Oxford Open Grand Rounds

00:00:00.000–00:00:07.040 Olivia: Thank you for joining us for our Oxford Open Grand Round September 2025 where we are asking the question what is next for AI in surgery.

00:00:10.600–00:00:20.800 Olivia: So just as a quick welcome and some housekeeping items here. This presentation is being recorded and will be shared with you via email after this event so you'll be able to go back to it.

00:00:21.680–00:00:31.160 Olivia: We will, you know, very much welcome your questions in the Q&A section of this webinar. So please do pop them in at any time in the Q&A function. we will try and answer them as we're going.

00:00:32.680–00:00:36.920 Olivia: We will also have time at the end to be able to answer your questions. So please do pop those in.

00:00:41.480–00:00:59.280 Olivia: It's very wonderful to meet you. If I don't know you, my name is Olivia. I'm the associate course director for the Masters in Surgical Science program and our PG Cert program in Patient Safety and Quality Improvement. And I'm joined here by our panelists. So, we'll go ahead and get started. Matt, do you want to introduce yourself?

00:01:00.800–00:01:07.000 Matt: Hello, welcome all. Matt Gardiner, I'm a consultant plastic surgeon based at Frimley Health and senior research fellow in Oxford.

00:01:08.680–00:01:15.680 Matt: And also, the associate medical director for research and innovation at Frimley and have an interest in AI and innovation within the NHS.

00:01:18.320 Matt: Over to you, Ryan.

00:01:21.280–00:01:39.880 Ryan: We could almost get Matt to do exactly the same because I'm also a plastic surgery consultant but up the road in Buckinghamshire and I'm also the associate medical director for research and innovation for my trust and I lead the innovation work for the Royal College of Surgeons the iHub and also editor-in-chief for Bulletin Journal.

00:01:41.680–00:01:44.160 Matt: So, we have a special guest, Katrina Mason.

00:01:46.560–00:01:55.760 Katrina: Hi everyone. So, my name is Katrina Mason. I am currently a pediatric ENT consultant in London working at the Evelina Children's Hospital which is part of Guy's and St. Thomas's.

00:01:58.000–00:02:10.479 Katrina: I also have a digital innovation role in that I'm the associate medical director and at Euphonia, which is an AI health tech company delivering real-world AI solutions in NHS trusts around the UK.

00:02:13.440–00:02:23.320 Katrina: I've been an NHS clinical entrepreneur and I am a graduate of the Oxford MSc Surgical Science and Practice Program. So, excited to be back today.

00:02:25.720–00:03:00.120 Matt: Excellent. Thank you very much. so, our question is what's next for AI in surgery? And before we kick off, I think it's fair to say that AI in healthcare is mentioned on a daily basis in the news. And even this morning driving in, I heard that the US and UK have agreed to invest heavily in AI within the in the UK. I opened the BMJ on Monday and again there was an article about how AI is transforming radiology. So, it is ubiquitous. It is exciting, but there's lots of lots to discuss.

00:03:03.120–00:03:17.280 Matt: To start with, we're going to move on and Ryan is going to give us a little bit of an overview of what is AI and what are we actually talking about when we mention AI before we go on and think about some of the applications that it's being used for.

00:03:19.640 Ryan: Thanks for that.

00:03:20.920–00:03:26.080 Ryan: So, the other thing to say is we have a set of slides, but actually we would much prefer this to be a conversation.

00:03:28.600–00:04:02.040 Ryan: So, to reiterate what Olivia said at the beginning, if you do have questions as we go along, please put them in the Q&A box. We can see those and we're very happy to have a bit of a mixture of slides and discussion. And the other thing that Matt left out was this morning also the Royal College of Surgeons will release their AI and surgery podcast as part of their data podcast series. So, there you go. It really is excellent and super so, what is AI? The term gets banded about a lot and the reality is it covers a lot of different things and you'll hear kind of LLMs and deep learning and machine learning and AI.

00:04:04.080–00:04:17.160 Ryan: So, I'm sure that for the majority of people on the call this is teaching you to suck eggs, but so if we take a step back, intelligence is just that ability to problem solve. So, or human intelligence that ability to problem solve.

00:04:19.279–00:04:25.760 Ryan: So, artificial intelligence is the ability for computers to solve problems in a way that humans would do.

00:04:27.640–00:04:52.320 Ryan: Machine learning is a subdivision of artificial intelligence and it's where the computer is gains the ability to solve problems, but isn't necessarily these aren't hard coded algorithms that it's just following a set path. It's learned through various techniques in order to identify ways to solve those problems and therefore have intelligence.

00:04:54.200–00:06:04.280 Ryan: And then deep learning is an even smaller field within the realm of machine learning whereby you take the kind of the way the brain is structured and the way our brains kind of process things in a kind of ordered structured way from one layer to the next to the next and they took that concept and put it into the field of computing. So, a program is able to kind of analyze for example a general picture to begin with and in that picture there's a kind of two circles the first layer is able to identify and a triangle below it and a and a kind of small moon shape below that. Then the next layer goes, oh, I see you've got these kinds of these shapes in in this order. Well, if you've got the shapes in this order, this is probably a face. And then you've got well, if you've got those shapes in that order and it's a face and you've got pointy ears and the next layer this may well represent a cat. So, it's kind of structuring your thing. If you take that back into clinical practice, about the kind of looking being able to spot things like tumors on mammograms or CTs or whatever it is, it's the same process, that kind of layer-by-layer decision-making in order to solve the problem that's put to you.

00:06:06.760–00:06:19.800 Ryan: So, that's kind of broadly speaking, and then GenAI, generative AI, is basically the kind of the front end that allows you to in a very kind of naturalistic way access that deep learning, that neural network that's sitting behind it.

00:06:21.440–00:06:27.360 Ryan: So, I hope that makes sense. You know, that kind of a super whistle-stop tour into the different terms that get batted about when it comes to AI.

00:06:29.400–00:06:41.960 Matt: And Ryan, this is not just appeared overnight, has it? It's AI and the application of it within healthcare has been around for quite some time. but it's accelerating a rapid pace, as I think we've all found in our day-to-day life as well.

00:06:44.000–00:07:12.320 Matt: But this is not happened overnight, you know, it's been years of work.

00:06:44.000–00:07:12.320 Ryan: No, yeah, absolutely. Yeah, the decades and decade’s worth of works kind of building slowly. I think really why we're all aware of it overnight, as you say, Matt, is probably that GenAI bit, because prior to the ability for anybody to be able to access the neural networks through a kind of I'm just going to type in, please create me a slide on what is AI.

