WEBVTT 00:00:00.000 --> 00:00:03.259 Imagine a future, perhaps not too far off, where 00:00:03.259 --> 00:00:05.780 orthopedic surgery, its precision, isn't just 00:00:05.780 --> 00:00:08.880 about human skill anymore, but err, also about 00:00:08.880 --> 00:00:10.980 artificial intelligence predicting complications 00:00:10.980 --> 00:00:14.080 before they even happen. Or picture this, an 00:00:14.080 --> 00:00:16.539 athlete's career -threatening injury being forecast, 00:00:16.620 --> 00:00:18.600 you know, actively prevented rather than just 00:00:18.600 --> 00:00:20.719 treated after the fact. Now, musculoskeletal 00:00:20.719 --> 00:00:24.219 conditions affect a staggering 1 .71 billion 00:00:24.219 --> 00:00:26.879 people globally. That puts immense strain on 00:00:26.879 --> 00:00:28.820 healthcare systems. So this isn't just while 00:00:28.820 --> 00:00:31.460 futuristic thinking, it's profound shift that's 00:00:31.460 --> 00:00:33.740 actually already underway. Today we're taking 00:00:33.740 --> 00:00:36.000 a deep dive into how artificial intelligence 00:00:36.000 --> 00:00:38.719 is truly revolutionizing orthopedics and sports 00:00:38.719 --> 00:00:41.140 medicine. How it's fundamentally shifting our 00:00:41.140 --> 00:00:43.770 approach from, say, Reactive Treatment to Proactive 00:00:43.770 --> 00:00:46.130 Personalized Prevention. Welcome to the Deep 00:00:46.130 --> 00:00:48.310 Dive. This is the show where we unpack complex 00:00:48.310 --> 00:00:50.289 topics, sift through the latest research, and 00:00:50.289 --> 00:00:52.070 really try to extract the crucial insights from 00:00:52.070 --> 00:00:54.429 all sorts of sources, all for you, our curious 00:00:54.429 --> 00:00:56.289 listener. And joining us for this fascinating 00:00:56.289 --> 00:00:59.030 deep dive is a true luminary in the field, Professor 00:00:59.030 --> 00:01:01.329 Moet Meemem, as a consultant trauma and orthopedic 00:01:01.329 --> 00:01:03.929 surgeon and a key contributor to significant 00:01:03.929 --> 00:01:07.489 policy work on AI in orthopedics, including the 00:01:07.489 --> 00:01:10.409 Orthopedic Research UK, that's O -R -U -K, policy 00:01:10.409 --> 00:01:13.060 paper. He brings this unique blend of clinical 00:01:13.060 --> 00:01:15.599 skill and, well, forward -thinking tech insight 00:01:15.599 --> 00:01:17.819 to our chat. He's here to help us get to grips 00:01:17.819 --> 00:01:19.879 with what's truly important in this, let's face 00:01:19.879 --> 00:01:22.819 it, rapidly evolving landscape. Right, let's 00:01:22.819 --> 00:01:24.900 kick things off with a few rapid -fire questions 00:01:24.900 --> 00:01:27.319 just to set the stage for our deep dive today. 00:01:27.739 --> 00:01:30.400 First question for you is this. What's the fundamental 00:01:30.400 --> 00:01:32.879 shift AI brings to medical practice, particularly 00:01:32.879 --> 00:01:35.019 in orthopedics? Is it more than just simple automation? 00:01:35.140 --> 00:01:37.159 Are we talking artificial intelligence or perhaps 00:01:37.159 --> 00:01:39.640 something more like augmented? That's a really 00:01:39.640 --> 00:01:42.200 excellent way to and yes, it's a crucial distinction. 00:01:42.680 --> 00:01:45.120 We're absolutely talking about augmented intelligence. 00:01:46.019 --> 00:01:48.500 The core shift isn't really about AI replacing 00:01:48.500 --> 00:01:51.280 skilled people. It's about amplifying their capabilities. 