in look at cells. So this allows to do an assisted evaluation where tiles are ranked by the scores. And so you, you can uh still have a an observer look at these S um cells but but you can um go through a slight, much faster because they're presented in the rank of the abnormality. So with that, I wanna very briefly uh talk about P 16 histology and um uh show you some early data. So uh as Sarah Feldman said earlier, P 16 can be used to um improve accuracy of uh cervical histology. This is kind of an example here, an H and E slide and the matching P 16 slide that highlights areas of precancer. And um it it has been P 16 has been evaluated as a tool to improve accuracy. And in some studies, it has shown remarkable effects in other studies. Um There have, there were some um issues with how P 16 was used and there uh it's, it's really important to have proper training and to use P 16 as an adjunct to uh uh um conventional morphology and not as, as a standalone marker. Um But, but this is an example uh from a study by Christian Bergeron where um when just looking at the H and E slides, there was a wide range of performance with some pathologists um having relatively low sensitivity for detecting uh um a precancer uh compared to a gold standard. Some do great, some are expert. Um We all uh highly accurate pathologist, others are um um lack sensitivity. And then when adding P 16 to that, um they converge. So this really helps um uh some of the pathologists that, that really were um further down here and for others, it doesn't make a big difference. So it really um kind of can equal out pathology performance, but that's a really important uh tool and, and this is um uh kind of the basis of the last um approach that we heard about. But I wanna emphasize what um Sarah said earlier. I, we, we think it's important that even though last is a dichotomous um uh result, primarily, it is important to try to qualify hil into CN two and CN three as much as possible. Now, um obviously this uh very strong P 16 stain uh really um is, is, is a great um target for automated um evaluation. We've done that we, we generated whole slide scans. Again, we cut up these uh slides in smaller tiles. And then the first step of the A I is really to, to identify the epithelium. And that's based on an algorithm that um Felipe Miranda Rui uh from uh Heber and developed and then we quantified uh P 16 in this Epithelium area with another A I based approach. And this is very early data again generated by Felipe and, and analyzed by Amy Tau from our group showing the distribution of P 16 in uh precancerous and a normal and the matching R OC curve. And that's uh very exciting early data. But as I said earlier, we have to do a lot of work uh to really get this to the next level and that's what we are currently doing. So um to summarize um as we move to implementation of some of these approaches, there is a lot of work to do Cyto reader. The automated te approach is uh much further along and we have uh several sites with installed uh systems. The automated P six in histology is an early development. We really focus on repeatability and portability issues. We need to look into platform specific versus platform independent training. Um And uh it's really important to think about clinical use cases um to uh uh get to the next development steps. So what is the actual clinical need and that may differ across different sites. And we're in a lot of discussions with many partners about that. With that, I wanna uh highlight the team members um that have done a lot of work for, for these projects. And I'm looking forward to your questions. Thank you. Thank you Nicholas for a wonderful talk, very insightful and uh comprehensive. Uh So II would like to ask you a question uh is see two here to stay forever after the last project. Well, I mean, uh so CIN two is, is a very challenging group. So what last does, I mean, last kind of emphasizes that you can um really like if, if you use P 16 on CN two, you can, you can uh get rid of some of the lower end of the spectrum and then, and, and reduce the over calling of, of um low grade changes as, as as high grade. That's, that's an important tool. Now, we, we think that it is still um uh useful information to, to separate in two and three, particularly for the reasons that uh Sarah uh addressed earlier. Like, like if we, if we have only one group, then it's gonna be very hard to decide um to, to make any decisions about um uh conservative management surveillance. Um because we, we cannot distinguish uh SYN three and sinn two and, and we don't wanna leave SYN three untreated. Now, ultimately, I think we wanna move away from morphology, we wanna move towards the molecular classification. And with that, we can probably do much, much better decisions. And that could include uh knowing the genotype, maybe knowing about P 16, maybe knowing about methylation. I'm not suggesting we run all these tests at the moment. But I, I'm, I'm saying we're, we're kind of in a process where we wanna move towards um really a more functional molecular classification of precancerous. And when we have those, we can probably um abandon some of these morphologic um um criteria. But before we are there, I think there's still some uh relevant information despite all the challenges with CIN two. Yeah. OK. And uh thank you. And we have a couple of uh very brief uh A I questions. Uh The first one would be in A I can we add on data on the initial training, data set without going through the whole process. Yes, absolutely. That, that is, that is very commonly done. So, so that there's, you can expand your training set and, and that's actually something we did. And, and I it was on that very uh very busy slide and I couldn't go into all the details. But something we've done in in that first development phase is basically generating a first training set, running it in the training set and then seeing, OK, what are we still missing? And then after