00:07:13.480–00:07:19.240 Ryan: That this was all a kind of difficult or kind of computer sciences topic that was inaccessible.

00:07:22.600–00:07:38.680 Ryan: But when ChatGPT-3 was probably the first real kind of public-facing generative AI version, suddenly opened up all of this possibility for anybody with a mobile phone or a or a web browser to have access to these incredible technologies.

00:07:40.440–00:07:46.560 Ryan: So, that that's probably the real reason why we're seeing it and think of it as an overnight thing, but you're absolutely right, Matt. This is decades of work.

00:07:47.440 Matt: Okay.

00:07:50.200–00:08:04.880 Matt: So, just to reiterate what said earlier and indeed Olivia, do put some questions in the chat. We're going to ask you also some questions and we'd love to have your feedback. So, I wonder what if people could put some info about their current use of AI. You know, are you using it in your day-to-day practice?

00:08:06.200–00:08:25.480 Matt: I'm an avid user of ChatGPT and other things. And we we'd be really keen to hear about your current experience. So, I think as we move on to the next slide, you'll be aware that there's all kinds of applications for AI at the moment. typically, the technology is often helped to support surgeons' vision.

00:08:28.960–00:08:37.000 Matt: But now it's really moving to the sphere of decision-making and trying to reduce, for instance, cognitive load on surgeons.

00:08:39.400–00:08:46.080 Matt: And it covers a really wide range of different tasks. and I and I don't know whether Ryan, you want to just touch on some of these.

00:08:48.720–00:08:52.320 Matt: I can certainly give our experience at Firmly where it seems to be pervasive now.

00:08:53.200–00:08:59.800 Matt: And also, which we'll come on to later, raising all kinds of ethical and professional queries as to how we deal with it.

00:09:01.000–00:10:59.839 Ryan: Yeah. So, the first thing w ant to credit Umang Patel, Microsoft's CIO in the UK because of some this slide is from him, the really what the   important bit of this slide is when we think about AI in clinical practice, a lot of people the first thing they think of is the kind of the AI doctor or for us the kind of the involvement of AI directly into surgery and navigation or whatever it would be and we'll talk about some of those shortly. But the reality is that that is the thin end of the wedge in some ways of the use cases. That there's a whole load of other use cases in health care that will have or is having a huge effect. And there is a bit of a kind of disconnect between the risk profile and the kind of the readiness. So, those kind of future-gazing the kind of patient-facing AI is further down the line. There's a lot of risk. people are really worried about it and go, "Well, I don't want to worry about AI in health care." but the reality is there's all this other stuff we do that we're applying AI to, be it to support with teaching, to support with appraisals, just making emails quicker or making notes in Teams meetings. None of that is high-risk clinical, but actually if you can attack that and get people used to using AI in a health care setting, you can have a huge impact on the workload, burnout. But also, if you can reduce that burden, you can increase the amount of clinical direct patient care that you can do in a much lower risk setting. So, the point of this slide is to really show you that yes, there is the really cool stuff coming down the line where Dr. AI is going to be super helpful, but let's not forget that when it comes to AI in health care, there's lots of things beforehand that have a much lower risk profile, but are ready to go and are being used already.

00:11:03.720–00:12:22.600 Matt: Yeah, I got I mean Leda has just popped a message in the Q&A, a former student of the course, saying that she's running a pilot on AI and scribe in ED. And I have to say we're from the health board, we've got Epic, a very big electronic health record, and I'm looking forward to the ambient AI that apparently is available through the system and may be available shortly to sit in the background taking notes, creating a letter, even ordering tests based on the conversation that we're having with the patient. but even that in the NHS is behind what many people are doing in other areas of their practice. one may have heard of Heidi Health for instance, and that's already very rapidly delivering support within dictating and assessing. And I think particularly around use of electronic health records, actually it's a big burden. You know, people talk about the sort of the overall burden of feeding these the software with all the clicks and information. And I see this as a great opportunity for reducing the burden on physicians and perhaps freeing up their time to do other things. I always feel that all of my time could be better spent doing something else rather than clicking.

00:12:26.200–00:12:34.520 Matt: At Buckinghamshire, I mean, your experience may be quite different. Do you do you have that? Or you know, there's often a big disparity between the implementation of these technologies.

00:12:35.680–00:12:47.200 Ryan: So, we don't have Epic, which obviously so we have a much simpler health record, but we just kicked off a trial with another company as part of the NHS clear program.

00:12:49.720–00:13:44.920 Ryan: And with so we we've tried to go, okay, what are the what's the problem you need to solve? Because that's always going to be number one. It can't just be that here's a shiny tool. So, we went to our chief medical officer and said, which are the highest admin burden areas in the trust? And then we are applying the we're accepted onto the NHS clear program and we are using 33N as another company to do an ambient AI in rheumatology and children and young adults because these are huge admin heavy specialties and the idea is to see does this really have an impact? Because the whilst these tools are really cool, the we're not unfortunately in a system where really cool counts as a procurement sign-off. So, you have to be able to show that you're delivering value through having these tools and ad and the hope is through these evaluations that's what where we're going to get to.

00:13:47.160–00:14:12.560 Matt: And as a research innovation leader at Bucks, do you feel that the research is keeping pace with the innovation? so, I was interested to see in the last couple of BMJ week from the president of the Royal College of Radiologists who's stepping down saying that actually AI probably was increasing the burden on radiologists and hadn't been delivering the benefits that perhaps have been touted.

00:14:15.120–00:14:31.520 Matt: And much like robotics, the technology and the investment in that technology was very rapid and certainly outpaced the sort of scientific evidence to support implementation but, robotics, a bit like AI in its broader sense, is shiny and cool.

00:14:35.000–00:14:39.160 Matt: And so, that really does push people to try and adopt even perhaps ahead of the evidence to support it.

00:14:40.960–00:15:06.720 Ryan: Yeah, so it's always a bit of a kind of challenge in in what you don't want to do is slow all innovation down to the pace of kind of pure research, but you need to make sure that you're collecting data. So, again, with Bucks and our clear program, we've done a lot of pre-work to understand what's the current admin burden, what's the current time frame for turning things around, what's the current face-to-face clinic time.

00:15:09.320 Ryan: So, you have all that data ready to go.

