Exploring AI Matters
Our mission is to help the policy community understand the breadth and richness of AI and the potential for such technologies, wisely applied, to augment all sorts of human endeavors.
Some AI tools are able to assist humans in performing tasks faster, more accurately, or more efficiently. Some, however, are inaccurate and unreliable. Who or what we hold accountable for these flaws, and what incentives we do or do not create for their correction will influence AI’s hand in how we work.
In this series we will refine, sharpen, and clarify your understanding of AI.
Exploring AI Matters
Episode 16 - Improving Colonoscopy with AI
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Colorectal cancer is one of the leading causes of death in adults worldwide. The most important diagnostic and therapeutic technique for colorectal cancer is the colonoscopy. Gastrointestinal doctors are working with computer scientists to develop AI-based technologies to help enhance outcomes.
We are fortunate to have Professor Doctor Julia Mayerle of the University of Munich as our guest. Dr Mayerle, Chair of the Department of Medicine at LMU, a recognized center of excellence for research in oncology, is an expert in gastroenterology and hepatology with a distinguished career at premier medical institutions in Europe spanning decades. Both a researcher and a clinician, Dr Mayerle has published extensively in a wide range of research areas in internal medicine. [2024-02-16]
Welcome to Exploring AI Matters. This podcast series, previously known as Mind the Gap Dialogues on Artificial Intelligence, will continue to appear in the ABA series to the extent that, in addition, all of the episodes, old and new, will now appear under our new podcast name, Exploring AI Matters. Thank you. Colorectal cancer is one of the leading causes of death in adults. While progress in diagnosis and treatment has been driving mortality down for decades, such cancers are expected to remain a leading cause of death for years to come. Reportedly, up to 75% of colon cancers could be avoided with regular screening and other healthy behaviors. The most important diagnostic and therapeutic technique is the colonoscopy, during which an experienced clinician carefully examines the interior surface of the colon. Potentially dangerous growth called polyps are removed by the clinician during the colonoscopy. As we have seen with other areas of healthcare, gastrointestinal doctors are working with computer scientists to develop AI-based technologies to help enhance outcomes. Recent use of AI has improved screening and therapy, substantially reducing mortality. Welcome to Mind the Gap Dialogues on Artificial Intelligence. I am Roland Trope, a national security lawyer.
SPEAKER_05And I am Mark Donner, a computer scientist. We are your hosts for this episode of Mind the Gap. In addition, we have two more hosts.
SPEAKER_00Hello, I'm Alma Adams, a national security lawyer.
SPEAKER_04And I'm Charles Palmer, a computer scientist.
SPEAKER_03Each episode will be led by two of us, with the others adding impromptu questions and comments as the spirit moves them. We are fortunate today to have Professor Dr. Julia Meyerley of the University of Munich, a recognized center of excellence for research in oncology. Professor Meyerley, Chair of the Department of Medicine at the University of Munich, is an expert in gastroenterology and hepatology with a distinguished career at premier medical institutions in Europe spanning the decades. Both a researcher and a clinician, Dr. Marley has published extensively in a wide range of research areas in internal medicine. Welcome, Dr. Marley. Can I call you Yulia during the interview, Dr. Marley?
SPEAKER_01Yes, please. Please do so. And thanks for the invitation. It's a great honor, and I'm hope that we will have a great podcast today.
SPEAKER_03Thank you. Yulia, could you explain to us what a colonoscopy is used for and why it is so important?