00:01:52.420 --> 00:01:54.659 Think of AI as an incredibly powerful co -pilot, 00:01:54.680 --> 00:01:57.840 maybe? A tool that performs tasks needing human 00:01:57.840 --> 00:02:00.340 intelligence, you know, learning, reasoning, 00:02:00.540 --> 00:02:03.120 spotting patterns, improving over time, but it 00:02:03.120 --> 00:02:05.900 does it at a scale and speed humans just can't 00:02:05.900 --> 00:02:08.680 match. It can sift through these huge complex 00:02:08.680 --> 00:02:11.300 data sets in moments. That lets clinicians make 00:02:11.300 --> 00:02:14.180 more informed decisions, boosts diagnostic accuracy, 00:02:14.439 --> 00:02:16.120 streamlines workflows. We've seen this quite 00:02:16.120 --> 00:02:18.139 clearly in studies, actually. Take mammography, 00:02:18.180 --> 00:02:20.400 for instance. Physicians working with AI achieve 00:02:20.400 --> 00:02:22.639 better diagnostic outcomes than either the AI 00:02:22.639 --> 00:02:25.599 or the physician working alone. So yes, it's 00:02:25.599 --> 00:02:27.439 definitely about making the human expert even 00:02:27.439 --> 00:02:30.039 better, not about replacing them. The idea of 00:02:30.039 --> 00:02:32.199 amplification, it's really powerful, isn't it? 00:02:32.259 --> 00:02:33.979 Building on that, then, from your perspective, 00:02:34.479 --> 00:02:36.960 where are we seeing the most surprising, perhaps, 00:02:37.180 --> 00:02:40.699 and impactful early wings for AI, either in orthopedic 00:02:40.699 --> 00:02:43.400 patient care or, say, sports injury prediction? 00:02:43.719 --> 00:02:45.719 What's genuinely getting you excited right now? 00:02:45.960 --> 00:02:47.680 Well, the early ones are genuinely impactful, 00:02:47.860 --> 00:02:50.199 yes, and often quite surprising in just how precise 00:02:50.199 --> 00:02:54.150 they can be. In orthopedics, The diagnostic capabilities 00:02:54.150 --> 00:02:57.289 powered by AI are really quite striking. Imagine 00:02:57.289 --> 00:02:59.629 AI algorithms, especially the deep learning ones, 00:03:00.050 --> 00:03:01.930 assessing conditions like osteoarthritis from 00:03:01.930 --> 00:03:04.469 x -rays with an aptitude that's comparable to, 00:03:04.530 --> 00:03:06.889 or sometimes even better than, a doctor with 00:03:06.889 --> 00:03:09.629 a decade's experience. That means faster, more 00:03:09.629 --> 00:03:12.430 consistent diagnoses for patients. Beyond that, 00:03:12.639 --> 00:03:14.939 we're seeing real strides in predicting patient 00:03:14.939 --> 00:03:17.819 outcomes. Identify who is at high risk for complications 00:03:17.819 --> 00:03:21.919 like parapathetic joint infection, or even predicting 00:03:21.919 --> 00:03:24.180 which patients might feel dissatisfied after 00:03:24.180 --> 00:03:26.599 their surgery. This allows for proactive intervention, 00:03:26.740 --> 00:03:28.280 doesn't it? Not just reacting after something 00:03:28.280 --> 00:03:30.879 goes wrong. And in sports, while that shift towards 00:03:30.879 --> 00:03:33.439 proactive injury prevention is frankly a game 00:03:33.439 --> 00:03:35.539 changer, AI acts like an early warning system. 00:03:35.860 --> 00:03:38.340 It analyzes these complex multidimensional data 00:03:38.340 --> 00:03:41.659 sets. wearables, biomechanics to predict muscle 00:03:41.659 --> 00:03:44.139 injuries, manage athlete workloads. You mentioned 00:03:44.139 --> 00:03:45.659 Health First earlier, they've deployed what, 00:03:45.840 --> 00:03:48.139 17 predictive models using AI? Generated nearly 00:03:48.139 --> 00:03:50.379 a thousand custom machine learning features to 00:03:50.379 --> 00:03:53.240 improve health outcomes. That really shows its 00:03:53.240 --> 00:03:56.550 power. That is remarkable. So, OK, if AI holds 00:03:56.550 --> 00:03:58.949 all this promise, what's the single biggest hurdle? 