00:15:12.080–00:16:04.640 Ryan: Then we've kicked off the trial with that data in mind and we've now got 3 to 6 months to start to look at actually what's the impact because the aim is to get to the end of the project with the ability to build a health economic case that either says, yes, this AI tool or this any tool is value-adding, and in which case we need to present this to the procurement board, or actually unfortunately this this isn't really making a dent. Katrina that from a euphoria point of view, and I know we'll cover euphoria much more, kind of how have you guys found that resistance to bringing in technology and having the kind of data, but then also the pilotitis when it comes to AI and yeah, how do you guys navigate that from a commercial point of view?

00:16:07.240–00:16:22.920 Katrina: Thanks Ryan. So just very quickly summarize the technology, we have an autonomous AI-driven clinical conversation agent. So essentially, we replace communication either with administrator or clinician. So, we have a consultation over a telephone. So that's what that's the core of the technology.

00:16:24.920–00:17:10.920 Katrina: And I think sort of one of those points what one of the really key learning points I had when I first started with the company about that value add in that your innovation really has to add value into the institution or the environment that you're bringing it into. And my biggest learning points was that I chased this really sexy deployment of our technology or I thought was the a really useful sexy deployment of our technology into head and neck cancer triage, but the reality was we couldn't actually replace consultations. We you know, we couldn't free up clinicians at that stage. So, we learned very quickly that even if you've got something that's sort of exciting or interesting, unless you're really going to save an institution money you can't even pursue it.

00:17:12.880–00:17:15.119 Katrina: And if you're not going to save an institution money, you're not going to have a buying customer at the end of it.

00:17:17.199–00:17:47.120 Katrina: So, what we have to really learn early on when we have our sort of whether it's pictures or early communications with trusts, conditions, individuals of trying to really understand whether our solution is going to have any real value to them and therefore can we sustain it as a company as it is it going to be viable in order to we going to be able to sell this at the end of the day so otherwise we're going to be like any other start up and we'll just burn.

00:17:49.240 Katrina: So, yeah.

00:17:50.280 Katrina: I've learned by a personal experience.

00:17:53.840–00:18:05.520 Ryan: It is it is a really interesting tension in that. How do you how do you prove the value and yeah. And so, this takes us so, Errol's asked a question about Errol another student of the course.

00:18:08.040–00:18:46.840 Ryan: So, thoughts and methods on validation and safety threshold assessment and completion for mass release public health care use of tools. And I think one of the I think interesting things when it comes to technology not just in health but generally across the board is we hold the threshold for introducing new technology higher than we do for the A the status quo but B kind of if you're introducing a new AI service, we almost demand more of it than if we're introducing a new clinician into the team. And so, I think there is a there is a balance to be had that patient safety is absolutely paramount.

00:18:49.560 Ryan: There's no that is non-negotiable.

00:18:53.440–00:19:18.520 Ryan: However, we need to be realistic about what the tech is what it's potential is now and what it's potential is in the future if we treat it well and train it well and kind of slowly incorporate it more and more into our workflows. So, there is that risk of kind of hyping the stuff that is that we have available now for what its future is going to be like.

00:19:21.040–00:19:34.760 Ryan: Does that make sense in terms of that I think there is this kind of over-high bar we put in saying, well, unless it's 100% perfect, we just don't want it at the moment. Despite the fact that the system we work in is not 100% perfect.

00:19:37.240–00:19:53.040 Matt: Yeah, I think there's certain elements there and you see that. I think for my simplistic point of view, if you look at electric cars and adoption of perhaps of autonomous driven vehicles is that somehow an autonomous vehicle cannot cause death or serious harm to a pedestrian. That's just unacceptable.

00:19:55.400–00:20:20.400 Matt: Yet, everyday humans are doing the same to a far higher degree than an autonomous car is likely to so, I think you're right that the threshold perhaps is different. there is a there is a a good website. I'll try and get the link looking at how digital innovation can be supported in terms of the research elements within the NHS and I'll try and post that in a moment.

00:20:22.400–00:20:52.080 Matt: And I think it does depend on the application and potential impact on patients. So, again, the threshold says for implementing an AI driven decision-making tool that is controlling a robotic surgery may be quite different to some element of administrative support that will help patients manage their appointments or attendance at the hospital and things like that. So, the these are you know, quite different to sort of implementations.

00:20:53.840–00:20:55.520 Matt: We've got another question to add to that.

00:20:56.400 Katrina: Yeah.

00:20:56.640 Ryan: Yeah, yeah, yeah, it's the same as that.

00:20:58.080–00:21:49.120 Matt: As the risk increases as you go to more direct care, absolutely. I mean, my co-pilot use of trying to you know, prioritize my emails yeah, is one thing. but absolutely sort of having high-level conversations with relatives, I think it would be a bit unfair for AI to cater for that at the moment, but that may come. and also, it's the perhaps when we come on to the ethics and personal elements, there are unintended consequences, I think, of implementing some of the technologies. So, if for instance you look at X-ray X- blood diagnostics, I believe there's some AI supported tools for making diagnoses, but the systems were able to diagnose and provide details of the demographics of the patient just by looking at the chest X-ray.

00:21:50.200–00:21:53.000 Matt: And you know, you start to get biases built into the system.

00:21:54.640–00:22:00.840 Matt: For instance, I think in the study black patients were under-diagnosed quite significantly.

00:22:02.480 Matt: And that wasn't known at the outset.

00:22:05.280–00:22:19.800 Matt: But it became apparent as they went on through the study. Certainly, for me health, we have established an AI research committee, which won the NHS Parliamentary Awards 2024. It's one of the first, get that plug in. and we've run some particular imaging projects.

00:22:21.760–00:22:38.480 Matt: And I think it is important that we do continue to develop really high-quality data and research activity to support the implementation. I think there is a there's definitely a danger of getting carried away with the shiny the shiny tool, and just implementing it without fully understanding its impact.

00:22:41.000 Matt: Okay.

00:22:42.280–00:23:04.360 Matt: So, Katrina, I wonder whether you could just dive in a bit more into perhaps as it was sort of come from the course through to your career with Euphonia and I mean obviously AI is built into that. We have AI within the course that we run, the healthcare innovation technology module. but it'd be really fascinating to see how you kind of use some of those transferable skills into your role.

00:23:06.200–00:23:30.320 Katrina: Yeah, absolutely. I think I can say that that Oxford MSc was very pivotal in my future career, my future life. And there was definitely a before the Oxford MSc and a subsequently post Oxford MSc and I've just taken a very sort of diverse career path as a result. So, yeah, I did the MSc in Surgical Science and Practice, which Matt and Ryan run the innovation module.