SPEAKER_01Well, Roland, you just rightly pointed out that a colonoscopy is an endoscopic procedure aiming to examine the colonic wall. So what you actually do, you have a scope, and there's a camera on the scope, and you introduce the camera into the colon and you will detect pathologies. So, and those pathologies can be inflammation, like we all know that somebody can develop diverticulitis, or you could have just an inflammation caused by a buck. And sometimes your physician would like to see how the extent of the inflammation or pathology, and therefore he uses a camera and just maneuvers the camera with which can be angled to every side into your colon and looks for it. The most important implication is, and this is also what you pointed out, is to detect pre-malignant lesions because those pre-malignant lesions, or you could call them polyps, will turn into colorectal cancer. And colorectal cancer is the third most common cancer worldwide. It's the second most common cancer in women. And if you just look from North America, colorectal cancer features among the highest incidents. So in 2020, in the United States, there were 147,000 people diagnosed with colorectal cancer, and that could be prevented if these small polyps would have been removed previously. So the United States were really at the edge of research 40 years back when they introduced the so-called national polyp studies. And these studies showed us that if you remove these small polyps, you can and you uh colonoscopy reduces these polyps by 90 percent, then you can decrease over time the colorectal cancer rate by 50 to 90 percent. And this might be striking to you why I'm telling you if I remove the polyp, then I remove the cancer, and then you tell me it's only 50 to 90 percent, and that is because of the missed rate. So you miss small polyps or for some different causes like a bad cleansing status or a tired endoscopist, you might miss more or less, and those polyps can then turn into cancer, and this is why there's a wide span on the efficacy of colonoscopy and removing polyps. And this is where artificial intelligence comes in and might help us in the future.
SPEAKER_03So let me clarify a few things. First, the you mentioned a camera. I take it this is a camera that has its own illumination in order to be able to see in there. And is the camera that's taking images, is it taking video, is it displaying the images on a screen? Is this in real time viewed by the clinician? Could you explain just a little bit more so we understand what the tool is that's that's being operated?
SPEAKER_01It's it's a real-time camera. It's like if you use your iPhone and you walk around and you can take a video or you can take a picture. So to record, you can you will record pictures, but you can also record videos to be analyzed later. But in general, it's the clinician examining the colonic wall on site in time, in real time.
SPEAKER_03And from what you said also, there are certain things that improve the performance of the colonoscopy and the detection rate. If the prep has been done well by the patient, if the clinician is wide awake, if the clinician is more skillful, let's say. But you're saying where those things are diminished, there's an error rate, if you will, or a missed detection rate. You say that AI can help reduce that error rate. How does AI do it? And does it do it after the procedure? Does it do it during the procedure as by examining the images? How does this happen?
SPEAKER_01So when they when AI started to enter the field of endoscopy, it was offline, and the images or videos were analyzed later on. And then there's a learning curve, but right now in 2022, it's a plugged-in system, and there are green green arrows or circles or squares which point out lesions on the screen to the examiner in real time. So if the camera stays on the spot and if the system detects something which the system believes might be a polyp, it will give you a square and then it will say analyzing, and then it will tell you adenoma or no adenomer. So they are non-malignant lesions, which the system tries to discriminate from pre-malignant lesions. And these you can then remove if it's an adenoma, and you go on with your investigation. So it's plug-in and real time.
SPEAKER_03Charles, you had a question.
SPEAKER_04Yes, I wanted to jump in here. This is fascinating, and the whole idea of this adjusting or augmenting the skill and expertise of the clinician, that seems to be a great place for the AI to help out. Have you been in situations where you disagree with the AI? Oh, yeah. What do you do in those cases? I know you you do what you need to do, which is in the best interest of the patient, but how do you feed back to the AI? Whoops, you missed that one.
SPEAKER_01Well, it's rather not the miss rate, to be honest. So the sensitivity of the AI systems is often higher than, if you want to say, the sensitivity of the clinician, but the specificity is much lower. So to put that in layman's turn, there is a very high number of false positive findings suggested by the AI system. And in those, as an experienced clinician, you just decide this is not something I want to remove. If you are insecure and you might disagree, then what you do, you're going to remove the polyp, bearing in mind that the complication rate of removing a polyp, so bleeding rates, perforation rates, and so the possibilities of endangering a patient is very, very low. So you just saw it, whether you remove it or not, and you you can disagree, and the false positive rate is 30%. The false negative rate, on the other hand, is pretty low.