00:03:59.129 --> 00:04:01.590 What's stopping its widespread adoption and, 00:04:01.590 --> 00:04:03.990 you know, realizing its full potential in this 00:04:03.990 --> 00:04:06.150 field today? Without a doubt, it really boils 00:04:06.150 --> 00:04:09.389 down to data, specifically its quality, its diversity 00:04:09.389 --> 00:04:13.110 and crucially, its accessibility. Look, AI models 00:04:13.110 --> 00:04:15.789 are only as good, only as robust as the data 00:04:15.789 --> 00:04:18.389 they're trained on. That's fundamental. And the 00:04:18.389 --> 00:04:21.529 reality is much of our health care data is still 00:04:21.800 --> 00:04:24.439 Well, it's uncaptured or unstructured. It's varied, 00:04:24.439 --> 00:04:26.759 stuck in different silos across different systems. 00:04:27.160 --> 00:04:29.259 This makes it incredibly expensive and time -consuming 00:04:29.259 --> 00:04:31.680 to clean it up, standardize it, integrate it 00:04:31.680 --> 00:04:34.279 into formats that AI can actually use. And often, 00:04:34.720 --> 00:04:36.519 we lack that really comprehensive data, don't 00:04:36.519 --> 00:04:38.699 we? Especially the contextual stuff, social factors, 00:04:38.980 --> 00:04:41.259 lifestyle choices, daily activities, all vital 00:04:41.259 --> 00:04:43.970 for accurate prediction, but often missing. On 00:04:43.970 --> 00:04:45.790 top of that, you've got the sensitivity of patient 00:04:45.790 --> 00:04:48.470 data. It's highly regulated and rightly so for 00:04:48.470 --> 00:04:50.790 privacy, but that does create significant barriers 00:04:50.790 --> 00:04:53.069 for AI model training and deployment. That's 00:04:53.069 --> 00:04:55.050 why initiatives to standardize data collection 00:04:55.050 --> 00:04:58.230 and ensure secure ethical access are just so 00:04:58.230 --> 00:05:00.490 critical. That paints a very clear picture of 00:05:00.490 --> 00:05:02.850 that foundational challenge. Okay, now let's 00:05:02.850 --> 00:05:05.769 unpack this core concept a bit more. You mentioned 00:05:05.769 --> 00:05:09.100 augmented intelligence rather than just... artificial. 00:05:09.199 --> 00:05:11.399 What does that actually mean for how clinicians 00:05:11.399 --> 00:05:13.500 and sports scientists will work day to day? It 00:05:13.500 --> 00:05:15.600 means a pretty profound transformation in how 00:05:15.600 --> 00:05:18.360 we interact with technology, I think. At its 00:05:18.360 --> 00:05:20.839 heart, artificial intelligence, AI, it refers 00:05:20.839 --> 00:05:23.180 to computational techniques enabling machines 00:05:23.180 --> 00:05:25.860 to do tasks that usually need human intelligence. 00:05:26.319 --> 00:05:29.279 Things like learning, reasoning, solving problems, 00:05:29.600 --> 00:05:32.120 spotting patterns, and crucially, the ability 00:05:32.120 --> 00:05:34.180 to get better iteratively as they get more data. 00:05:34.360 --> 00:05:37.180 Within AI, we've got these specific powerful 00:05:37.180 --> 00:05:39.620 tools. There's machine learning, or ML, where 00:05:39.620 --> 00:05:41.939 algorithms learn from data without being explicitly 00:05:41.939 --> 00:05:44.560 programmed for every single scenario. Then deep 00:05:44.560 --> 00:05:46.959 learning, DL, which uses these layered neural 00:05:46.959 --> 00:05:49.519 networks to process huge amounts of complex data. 00:05:49.879 --> 00:05:51.879 It's particularly good at image and pattern recognition. 00:05:52.560 --> 00:05:55.199 And natural language processing, NLP, which helps 00:05:55.199 --> 00:05:57.620 computers understand, interpret, and even generate 00:05:57.620 --> 00:06:00.519 human language. Now the augmented intelligence 00:06:00.519 --> 00:06:02.519 perspective, which bodies like the American College 00:06:02.519 --> 00:06:05.100 of Physicians strongly advocate, views AI not 00:06:05.100 --> 00:06:07.540 as a replacement for human expertise, no, but 00:06:07.540 --> 00:06:10.339 as a powerful enhancement, a tool, and the evidence 00:06:10.339 --> 00:06:12.360 consistently backs this up. When physicians work 00:06:12.360 --> 00:06:14.279 with AI like reviewing mammograms for breast 00:06:14.279 --> 00:06:17.100 cancer, the diagnostic outcomes are demonstrably 00:06:17.100 --> 00:06:19.860 better compared to either working alone. It's 00:06:19.860 --> 00:06:22.199 about leveraging AI's knack for processing vast 00:06:22.199 --> 00:06:25.180 information, spotting subtle patterns, allowing 00:06:25.180 --> 00:06:27.360 the clinician to focus their highly trained judgment 00:06:27.360 --> 00:06:29.910 where it truly matters. It should lead to a more 00:06:29.910 --> 00:06:32.089 efficient, more precise healthcare journey overall. 