00:23:31.920–00:23:52.280 Katrina: And the way I viewed it was this was an opportunity for me to acquire the skill sets that I wasn't necessarily going to acquire on the job through my sort of surgical training within my NHS trusts. And I wanted to essentially become a more well-rounded physician, surgeon, clinician overall.

00:23:54.000–00:23:59.440 Katrina: And I had no sort of objective other than and that. So, when I went in, I certainly didn't have a digital health sort of string to my bow at that point.

00:24:02.160–00:24:23.960 Katrina: And it really was the module that I did in the email that was just sort of mind-blowing. It was it just opened my eyes to individuals who are revolutionizing health care. I think one of my favorite lectures was the gentleman who generated organoids, who has just done one of the biggest sells out of an Oxford Innovation scheme.

00:24:25.520–00:24:31.080 Katrina: And I think that was the gentleman who made a machine that essentially keeps organs alive outside of the body from really livers.

00:24:32.680–00:24:36.080 Katrina: And on that module I also met the CEO of Euphonia, who's Nick de Pennington.

00:24:38.640–00:25:00.200 Katrina: He's an incredible individual and I was just sort of on that masters and just thought, "Wow, that technology is really relevant to me in the ENT and head neck care." consulting head neck up at the time, so we obviously had a masters assignment that I had to do, so I asked him how I sort of Matt, do you think could message Nick and see if he'd like to collaborate on a project idea.

00:25:02.160–00:25:10.480 Katrina: And the rest is history as it were. So, I quickly pitched an idea to the CEO of this company who, like most innovators, like, "Yep, brilliant. Let's do it.

00:25:11.920–00:25:20.320 Katrina: Let's get a grant. Go, go, go." and within a few months rapidly, we had like 180 grand from the national grant funding to generate the idea, which was the head and neck risk calculator.

00:25:22.280–00:25:23.200 Katrina: Which I've told you the punchline.

00:25:25.520–00:25:27.400 Katrina: It hasn't gone live in clinical practice.

00:25:29.160–00:25:43.920 Katrina: But, so yeah, so then I continued went back into training. I took a year out to do that. COVID pandemic, went back to training. and since then, I've just grown with the company. I just did a bit sort of ad hoc work here and there, then became a fellow.

00:25:45.520–00:25:47.160 Katrina: And then have sort of become associate medical director.

00:25:49.520–00:26:00.360 Katrina: And what's really nice now is its primarily ophthalmology-based company, looking at the ophthalmology consultations and communications, but we're now re-lighting the fire of ENT head and neck.

00:26:01.680–00:26:11.200 Katrina: And I am also using my sort of clinical acumen and clinical skill set to help provide wrap-around to our technology, which of course is the software as a medical device technology.

00:26:13.320–00:26:18.040 Katrina: But what I think we've realized is that there's limitations in just purely being software as a medical device.

00:26:19.760–00:26:37.400 Katrina: And in order to really add value to the trust that we are in, we actually need to have an element of humans in the loop or human clinician care provided either side. so, we've now actually got CPC validation registration, and that's also the element that I'm helping lead.

00:26:39.600–00:26:52.240 Katrina: Because I think actually at this stage in innovation, we do need to acknowledge there are some limitations in AI technology. that you can't do everything. You can't truly sort of provide the whole service, and that's what we're now trying to do.

00:26:53.480–00:26:56.760 Katrina: So, we do it with we have some clinical clinicians on board that wrap around other side of the product.

00:26:59.000–00:27:57.960 Ryan: Can I Can I pick up on that Katrina that one of the one of the kinds of key principles of the course is not to kind of create the next wave of clinical entrepreneurs that are going to go off and startup companies, which if you do then fantastic. But actually, I think Katrina has hit on something that's absolutely key bit for AI or robotics or any other technology that actually what is required the technology is great, but unless you have a workforce that understands the technology is exactly having the conversations that we're having around on the webinar around risk and implementation and all the other bit and value add you need that workforce that gets it that can be the internal clinical innovators that can help bring this technology to the front line because good technology on its own will not reach the front line without that ability to understand both sides of the equation the clinical and the and the commercial.

00:27:59.000–00:28:22.280 Ryan: Should we get to the next slide because I think that we're just creeping on towards the different areas of surgery and Katrina has absolutely kind of touched on beautifully the where AI is being used in the kind of post-op and big plug for Bucks. We were the very first hospital in in the world to use Ufonia.

00:28:23.440–00:28:40.360 Ryan: But the reality is a bit like when we mentioned AI in healthcare the first thing you think of is the kind of the super cool stuff at patient facing. In a similar way when you talk about AI in surgery a lot of time people think about the operative period, but the reality is that's one part of the one part of the journey.

00:28:44.040–00:29:05.680 Ryan: Matt or Katrina did you want to kind of talk about the kind of pre-op part of.

00:28:44.040–00:29:05.680 Katrina: Yeah, I can talk a bit about sort of my experience. So, with Euphony, the technology that we have called Dora, so she has a conversation with patients. And what we're trying to do is replace consultations throughout that surgical journey or that patient pathway, particularly in cataract surgery.

00:29:07.680–00:29:28.000 Katrina: So, what's really nice is that an AI system can actually have a pre-operative assessment conversation with a patient so that can free up the nurse-led appointment that can highlight patients who are perhaps not suitable to certain operative environments, particularly in cataract surgery. A lot work is getting done in the independent sector where they can't actually safely look after quite complex individuals.

00:29:29.680–00:29:35.720 Katrina: Then you can wrap that around at the end by having a post-operative follow-up consultation again with the AI-generated telephone consultation.

00:29:37.560–00:29:48.160 Katrina: I think what's nice with if you're sort of collaborating with institutions and researchers and surgeons themselves, you can add in extra things. So, do you want to do some PROMs measures? Do you want to get a bit more data as to whether the surgery was successful?

00:29:49.720–00:29:53.680 Katrina: That's easily sort of pigeonholed into that those series of conversations that can be had.

00:29:56.360–00:30:46.400 Matt: So, yeah, that's the sort Yeah, so especially the so, one of Tom Revington, who's one of our panelists as well, director of the courses it was put a message to us saying that he feels the victim of AI and often everything is now moving towards the first point of contact will not be with the human, but will be with some kind of chatbot or other technology that that isn't necessarily human orientated. And I think with Euphony, you know, do people get a choice of speaking to a human or is it called Dora and if that doesn't go well, then potentially on and I and I think there is that danger that suddenly humans are held in reserve somewhere down the line and that that is a great incentive for personal interaction.