SPEAKER_03Can I ask a corollary to Charles' question? When AI is used in other areas, such as in military uses, one of the major challenges to transitioning from developing it to using it is that the end users don't trust it. They're not willing to rely on it. And if you don't rely on it, you don't end up becoming expert in its use. Did you and your colleagues find that you had difficulty making that transition to trusting AI? Or did the research lay the groundwork so that wasn't a problem?
SPEAKER_01I don't think that we are in the situation that all of us really trust the system already. And I think it's very important in medicine that we understand that it's not taking out over our skills, but it's used to help us and support us. And so I would trust my clinical experience, and here comes the problem. I mean, I've been educated in an era without AI. So I'm used to taking my own decisions, but that might change in future generations when AI is already introduced in your training. And just as an example, which I found fascinating, is if you just look for the eye movement of an experienced endoscopist examining the colonic wall, the eye movement and the um is much well spans a greater area than if you look at a trainee who has AI to help and just looks for the squares turning up on the colonic wall somewhere. So I'm I'm not guided by the squares and looking for the squares, like in a computer game, but I'm just looking around and you can actually measure that. And my strong belief is that we should continue training our residents without AI, and maybe even looking at their um at the eye movement, take that as a measurement of quality control before allowing them to use AI.
SPEAKER_03As you were saying this, you were holding your hands up on either side of your face, and I got the impression that the eyes sort of move in a circle around.
SPEAKER_01Oh, they do, they do, they do. I mean, if you if you look into if you have to think about the column, it's like looking into a tube, but the it's a folded tube. And you what you have to do, you have to look behind the folds, and this you can only do by maneuvering your scope and really trying to angle it to look behind every fold. And I mean, there is a very good example where AI and colonoscopy does help us a lot, and this is actually it was brought up by Google. So by measuring the eye movement and also the camera, they they made a map of the colon. So the whole after the examination, it shows you which areas you have visualized, how long you have visualized those areas, and that relates to the number of polyps you might have detected if there are any. And that is very helpful because you recognize what you have not seen, and this you won't you won't um recognize during the examination. I mean, you you you need to try to be very careful. Another thing where AI really helps us in colonoscopy is that it can actually give us an objective measurement of the cleansing status. So if the preparation wasn't good, and if you still have too much feces in your colon, then obviously either you um the examination will take ages because you have to remove all the remains by flushing, or the quality is bad. And I mean, in in it, there are different environments. Like if you do your your colonoscopy, and I'm just talking about German situations because I'm not that much aware of the Americans, and I don't want to frighten you, but a a physician in Germany in a private practice does between 15 to 20 colonoscopies a day. So that is a very really short time span he can spend on an examination. So, as a quality control measure, the pullback time should be at least six minutes. So AI can actually measure those six minutes just by knowing now I've visualized the end of the colon, and that is the vasula of Bauhini, that is the connection between the small intestine and the large intestine. So if the AI recognizes this or the sequel pole, and then you start to pull back, it starts to count. If you have not taken up six minutes before you actually remove the scope from the bowel, then you get a red light. And that is what is definitely increasing the quality. So if you are under time pressure, that's helpful. And it's helpful for the patient.
SPEAKER_03Just one more question before I Mark wants to ask a number of questions. I think. When did you and your colleagues first start to consider AI as a tool to be used in colonoscopies?
SPEAKER_01So the first study was published back in 2003. So it took nearly 20 years until you have plug-in systems, which can be used in in any practice and with any endoscopy system. And there are a couple of companies who've actually invested a lot of research um in these technologies and brought them to the market.
SPEAKER_05So is AI best suited to detect certain types of polyps? Are there other benefits from using AI?
SPEAKER_01Well, I think I've had a couple of benefits, but let's be um honest. I mean, distinguishing between the different polyps is the most difficult part. So currently we say, well, that's an adenoma, and that is a non-adenomer. But for example, there's this certain type of adenomas which is flat and spreading.