you, you identify what you're missing you can enrich your training set for specific um elements from that. And, and that's what you can also do for uh to, to when you move to a different platform. For example, if you use, if you wanna use a different scanner and, and the, the, the scanning, the, the the scans look somewhat different, you can enrich your training set and make it broader and, and more portable. The the alternative is to to directly train on a different system. So, so to have like two algorithms and, and the, the it really depends on the specific question when you would do what? But the like augmentation of your initial training is that, that's, that's standard. The other thing that that is uh currently do done in by, by almost all groups is that you can modify your training set, you can, for example, rotate tiles, you can um make them a little blurry. You, you can modify your training set and then augment it that way and, and get like a better performance. So that, that's kind of a wide range of things. But Aug like augmented training sets are very common. Yeah. OK. I do have a final question by professor at my mentor. Uh He says he asks, how far is today? The progress and the use of A I in colposcopy? And may I add to that uh what would be the, the utility of a colposcopy in the future? Uh Given that uh uh I have read recently an editorial by Mark uh Schiffman uh commenting on the differences between C A and three. For example, can three New Zealand uh experiment were much more different than the Can three is that we uh come, come up uh uh during the scr cervical screening process uh today. So maybe uh the, the risks for uh recurrence are much more different in a different in, in different cases of uh can three. So uh could this be uh a AA a AAA place where colposcopy uh combined with A I might be useful in the future? Yeah. So, yeah, II think these are two slightly separate question. I'm, I'm trying to address both of them and, and by Professor Agresta. I so um it, it is uh there's a lot going on on um A I based Colposcopy and, and we um I'm actually currently at wh O where we had a meeting on new technologies and, and we discussed a lot about that. So they're, they're, they're large uh uh studies, large trials ongoing um around the world. This is, this is it, it's a very similar application. We have a flawed um manual evaluation. Uh We have a, we have, we have actually like a much more flawed approach uh than, than what we have for histology and cytology with vi a really being um uh not reproducible and, and having low accuracy in most places. But so the automated approaches that have been developed so far are, are very, very attractive to improve that. And they would provide like a really like a um uh and point of care evaluation that, that, that can allow to make a treatment decision in a screening setting. And, and, and that is I think a major breakthrough that will be very important for um for elimination of cervical cancer. So we need to evaluate those in large studies that's underway. And I think in the next two years, we'll know uh how far we will get with that with regard to your second to the second part of your question there, there is definitely heterogeneity of CM three. So all of these like and, and we already and the same markers that I described for discriminating like a different class of CN two can also be used for CN three. I mean, we have CN three caused by types who who hardly ever um progressed to cancer. So that, that's a different story than P 16 related CM three where we probably want to treat uh as quickly as possible. And so the, the point about the New Zealand CM three is that these were CM threes that were detected very late and they were very advanced and uh uh cm threes compared to some of the CM threes that we see in uh heavily screened populations or in trials like the a trial where we have very aggressive screening and management procedures So basically, there's a spectrum from very earliest in threes that may, in many cases uh go away by themselves or those that have persisted for a long time and likely have a very high risk of uh progressing at the moment. We don't have a fully reliable marker of progression to of, of invasion. That's like if we have that, we could make that decision on the spot and say, yeah, this is one that we need to treat. This is one where we can do more surveillance. But again, with all the work on biomarkers, um this is that isn't a possibility that we will at some point have a marker to differentiate that. Oh, thank you. Thank you Nicholas for uh for your answers for your uh time and uh we wish you a nice life. Thank you. Uh Thank you for your participation, participate, all of you. Uh Let me now, just welcome uh Sophia uh to uh say hi uh on behalf of the board of the dog Sophia. One, thank you, first and foremost IV for accepting and working with Endo for uh doing this webinar. Um Today, I would like to thank all of the speakers for offering their time and their kindness to speak to all of us today for the enlightened speeches. I'm sure that all of us, the trainees and young specialists have earned something today. Thank you all for your time and uh having uh dryness of mine. Thank you very much. OK. So uh let me also thank uh the IPV S Board, the Officers, the Education Committee for this wonderful initiative uh for the joint initiative with OC. I hope all the the young trainees uh were fascinated by these talks. We tried to uh shed some light on gray zones in cervical cancer um prevention in the cervical uh disease uh in general. And uh to also to uh to give insights in the future of uh cervical cancer prevention. Uh We hope uh to see you all in the I PBC in the coming I PBC 2023 in uh Washington DC in the US. And uh let us hope that we uh we will have uh AAA very uh interesting a series of webinars during 2023 as we did in 2022. So stay tuned and uh thank you all for attending. Have a nice uh evening everyone. Bye.