00:06:32.350 --> 00:06:34.509 That's fascinating. So it's really about collaboration, 00:06:34.610 --> 00:06:37.589 isn't it? Where the human brain and the machine's 00:06:37.589 --> 00:06:40.670 processing power kind of create a sum greater 00:06:40.670 --> 00:06:43.410 than their parts. Can you give us some specific, 00:06:43.649 --> 00:06:46.769 tangible examples of how this augmentation plays 00:06:46.769 --> 00:06:49.649 out in a clinic or a sports science setting? 00:06:50.269 --> 00:06:51.810 Where's it making a real difference right now? 00:06:52.069 --> 00:06:54.329 Absolutely. Let's think about something as fundamental 00:06:54.329 --> 00:06:57.290 as clinical documentation. It takes up so much 00:06:57.290 --> 00:07:00.519 time. Digital scribes, like Dax Copilot from 00:07:00.519 --> 00:07:03.339 Microsoft, they're AI -powered, voice -enabled 00:07:03.339 --> 00:07:06.220 solutions. They basically listen in during patient 00:07:06.220 --> 00:07:09.000 interactions and automatically generate comprehensive 00:07:09.000 --> 00:07:11.240 clinical notes. Now, this isn't just about saving 00:07:11.240 --> 00:07:13.480 time, though that's significant. Studies show, 00:07:13.480 --> 00:07:16.120 I think it was 70 % of physicians using it reported 00:07:16.120 --> 00:07:18.560 better work -life balance, less burnout, and 00:07:18.560 --> 00:07:20.879 perhaps even more importantly, 93 % of patients 00:07:20.879 --> 00:07:23.139 felt their doctor was more personable, more engaged, 00:07:23.199 --> 00:07:24.680 because they weren't constantly looking at a 00:07:24.680 --> 00:07:26.920 screen and typing. It frees clinicians up to 00:07:26.920 --> 00:07:29.019 actually connect with their patients rather than 00:07:29.019 --> 00:07:31.139 wrestling with the admin side of things. That's 00:07:31.139 --> 00:07:32.839 a huge impact on the human side of medicine. 00:07:33.259 --> 00:07:35.100 Exactly. And we see this powerful augmentation 00:07:35.100 --> 00:07:38.220 in diagnostic fields, too. Radiology, pathology, 00:07:38.759 --> 00:07:42.199 ophthalmology, dermatology. Here, AI systems 00:07:42.199 --> 00:07:44.779 can analyze images, identify anomalies, or subtle 00:07:44.779 --> 00:07:47.079 patterns the human eye might miss, or maybe just 00:07:47.079 --> 00:07:49.319 take much longer to spot. Again, it's not saying 00:07:49.319 --> 00:07:52.439 the human is redundant. Far from it. The AI provides 00:07:52.439 --> 00:07:54.819 a powerful second opinion maybe, or a rapid first 00:07:54.819 --> 00:07:57.360 pass, flagging up areas of concern for the human 00:07:57.360 --> 00:08:00.699 expert's focused review. Even seemingly mundane 00:08:00.699 --> 00:08:03.980 admin tasks, patient scheduling, resource optimization 00:08:03.980 --> 00:08:06.259 in a hospital, they're benefiting from AI too, 00:08:06.519 --> 00:08:08.199 making things run smoother, more efficiently. 00:08:08.780 --> 00:08:10.980 There's a quote from Dominique Rothenflue, a 00:08:10.980 --> 00:08:13.339 highly respected orthopedic surgeon at Oxford 00:08:13.339 --> 00:08:15.800 University Hospitals. He read it quite profoundly. 00:08:16.600 --> 00:08:18.639 Machines may not necessarily replace physicians, 00:08:19.079 --> 00:08:21.100 but physicians making use of AI will at some 00:08:21.100 --> 00:08:23.519 point replace those not using it. And the sale 00:08:23.519 --> 00:08:25.579 of the other. Now, that's not meant as a threat. 