SPEAKER_03I I need to interrupt you. I I don't think you've defined adenoma for us, and our audience may have lost you for a moment.
SPEAKER_01Okay, so adenoma is a polyp, and they're different types of polyps. So they they can either grow like a cauliflower, and then they are easily recognized, or they can be flat and spreading and might have a slightly different color. So they are slightly yellowish, and there might be some mucus on top, and those are notoriously difficult to detect in colonoscopy. So basically, those so-called serrated adenomas or flat spreading adenomas are those which are which are known to turn into cancer, which are mainly located on the right side of the column. So that's the side you reaches last, and they are the ones which are frequently missed and are probably responsible for the so-called interval cancers. So if you if there's a cancer development between a screening screening colonoscopy and the next scheduled appointment, so and I mean the AI has different systems, so there's the so-called CAT C that is just highlighting lesions, and then you have CAT X, which tries to predict histology, and then you have CAT Q, which is the assessment of the quality of the examination. So Cat C is pretty good right now, so lesions can be detected for CAT X. There's a lot of work to be done, especially because you have to map the endoscopic picture to histology, and that is simple as long as you have removed a polyp and send it for pathology. But the biggest obstacle is that you will not remove healthy areas. So you need to prospectively set up a database where you have a control cord, and that is the biggest challenge in setting up a system which can predict histology. And to your second question, I mean AI could, for example, help in semi-automated reporting. Like you pull back, you pull back your scope, you're visualizing all the areas, you have your squares, and then in your report, it could already say before you even start typing, there were two adenomas on the right-sided colon, pulled like a cauliflower. I don't want to use the medical terms for that, and you would just start to correct it and not not to have rewrite it from scratch. And that might be pretty helpful. It could also just add your pullback time, it could add a lot of features, like how long you needed to actually reach the Valvola Bahini in the first place, how long you it took you to remove the polyps, which instruments you used. You could uh it could already record the instruments and give them codes for reimbursement. It might even in the future judge whether a bleeding has to be expected from the side where you remove the polyp. It could suggest your next appointment to actually screen. So there are many options which you could feed in the system helping to actually adhere to guideline recommendations.
SPEAKER_05So there's a lot of work to be done, clearly. So we understand from our earlier conversation that malignant start out polyps start out as benign ones, which can then be removed if they're detected. And you basically said with AI, you found that the benign polyp detection rate increases what seems to me to be a relatively modest amount from about 25% to 33rd, 33%, but that this results in a 2x reduction rate of missing of the of the adenomas, the polyps that need to be removed. How does it happen that that's such a huge improvement?
SPEAKER_01So you schedule an appointment after your first screening colonoscopy, if you've if you've found polyps. So if you miss the polyps, it's a different time frame. So and if in even if the polyp is relatively small, you give it a long time to grow, and then it can be cancer. So if you decrease the miss rate, then the the advantage is that you are more sure that your next screening colonoscopy will be at the right time, and you will not need to schedule in one in between, and that is cost efficient and reduces healthcare costs, and it also reduces, say, the burden for the patient because you undergo cleansing and a colonoscopy is not something you would you would want every day, and you would remember as one of your best experiences. Experiences in life. So we need to we need to somehow ensure that we are not losing our patients because we do procedures on them every other day, which they totally dislike.
SPEAKER_00So compliance is something which is very important. So I've never heard anyone describe a colonoscopy to me as the best experience of their life, but hopefully this conversation we're having and the advantages we're seeing, the technology will make it closer to that. But listening to you talk about your work and the benefits of AI and sort of the expertise that you're bringing to this conversation makes me think about an important distinction. And I'm going to make the statement and you're asking you to tell me if this is sort of a fair assessment of the role of AI and colonoscopies. And it's really to help identify sort of regions of interest during the colonoscopy, maybe you know, help mark abnormalities that are consistent with the polyps, but the diagnostic assessment, the decision making around how to manage suspicious polyps remains in the judgment of the clinician or the doctor, right? And is that a fair assessment? Because we've talked a lot about what a colonoscopy is, failure rates, all of these sort of technical terms, but that led me to think about you using this as a tool, as another tool in your toolkit in order to do your assessments for the patient's well-being.