00:30:48.520–00:31:12.840 Katrina: So just to speak to that point Matt, to give some technical details, it depends on how it's been commissioned by the trust or the region. but generally, if you're not suitable for an AI telephone conversation, you just have your standard pathway. So, there's no sort of change, you just do your usual come in face-to-face appointments but generally most people we at least attempt a phone call with and then if it's not suitable we can pass that back to the trust according to appropriate SOPs.

00:31:14.960–00:31:28.520 Katrina: I think just coming back to Tom's point, I think at this stage in the sort of technology, I think it's about augmenting what we can do as opposed to truly replacing us.

00:31:30.840–00:31:36.960 Katrina: But I very much have a sort of quite philosophical approach about this. I think we really need to be aware of our limitations as humans.

00:31:38.720–00:31:43.600 Katrina: I have good days and bad days. I can't speak to everyone in every language. I go to sleep at night.

00:31:46.840–00:31:49.480 Katrina: And I've got an IQ that's at a certain level. I can't exceed that.

00:31:51.760–00:31:57.640 Katrina: And I think if you look at Yes, so if you look at ChatGPT-5, now I think three have the IQ of Einstein.

00:32:00.200 Katrina: We cannot compete with that.

00:32:02.720 Katrina: We can't.

00:32:04.160–00:32:21.880 Katrina: And I don't think we should. And I think there's going to be a point where not only is it better from an intelligence understanding perspective, but actually from the subtle nuances of human communication in all you know, empathetic communication, I think it's just going to be better than us.

00:32:23.640–00:32:32.960 Katrina: So, I think we're right to question, challenge, safely adopt, but I think 5 years, 10 years from now this we're going to have been a completely different world.

00:32:35.440 Matt: Mhm.

00:32:35.920 Ryan: And I welcome that.

00:32:38.040–00:32:48.160 Ryan: Well, sorry, I will just very quickly just add that we will pass the inflection point where at the moment if you use AI, there's a kind of clinical liability risk.

00:32:49.400–00:32:55.240 Ryan: We will hit that inflection point where if you don't use AI, there'll be a clinical implication risk.

00:32:57.200–00:33:21.720 Matt: Well, so, a simple practical application that we've been using is to generate patient information. So, kind of talking to your point there, Katrina, where you know, you're in a busy surgical clinic and actually I've used ChatGPT to generate really good high-quality patient information. You can sort out your prompt and with the prompt you can get it exactly the right language level that you want.

00:33:23.440–00:33:25.840 Matt: You could put in pushing capabilities and it'll try to adjust the information.

00:33:27.840–00:33:46.920 Matt: But even having done all of that, it doesn't really meet the sort of the top level of patient information generation that's accepted. so, people may be familiar with the patient information forum that provides guidance on how to generate good patient information. Because patients haven't fed back on that information.

00:33:49.520–00:33:55.360 Matt: And so, to make it really top-notch, actually need to go through this process of engaging patients, getting feedback, and everything else.

00:33:57.400–00:33:58.240 Matt: And it's Yeah.

00:33:58.920–00:34:34.560 Matt: What How do you call that? Because actually, is it better to give more people better quality information that it might not be absolutely perfect or slow down the process and sort of get it badged, but take forever generating it and its sort of then out of date? So, I think that's partly going back to the tensions around getting on with it versus proper evaluation and you know, correct implementation, which will delay things, I think. and that perhaps will deny people whether it's patient information or better care for instance.

00:34:36.360–00:35:10.520 Ryan: And that do you think there is because a bit like it's really easy to be a bad doctor and it's really difficult to be a really good doctor. And it's really easy to put a prompt into GenAI to create a patient leaflet. But actually, if you spend time and you really understand the kind of nuances of prompt engineering, which is why that that there are now jobs which are that that the job title is kind of chief prompt engineer, that you can get even with where we are at the moment you can get to really high-quality output, but you've got to understand how to use it.

00:35:11.800–00:35:46.720 Ryan: And to I think the point Katrina made earlier that 5-10 years down the line, so if we think about med students that start have started medical now will qualify in 2030 and they'll be consultants in 2040. This technology is only going to get bigger and bigger and so we need to make sure that we are training our next generations to understand the technology and the and the nuances to make sure they do produce the good patient information leaflets. I realize that's not a big thingy to jump over, but do you know what I mean?

00:35:47.600–00:35:51.000 Ryan: Yeah, it's hard to do it well and we need to be training people to do it well.

00:35:52.440 Matt: Yes, I know completely agree with that.

00:35:54.320 Matt: Yeah.

00:35:55.600 Matt: Katrina.

00:35:57.520–00:36:02.520 Katrina: Yeah, I thought you just volunteering yourself to write the next surgical curriculum for Oxford now as an AI arm.

00:36:05.800 Katrina: But we have to think this way.

00:36:08.320–00:36:18.720 Katrina: We can't resist this. It is coming. It's here already. It's I it's terrifyingly quick how quickly it's iterated and developed. I haven't seen anything like this in my lifetime.

00:36:20.200–00:36:24.520 Katrina: It's only going to get smarter and clever and yeah, I think an understanding of that.

00:36:26.320–00:36:57.840 Katrina: To match your point that you were talking about generating sort of patient information leaflets like on the cuff using chat GPT. With the new GPT-5 that came out, Sam Altman has very actively sort of lauded the health benefits of his app. And even though it's not a medical device, it's not intended for health care use so outside of regulatory approvals, they're actually very clearly advertising this and marketing it as almost like an adjunct for the patient. So, it's like the champion of the patient. It's allowing patients to digest information.

00:37:00.520–00:37:08.960 Katrina: What questions should I be asking? How can I interpret these results? So, it is actually really interestingly being proposed for and used for a health care setting.

00:37:10.200–00:37:28.200 Matt: Well, I think that that's fascinating and that's where in increasingly medical so professional control of information has been lost. I mean, of course there's lots of advice available on the web. People have always gone to in the last you know, 10, 15 years to seek advice.

00:37:29.320 Matt: But I'll give you one recent example.

00:37:31.160–00:37:52.200 Matt: I was researching and we had a trial comparing which pair of flexor tendons in the finger, either one tendon or both tendons. And I was trying to recruit a patient to the study and she read my patient information leaflet, said it was an excellent trial, but I've asked chat GPT and it says that I should have both tendons repaired and therefore I'm out. and I and that took me back.

00:37:56.640–00:38:13.120 Matt: Patients have always you know, sought advice from maybe their relative or perhaps they have done some searching online, but I think chat GPT and others are held at a much higher level that somehow they're superhuman and they're not it is superhuman as knowledge and it is a source of truth.