00:08:25.699 --> 00:08:27.720 It's more of an imperative, really, for professionals 00:08:27.720 --> 00:08:29.519 to adapt, to integrate these tools into their 00:08:29.519 --> 00:08:31.399 practice, to ensure they stay at the forefront 00:08:31.399 --> 00:08:33.179 and keep delivering the highest standard of care. 00:08:33.440 --> 00:08:35.240 That is a powerful statement about the future 00:08:35.240 --> 00:08:39.110 of practice. Okay, so AI is clearly reshaping 00:08:39.110 --> 00:08:41.950 the before picture diagnosis planning. Let's 00:08:41.950 --> 00:08:43.909 delve into that a bit more. Before a patient 00:08:43.909 --> 00:08:46.149 even gets near an operating theater or an athlete 00:08:46.149 --> 00:08:48.549 steps onto the pitch, AI is already at work. 00:08:48.889 --> 00:08:50.570 Can you walk us through some of these really 00:08:50.570 --> 00:08:53.250 critical preoperative applications? Yes, the 00:08:53.250 --> 00:08:56.330 preoperative phase is, er, tremendously fertile 00:08:56.330 --> 00:08:59.379 ground for AI. It's transforming how we diagnose, 00:08:59.480 --> 00:09:02.019 predict outcomes, plan interventions. Let's start 00:09:02.019 --> 00:09:05.259 with imaging and diagnosis. Traditionally, diagnosing 00:09:05.259 --> 00:09:08.360 something like osteoarthritis, OA, from x -rays, 00:09:08.720 --> 00:09:10.820 has been manual, often subjective, depends who's 00:09:10.820 --> 00:09:13.299 looking, right? Now, AI algorithms, especially 00:09:13.299 --> 00:09:15.820 these convolutional neural networks, CNNs, they're 00:09:15.820 --> 00:09:18.860 basically AI systems really good at seeing and 00:09:18.860 --> 00:09:21.360 interpreting patterns in images, a bit like our 00:09:21.360 --> 00:09:23.259 own brain's process vision, they're automating 00:09:23.259 --> 00:09:26.279 this. For instance, one significant study developed 00:09:26.279 --> 00:09:29.259 a CNN model for hip OA diagnosis using hundreds 00:09:29.259 --> 00:09:31.519 of radiographs. And what they found was compelling. 00:09:31.980 --> 00:09:34.639 This AI model showed diagnostic aptitude comparable 00:09:34.639 --> 00:09:37.080 to a physician with 10 years clinical experience. 00:09:37.159 --> 00:09:41.000 10 years? Wow. Exactly. So what that means for 00:09:41.000 --> 00:09:43.960 you, the patient, is potentially faster, more 00:09:43.960 --> 00:09:47.460 consistent, highly accurate diagnoses. Similarly, 00:09:47.799 --> 00:09:50.700 AI can accurately classify the severity of knee 00:09:50.700 --> 00:09:53.620 OA using the standard Kellgren -Lorentz, or KL, 00:09:53.919 --> 00:09:56.690 greeting system. It offers better reproducibility 00:09:56.690 --> 00:10:00.250 than manual evaluation. Some newer approaches, 00:10:00.590 --> 00:10:03.169 deep Siamese CNNs, can even give probability 00:10:03.169 --> 00:10:06.409 distributions for ambiguous cases. That's incredibly 00:10:06.409 --> 00:10:08.730 useful for clinical decisions, helping clinicians 00:10:08.730 --> 00:10:11.370 gauge the AI's confidence level. But there's 00:10:11.370 --> 00:10:13.669 a key challenge here, one we absolutely must 00:10:13.669 --> 00:10:17.259 address, generalizability. Algorithms trained 00:10:17.259 --> 00:10:19.860 mainly on data from one hospital might struggle 00:10:19.860 --> 00:10:21.679 with different labeling systems or radiation 00:10:21.679 --> 00:10:23.419 protocols elsewhere. Right, so it might work 00:10:23.419 --> 00:10:25.720 brilliantly in one place, but not another. Precisely. 00:10:25.860 --> 00:10:27.620 So for broad use across different hospitals, 00:10:27.779 --> 00:10:30.159 different clinics, these models need really diverse, 00:10:30.419 --> 00:10:32.379 extensive data sets for training. That's crucial. 