SPEAKER_01I totally agree. And I mean, the the the difficult thing in answering your question is I have to understand whether I would call myself a dinosaur. And I mean, we usually stick to the technology we have been educated or taught because we know that works, and we are not always readily ready to accept new novel technologies. So I look at AI presently exactly as you pointed it out, as a system which aids me to improve my care, but I would not rely on AI making a diagnosis. So I would be absolutely ready to overrule any square showing up on my screen. And I mean, I had a funny experience a couple of weeks back. So I had to scope a cardiologist, so a colleague, and he decided that he would like to do the colonoscopy non-sedated. Only very few patients asked for that, but he wanted it, and I was happy to do that. So we had a very nice conversation, and I did this, I did the examination using artificial intelligence, partly because I wanted to show him the novel technology we are developing, and we were discussing side by side whether to believe or not. And it was interesting that even the cardiologist, well, it's too slow, and we have to look for too long time on that. Both of us have already decided there's just nothing, and that's a fold, and you need to look at that for another 20 seconds. Let's go on. So there is a lot of improvement if you look at it, but I'm pretty sure that there's a lot of potential. And I'm sure that in a couple of years' time, every practice will have these technologies and patients will ask for it because it gives you, it secures you. It's that it's there's no way harming. And you will be, I mean, you're lawyers. Somebody will sue us if we've missed the cancer, if we say, well, we've missed the cancer, but AI also missed the cancer. Maybe we are pretty, maybe we are better off uh if if you take us to court.
SPEAKER_03It's it's interesting. You you've changed my concept of bedside manner now. But I've also been in procedures where I was not sedated because the procedure required that I remain awake. And this really makes me ask you to what extent during this transition to the use of AI, have you been telling your patients that you're using the tool? Is it important? I mean, you you just said in the future patients will want to require that doctors do this, but if the current patients are not being told by their doctors on now using AI, do you need to get consent from them to use AI? So, how I guess the larger question is how is the application of AI-assisted techniques changing your practice of medicine and changing the way you relate and talk to your patients?
SPEAKER_01So we have not we don't need to inform our patients that we would use AI, because basically we are not informing our patient which scope we're gonna use. I mean, if they ask, for sure, we will say we have an Xierra 190 system, which is the most up-to-date system from Olympus, but I don't want to advertise that right now. So if he comes up and he's asking, for sure we will reveal this information. But we also not telling our patient which snare we use or which forceps. Right. So legally we we we are not we don't need to do that, and I don't think that it does help the patient. I mean, if I've got a patient who's very unsecure about the procedure and is asking a lot about missed polyps, then I will I might tell him and say, look, we have even developed a system to support us, and for sure we will use it, but it's only I would rather use that to elevate his feelings and give him insurance than advertising it.
SPEAKER_05Thank you. In your earlier comments, you you reflected on the fact that the the practitioners you're training using AI techniques seem to work differently than people trained without AI. Do you think that that the next generation of practitioners, the ones being trained with AI, are at risk of not developing skills or becoming expert in judgment by relying on these AIs?
SPEAKER_01I think you're pointing to the biggest problem in artificial intelligence or machine learning and medicine. I my biggest fear is that we change our training. And what we don't recognize in our daily practice is that the intake of information we do is not only, for example, pictures. To give you an example and stay in the field of colonoscopy, but there would be much better examples, is if I push the scope, I get a feeling of the tissue and the elasticity. I also recognize if I inflate the colon by air, how it inflates. And that is an information which I subcortically incorporate in my interpretation. So in a cancer, the the inflation is reduced because there is infiltrative uh growth and that reduces the elasticity and the the possibility of movement, but the system is only analyzing pictures and it's mostly static pictures, at least as we have trained them on pictures so far, because there were no movies movies to train it on. So it lacks information, and I'm not even sure that we understand what information we take in and amalgate in our decision making. And I mean the system can only be can only do what it you train it for if you don't know what you were trained for. It's hard to actually develop a system which will eventually be better than your own judgment.