00:38:14.560 Matt: And so, I have concerns around that.

00:38:18.440–00:38:23.600 Matt: And that's where you know, trust in the resources and perhaps some curation of this is really important.

00:38:25.040–00:38:44.440 Matt: I'll give you another example. Say we're doing patient information using ChatGPT, but for some reason it always misses out complex regional pain syndrome. whenever I ask it to deliver the risks of hand surgery, which is uncommon, but nonetheless something that really does need to be in a in a consent form. But it too I do think sometimes it's not quite there.

00:38:46.000–00:38:49.800 Matt: And that's where I think some curation of it at the moment is important.

00:38:51.520–00:38:55.280 Matt: Okay, should we should we move on? we've got a couple of questions coming through.

00:38:56.840–00:38:59.960 Matt: Perhaps if we go on to the next slide.

00:39:01.280–00:39:14.120 Matt: Chee Sum Eng, I see thanks for your message on the use of AI for robotic surgery. Do you have any thoughts on whether particular types of more repeatable surgeries might be suitable for first application of this technology and what some of the barriers to adoption might be?

00:39:16.800–00:39:27.360 Matt: Right, good question. My feeling with robotics is that it it's sort of augmenting the experience of a human surgeon doing the robotic surgery.

00:39:29.000–00:39:32.960 Matt: It for instance is perhaps improving the movements.

00:39:34.440–00:39:36.360 Matt: It's perhaps giving some augmented view of the field of surgery.

00:39:39.120–00:40:03.800 Matt: Perhaps even identifying tumour margins and things like that. in terms of repeatable surgeries, I think the sort of scientific evidence to the application of robotics within surgery is still catching up and it's got quite some way to go to really clearly demonstrate the benefits of robotics. So, time will tell whether that's the case. I think the regregulatry yeah.

00:40:04.200 Matt: Yeah.

00:40:04.640–00:41:19.200 Ryan: I was just going to add very quickly onto that. So, if you look at the work of Johns Hopkins with their automated robotic surgery, so they're they've got their robots to be to do suturing and recently they did a cholecystectomy on a pig using GenAI prompting. So, there are bits that are coming. I think there's a couple of bits of technology that still have to happen for anything to really even approach that and computer vision is absolutely key and especially moving targets. So, a lot of the really cool navigation stuff that you see where you have a CT kind of anchored onto the patient you wear your headset and as you move through the patient you can see at the moment you need physical anchors to lock that image in place. And so, to your question about what are the potential areas? Well, when you've got areas that have got predominantly bone, so you can see there could be a role in the in the not that distant future where everything's locked in place, you the CTs in place and you can align a prosthesis. Apologies if there's any orthopedic surgeons in the audience, but that's only where I could see the that role for accuracy and standardization.

00:41:24.280–00:41:38.200 Ryan: But I think the reality is as Matt said, I think we're still a fair way off that and there's quite a few steps that have to happen and quite a few more practical things that will happen first before we're really into the point where we're looking at autonomous robotic surgery.

00:41:40.360–00:41:54.080 Matt: This is leading to all kinds of new words and terminologies. I came across Gaussian splatting. Have you come across that? But that's an interesting technology that's improving the quality of the sort of 3D rendering I think of these sorts of overlays.

00:41:56.080–00:42:07.520 Matt: Okay, so I think we've covered off a bit of this. automating documentation, we've talked a bit about ambient use of ambient AI preoperatively saying your consultations.

00:42:09.520–00:42:36.800 Matt: One thing perhaps we could just touch on is surgeon well-being and whether we think that AI has a role there. Brian as editor of the bulletin and I know that you've run a series of articles around sort of the psychological elements of surgery and supporting surgeon well-being. I've heard of reports of people using ChatGPT for instance to sort of counsel them and help support there you use it.

00:42:38.040–00:42:39.120 Matt: I've not tried that. I've had a few conversations.

00:42:40.480–00:42:47.760 Matt: My sons had a few interesting conversations with ChatGPT. You can might have to stop that. But I wonder whether you could talk to that a little bit.

00:42:49.120–00:44:51.240 Ryan: Yeah, so I think there's probably a few bits of this. So, a bit like everything in AI, the data in kind of the quality of the data in is what derives the data out and I've been using it pretty solidly every single day for the last however many years it is now. Is it coming up to 3 years since GPT 3 came out? So again, if you if you craft your prompts well enough and I'm not saying I'm a kind of expert prompt engineer, but I kind of saw something somebody had posted about a particular prompt that really dove deep. Dug deep. Let's go with dug deep. Because it has all the data put into it. So, it was like kind of all of everything you know about me, what are the kind of what are my strengths, what are my weaknesses, but really but literally it says, but don't stay at the superficial kind of go down and look at each of those domains and identify ways I could be and the output was slightly intimidating. But you can imagine that that in surgery that because we have this whole thing about videoing video logbooks are a thing and whether we should be doing that to really prove the quality of training. So, moving away from indicative numbers, but the reality is the three of us as trainers, we're not going to sit and watch 3 hours per operation times 10x for a single trainee. But you could put all of that into an AI model and say, "Critique this video and find the areas that this particular trainee does well, find the areas that they need to work on and suggest some exercises or suggest some things that may help with those particular areas." So, I think there's a huge set of potential we haven't even uncovered in terms of where that would go, but kind of taking a step back in terms of well-being. So, yeah, you have the kind of AI counslor, which is fine and slightly gimmicky, but is interesting.

00:44:52.720–00:45:12.480 Ryan: But there is just that kind of taking the workload off of us of the kind of repetitive monotonous tasks that that bring us down, increase that risk of burnout, and reduce staff retention. If you can take some of that stuff, actually, there's a huge amount of kind of morale-based bonuses that you can gain.

00:45:15.240–00:45:20.760 Matt: Yeah, which is perhaps not available day-to-day just because of, you know, the resource implications for the hospital training deanery, etc.

00:45:24.400–00:45:45.440 Matt: Okay, should we move on then? I realize actually time is flying by. We've been nattering away. we've got a few other things just to talk about, think about the future actually. we've talked a lot about the sort of current implementation. but, we probably ought to move on to the future and also perhaps some of the ethical and professional issues that people have raised in the questions as well. if Olivia could move us on.

00:45:50.760–00:46:03.200 Matt: So, one of the questions around robotics was, you know, what are the quick wins or where is AI going to support us at the moment and perhaps with the adoptions in the future.