SPEAKER_05This reminds me of something that the late Al Newell told me years ago. Early in his career, he worked on training Air Force operators of an early warning system. And they had done an analysis that predicted a particular capacity of a team of people doing this work. And what they discovered was that after the team had been trained and been practicing for a while, their capacity became substantially higher than was considered possible. And what they ended up doing was attributing this to what they called system learning. And what I hear you describing is things that you may not be able to articulate just yet, but that are things that you've internalized into your understanding over the course of your years of experience, with things like, oh, if I put some more air in and I see it expand, I get a sense of how you know how elastic the tissue is, and that gives me information. At some point, one could imagine training systems to do that, but that needs that needs again the judgment of experienced practitioners to say, yes, there's actually more to it than just the image.
SPEAKER_01Right. You're absolutely right. And I think that is that is the way we have to go. And I mean, if we can incorporate all this information in the coming years, then the system will get better. And also sometimes we have limitations. Like there was a system they developed earlier, which was not AI, but where they tried to actually visualize the the colon on three different screens. So usually you just have one screen and that looks straight. But there the camera did not only look in front of you, and by angling it, you look left or right to the um behind the falls, but there were cameras on each side, and that gave you something like a circular view, but I couldn't do that because after 20 minutes looking at three screens, I could not concentrate on the adenomas or what I would look wanted to see, but it just got me a headache. And you could say, yes, you can you can train yourself, but I think there each generation has limitations to how much information we can intake. And the younger generation, growing up with tons of computer games, very rapidly moving options on on different sides, might be much better equipped. And there are studies, for example, that somebody who has done a lot of computer gaming and a lot of joysticking in their youth are better or do learn colonoscopy or endoscopy more rapidly. We all reach the same level, but they are quicker.
SPEAKER_05Fascinating. So as you look ahead for say the next you know, five to ten years, do you think patients will start to insist on the use of AI in screenings? If so, what might patients need to understand for that adoption to progress more effectively and thereby save more lives in the near term?
SPEAKER_01I think they should ask for it because I'm pretty sure that there's a rapid development curve, and we will get better very soon. And I think there's nothing for them to lose. The only thing what could be lost at present is that you remove unnecessarily some non-adenomal polyp, but the risk of complication is so low that I would not mind doing that.
SPEAKER_03But that takes me back to the question we were talking earlier about. If you don't tell them that you're using AI and you don't tell them that you're not using AI, how does a patient know that the tool is available and how do they know to say to their doctor, do you use it? And if the doctor says, Oh, I'm I'm not trained to do that, that maybe they want to go to a doctor who has been trained to do that.
SPEAKER_01So there's no official training for using it, it's pluck and play. That's one thing. Um as it's a novel tool, it's very likely that a doctor will tell its patient, hey, I've my I'm equipped with the novelest technology available on the market for your benefit. So we will let them know. And there are also patient advocate advocacy groups who know about it and especially ask it and broadcast it on their website and use it in their flyers. So I think that the spread of the word will be out in the world very, very soon. I don't think that there will be few people not knowing about the possibilities.
SPEAKER_03I know this is a hard thing for you to answer, but since you you've been in the United States several times, you're here at a conference in Florida right now. Can you give a sense during the next four or five years whether you think the US patients or German patients are more likely to be comfortable with this procedure?
SPEAKER_01The procedure, the AI or the colonoscopy? No, well I would be let's let's let let's make let me make one statement. I would hope that Americans and Germans, to be frank, everybody would accept colonoscopy as the only screening examination we have so far which really reduces cancer mortality. It's not mammography, it's not PSA screening. The only exam or the the only uh procedure we can offer to our patient which really reduces mortality, cancer mortality is colonoscopy. Having said that, only 12% of the illegal population accepts to come for screening. Wow. And so I don't bother about AI. If we can increase the number of people coming to our practices asking for screening colonoscopy, then we are winners.