00:46:05.760–00:46:13.680 Matt: We've talked we've touched upon briefly sort of the imperfect data and I think Ron, this is your slide in terms of garbage in, garbage out.

00:46:15.360–00:46:18.000 Matt: That's an issue and we've talked a bit about the biases that might become inherent in some of these systems.

00:46:20.000–00:46:21.000 Matt: I wonder whether it's just worth covering that off in a bit more detail.

00:46:22.480–00:46:50.400 Ryan: Yeah, yeah. So, I think a couple of things to this slide. if anybody, especially those on the course, would have heard this and everyone has heard me talk about AI elsewhere. The cake there is the analogy of what we see or what we think about it in terms of AI and healthcare is a bit like the most beautiful cake with wonderful icing and that icing represents the AI that has it just looks so good. But as soon as you cut into the cake, the whole thing collapses because the sponge is undercooked and underprepared.

00:46:52.360–00:47:48.320 Ryan: And in some ways that's where we are with NHS and healthcare data. Our data isn't yet strong enough to really get the benefits out of AI. And that has implications. It means that what the kind of insights we're asking it to deliver are on imperfect data. And the problem with AI and certainly gen AI is it's a bit like Hermione Granger. so, it is the perfect teacher's pet. If you ask it a question, it will answer that question. Whether it understand or not whether it has the data to back up its answer, but it's programmed to go, "You've asked a question, I must answer it." and it doesn't There's no way yet of understanding where those hallucinations are and how kind of subtle. And I know Matt, you gave some great examples of hallucinations, but it wants to help. It wants to give you an answer. And in doing so, it amplifies bad data.

00:47:51.000 Ryan: It fills in gaps in incomplete data.

00:47:53.960–00:48:19.880 Ryan: And one of the issues we've got to be really careful of in the future, is as it trains on new data, a huge proportion of that data is AI-generated data. So, there is a risk that we start to get these kinds of loops that the term nobody's using yet. I'm going to I'm going to I'm coining this, so if you hear it in the future, you heard it here first, of echo chambers, as in AI echo chambers, where the thing is just building up on its own bad outputs.

00:48:23.840–00:49:05.680 Matt: Yeah, I think we've all had the experience of asking ChatGPT something, it gives you a completely wrong answer, and you point that out to it, and it goes, "Oh, I'm so sorry. Yes, you're absolutely right. Yes, it's that, isn't it?" Like, not fast, not by the skin of a duck's back, water off a duck's back, you know, carry on. And you just wonder how many things you are wrong that you don't pick up. Okay, Barry asks in the questions, do you think it will evolve primarily as a decision support system that might enhance surgeon's intuition skills, or do we think it may take over a significant part of interoperative decision-making independently? Well, I think there's only one direction that's going to balance, and that is the latter.

00:49:06.800–00:49:31.680 Matt: You know, for instance, it might be that you're doing robotic surgery, and it will determine whether or not you've completely resected a tumour, because it will be supporting your vision and understanding of where the tumour is. And ultimately, it'll come down to that that technology to make that decision, rather than the surgeon deciding whether or not to remove a bit more tissue or not. So, I think yeah, it that's going to happen.

00:49:34.080 Matt: I'm almost certain of that.

00:49:36.360–00:49:44.000 Ryan: Yeah, there's already some great work by Professor Cahill in Ireland, who because you think it's just data. Can't stress it enough, this is all just data.

00:49:45.320–00:50:21.080 Ryan: So, he's looked at bowel resections, kind of loads and loads of videos, and looked at where bowel resections have been made relative to a tumour and for future anastomosis some by expert surgeons or senior surgeons and more junior surgeons and then he shows this amazing heat map of where those incisions are made which can then be overlaid on the image. So, when you've now got your new patient in front of you it's using all that data to heat map where an expert surgeon or senior surgeon would do it and where a more junior surgeon would do it.

00:50:23.760 Ryan: And all of that is building that future.

00:50:25.920–00:50:34.440 Ryan: Well, where's that going? Yes, it's giving a bit of kind of navigation to the surgeon now, but in 5 10 years is that still the point of just giving navigational input?

00:50:37.960–00:50:53.960 Ryan: This technology is just going to grow and grow and unless as clinicians we are there and understand it and can help shape it, this is coming regardless of whether we're there or not and we need to make sure we do our bit to kind of guide its real-world implementation. So going back to that importance of the clinical innovator.

00:50:56.600 Matt: Yeah.

00:50:57.720–00:51:06.480 Matt: And I think Errol makes a good point here that we've talked about robotic surgery and the benefits there, but actually open surgery may well benefit from this as well.

00:51:08.200–00:51:15.280 Matt: I know you've got a fancy pair of meta-Ray-Bans Brian. Katrina, have you joined the cool club? I'm not there yet.

00:51:16.480–00:51:17.400 Katrina: Needs to be painted though.

00:51:20.280–00:51:26.240 Matt: But you know, I think absolutely you could envisage wearing some sort of glasses or headset.

00:51:28.400–00:51:36.280 Matt: You might be able to have along the lines of proxy some sort of proctoring or someone be able to dial in and provide support whilst you're operating.

00:51:38.160–00:51:46.480 Matt: You could effectively phone a friend or phone an AI assistant to get some advice on what you might have to do next in terms of surgical technique or decision-making.

00:51:49.520–00:51:59.960 Matt: So, I think that sort of technological support is kind of here, but not obviously widely adopted. but the technology's all there, I think, isn't it?

00:52:02.680–00:52:11.160 Ryan: Yeah. And it is those kind of risk barriers that are kind of holding it back, rightly so, from healthcare, but those will be crossed in time.

00:52:13.560 Matt: Yeah.

00:52:15.000–00:52:19.400 Ryan: Yeah, Grant asked with overconfidence in AI could make us less skilled in less skilled in identifying hallucinations.

00:52:21.840 Ryan: And this is a really good question.

00:52:23.160–00:53:14.280 Ryan: There's a paper, is it from Stanford, that have looked at people that use gen AI more regularly are becoming stupid or less intelligent. So, I think there is an element of that. I think the other side of that that question, like so we we've been using skin analytics for AI detection of skin cancers in our trust, and it's had a fantastic im- impact. One of the pushbacks is, "Oh, it's going to de-skill the dermatologists and plastic surgeons in using dermoscopy." Well, possibly, but that's because dermoscopy is in time going to become an outdated skill. So, you don't need to use it because you'll have the tools as long as the tools we have enough confidence in them, in the same way that we don't stick our finger into urine to taste whether it's still sweet.