SPEAKER_05Yes, well, that's a sobering statistic that you just gave us. Very sobering.
SPEAKER_01And it's reimbursed, it's advised, it's reimbursed. We have a cancer screening month, so it's it it it's it's a nightmare that people still do not accept to come.
SPEAKER_05Wow, that that by itself is just eye-opening. So sort of a last question here: what do you hope to see from future developments in AI techniques in your field?
SPEAKER_01Well, I would I I would think that I wouldn't invest that much on really making a diagnosis, but I would rather uh invest in supporting our systems. To give you an example, if you if you do 20 colonoscopies a day, you get tired at one point. If the system tell you tells you your eye movement is slowing down, like if you are in your car and your car tells you, please take a break, that would be very helpful for the patient. And it's simple. And so if if you have your quality measurement systems installed, that I guess will be very helpful and not very challenging in the development.
SPEAKER_03I get worried though.
SPEAKER_01If the Google thing, if you have a very experienced endoscopist in another room and a junior or less experienced endoscopist in in the next room doing an endoscopy, wouldn't it be great if there's a red light in the room of the experienced endoscopist if the machine recognizes that the resident hasn't seen enough of the mucosa? Like pull push back too fast, haven't visualized it properly, so that there's a red light, and you can just walk over and say, Can I just assist you for a moment? There seems to be trouble.
SPEAKER_05Wow.
SPEAKER_01So I think there are tons of tons of applications you could think of, which not necessarily would interfere with the medical ethos and understanding of our profession and therefore be readily accepted by the community, still supporting patient care.
SPEAKER_03And you don't feel that doctors would feel they're under too much surveillance by that kind of quality control.
SPEAKER_01I think we need to be under this surveillance. I mean, if you say, I mean, in the States it's it's less obvious, but in Germany, for example, from my personal research field, we have these comprehensive cancer centers. And pancreatic cancer is a deadly disease. And one option is to resect the patient in a very early stage. So our comprehensive cancer rules ask for a mortality which is below 5%. And it's still a high mortality rate, if you want to say you undergo a resection and you die, your mortality rate is 5%, it's not good. The mortality rate recorded on the insurance database in Germany is 11%, showing us that only a very small number of patients are resected incentives, which are quality control. So we have to survey or we have to allow to be surveyed in the interest of our patients.
SPEAKER_03Well, I Mark, I'm going to leave the final thank you to you, but I'd like to say before you do, Dr. Meyerly, I find your explanations extraordinarily articulate, especially since you're expressing yourself in a language that's not your native language, and listening to you clinically explain so clearly. I deeply appreciate it. Outstanding. Yes.
SPEAKER_05Thank you. Thank you for being our guest today, and thank you for opening our eyes, certainly my eyes, to a number of amazing things. It was quite a wonderful conversation. Thank you very much.
SPEAKER_01Thank you for asking me to join. It was just a very nice afternoon, and I hope that our audience will enjoy listening.
SPEAKER_03I think they will.
SPEAKER_04No doubt. You have a real gift for explaining what is typically unexplainable.
SPEAKER_03We thank the business law section of the American Bar Association for their generous sponsorship of the production of this podcast. We welcome questions and comments from listeners. Send email to comments at mindthegapdialogues.com. We read all comments and questions and will try to respond in the letters section of a future episode. If you are writing about a particular episode, please do mention a specific episode number. Please also do include pronunciation tips to help us properly say your name when we reply in a subsequent episode. See you next time on Mind the Gap Dialogues on AI.
SPEAKER_02Thank you for listening to the Aviate Business Law Section Podcast Series to the extent that the section offers a robust collection of content. To explore more about this topic, or to learn about joining the section, visit anbar.org slash bizlaw. That's B-I-Z-L-A-W.