00:53:16.040–00:53:21.520 Ryan: And that's a skill that luckily, we're all de-skilled in, and we don't need it anymore because tools moved on. And I think in many ways this is similar.

00:53:23.880–00:53:49.880 Matt: It is important, though, for instance, in robotics, I believe with, you know, robotic assistant radical prostatectomy, for instance, the surgeons are less exposed to doing it open or if they have a major complication, converting from a robotic-assisted procedure to a open procedure may be more challenging and actually the trainees coming through that program now may be less familiar.

00:53:52.160–00:53:58.480 Matt: I suppose the same could be said for laparoscopic surgeon surgery and having to open an abdomen but perhaps to slightly lesser degree.

00:54:00.120–00:54:07.520 Matt: So, I think they're legitimate concerns aren't they around the loss of you know, human skill set because these technologies are taking over.

00:54:10.000–00:54:25.080 Matt: But I'm sure this is repeated throughout Australia in terms of technologies coming supplanting what was done previously and actually freeing up humans to do other things or contributing in different ways that we don't currently have the time to do.

00:54:27.280–00:54:28.960 Matt: So, times up actually. We're actually five minutes away.

00:54:31.040–00:54:32.320 Matt: And I wonder whether we just finish We've had a good number of questions.

00:54:33.480–00:54:37.680 Matt: Please do put any final questions in the chat or the Q&A box rather.

00:54:39.680–00:54:45.440 Matt: Perhaps we just finish on thinking about some of the ethical and professional issues. We have covered off some of them.

00:54:46.880–00:55:01.200 Matt: One thing I wanted to just touch on was patients in all of this of course. We've talked about patients but I I'm not sure what my patients think about this technology. I don't know what they think about me using ChatGPT to develop some of the patient information for instance.

00:55:03.440–00:55:05.000 Matt: Katrina Ryan, do you have experience of this? Have you had pushback?

00:55:06.080–00:55:21:00 Matt: Particularly Katrina, with your phone use. Is this a technology that discriminates against those who are not tech-savvy or less engaged in sort of?

00:55:21:00-0056:32.00 Katrina: So patient public engagement has been very much at the core of our technology co-creation. Well, I think what's really nice that is that even though it's like an AI solution, the actual interface is speech communication. So generally, it's fairly accessible for most people. But we have done a lot of work on certain subsections of society into whether this technology is acceptable and usable by them. So, non-English first language speakers, we've got an NIHR trial on that at the moment Moorfields. we've listened to cancer patients; we've listened to people with learning disabilities. So, we have actively sort sought out perhaps patients where you would naturally think that AI doesn't naturally sort of fit in with them, but I think it is a really important thing to say that any technology that goes into a patient's hands has to have PPI involvement from the outset. And what what's quite nice is with a lot of regulatory processes now for software as a medical device, that's almost like integrally built into that process, particularly in FDA in America and more so in the UK and the EU.

00:56:34.200 Katrina: Yeah, welcome it, advocate for it.

00:56:37.600 Katrina: Has to be done.

00:56:38.720–00:57:02.040 Ryan: And I think I think the fact that gen AI is so ubiquitous in so much of the kind of life. I now use it to create bedtime stories for my kids. Like it it it's just across the board. So, the idea of someone coming in and you say, "Oh, well, we're using generative AI for whatever." They're Oh, yeah, well, of course you are. I used it to work out what to cook based on what was in my fridge the other day. So, that makes perfect sense.

00:57:05.680–00:57:08.760 Matt: Well, it I've not got to that point of getting gen AI to create stories yet.

00:57:11.000 Matt: I'm still proper parenting, right?

00:57:14.680–00:57:16.600 Ryan: I'm an augmented I'm an augmented parent now.

00:57:16.840 Ryan: I'm an augmented parent.

00:57:17.720–00:57:29.080 Matt: Okay. So, as we come to close, the I mean, finally, surgeons must remain in control. Well, do you think we can? I mean, is it already slipping from our fingers? Are we going to ultimately be subservient?

00:57:31.680–00:57:35.360 Matt: Are we going to be just referral to all of this or do you think we can maintain control?

00:57:39.600–00:57:42.400 Katrina: So, I would argue that we shouldn't be in control if there's something that's better than us.

00:57:43.720–00:58:07.400 Katrina: We may not be there yet, but I think we will be there. And I mean, surgeons are the most egotistical, arrogant subsection of healthcare, and I think we need to be the first to sort of lead by example because if there's something better than us, we can't advocate against it. We have to we have to utilize that. Yeah. I don't know what your thoughts are.

00:58:08.240–00:58:11.520 Ryan: Katrina, I can recommend a good counslor for you if you if you require it.

00:58:12.560–00:58:28.800 Ryan: So, I I would actually I would agree disagreeably and say that I think that we will stay in control at least for our careers and hopefully next year, but the role of the surgeon and the doctor in general will change.

00:58:30.880–00:58:40.480 Ryan: And I my argument is that I don't think AI is going to replace us, but those that can use it and do use it effectively and understand it will replace those that don't use it.

00:58:43.600 Matt: Yeah.

00:58:44.560–00:59:16.160 Matt: And overall, our feeling is that it will drive improvements in patient care and ultimately bene- benefit human race in terms of access to surgery. Now, don't forget that you know, billions of people don't have access to high-quality surgery. So, whilst I think probably our discussions have focused on high-income countries, our experience in the UK, for instance, actually there's huge potential to improve the quality of surgery and access to surgery across the globe. So, perhaps at that point, we will draw to a close.

00:59:17.080–00:59:29.320 Olivia: Wonderful. And I'm now rethinking all of my credentials. So, thank you for a little bit of the of the shock there, but it was really wonderful to see everyone today, and I recognize that we could spend my goodness just weeks and weeks on this.

00:59:31.600–00:59:55.480 Olivia: Lucky for you, this is just sort of a a quick teaser of some of the information that is available in our courses. So, I mentioned Matt and Ryan teach on our healthcare innovation and technology module, which you can find in any of these areas, the Masters in Surgical Science and Practice, the PG Cert in Patient Safety and Quality Improvement, or any of the Surgical Science and Practice Short Courses if you wanted to just take this one as a standalone module.

00:59:57.600–01:00:21.120 Olivia: If you want to stay in touch for any of these key updates and hear about some of these courses that are coming up, or if you're interested in joining us for future Oxford Open Grand Rounds, please do, you know, sign up to our mailing list. You can find us here on the QR code. you can find us via email. You can search for us on Google, wherever you get your information about Oxford, we will be there. And with that, thank you so much for joining us today.