00:10:32.600 --> 00:10:34.659 That diverse data point keeps coming up, doesn't 00:10:34.659 --> 00:10:37.019 it? Okay, so beyond just diagnosing what's happening 00:10:37.019 --> 00:10:39.460 now, how is AI helping us predict what's likely 00:10:39.460 --> 00:10:42.200 to happen for a patient's future? Yes, this is 00:10:42.200 --> 00:10:45.269 where AI truly shines, I think. in its proactive 00:10:45.269 --> 00:10:48.830 capability, moving us from just reacting to treatment 00:10:48.830 --> 00:10:51.409 towards actually having some foresight. Machine 00:10:51.409 --> 00:10:54.149 learning and deep learning models are now significantly 00:10:54.149 --> 00:10:56.330 outperforming traditional statistical methods 00:10:56.330 --> 00:10:58.649 in predicting, say, the structural progression 00:10:58.649 --> 00:11:02.409 of OA, or identifying individuals at high risk 00:11:02.409 --> 00:11:05.129 for needing a total joint replacement. a TJR. 00:11:05.570 --> 00:11:07.809 For example, one study developed a deep learning 00:11:07.809 --> 00:11:10.470 program using baseline knee x -rays to predict 00:11:10.470 --> 00:11:12.990 medial joint space loss. That's a key sign of 00:11:12.990 --> 00:11:16.090 OA getting worse. Their model achieved a significantly 00:11:16.090 --> 00:11:18.870 higher AUC score. That's a metric for how well 00:11:18.870 --> 00:11:21.330 a model distinguishes between groups, like those 00:11:21.330 --> 00:11:23.690 who will progress versus those who won't. Their 00:11:23.690 --> 00:11:27.970 AUC was, I think, 0 .799 compared to 0 .660 for 00:11:27.970 --> 00:11:30.250 traditional models. That indicates a much better 00:11:30.250 --> 00:11:32.289 predictive ability. Others found deep learning 00:11:32.289 --> 00:11:34.090 could predict the need for total knee replacement, 00:11:34.289 --> 00:11:36.950 TKA, just from radiographs, with pretty impressive 00:11:36.950 --> 00:11:39.250 sensitivity and specificity. And that's not just 00:11:39.250 --> 00:11:41.370 academic, is it? That has real -world implications. 00:11:41.850 --> 00:11:45.289 Absolutely. This ability to predict lets us potentially 00:11:45.289 --> 00:11:48.629 foresee osteoarthritis progression much earlier. 00:11:48.970 --> 00:11:50.909 It enables proactive interventions that might 00:11:50.909 --> 00:11:53.029 slow the disease down for a patient, or at least 00:11:53.029 --> 00:11:55.690 help them prepare for what's likely coming. It's 00:11:55.690 --> 00:11:58.169 about being prescriptive, not just descriptive, 00:11:58.730 --> 00:12:01.480 giving people options earlier. So it's not just 00:12:01.480 --> 00:12:03.340 about what's happening now, but what's likely 00:12:03.340 --> 00:12:05.740 down the road. And how does that translate into 00:12:05.740 --> 00:12:08.120 improving the patient experience? Especially 00:12:08.120 --> 00:12:09.940 given, as you said, quite a few patients report 00:12:09.940 --> 00:12:12.500 dissatisfaction even after a technically successful 00:12:12.500 --> 00:12:15.539 surgery. Exactly. Patient satisfaction is a critical 00:12:15.539 --> 00:12:17.899 challenge. It's quite sobering, really, that 00:12:17.899 --> 00:12:20.299 20 to 30 percent of patients report no improvement 00:12:20.299 --> 00:12:23.240 or even dissatisfaction after a total joint replacement. 00:12:23.899 --> 00:12:26.120 So AI models are now being developed specifically 00:12:26.120 --> 00:12:28.679 to predict this dissatisfaction. These models 00:12:28.679 --> 00:12:31.220 can identify influential factors, things like 00:12:31.220 --> 00:12:33.940 preoperative functional scores, maybe mental 00:12:33.940 --> 00:12:36.019 health status allowing us to spot high -risk 00:12:36.019 --> 00:12:38.419 patients beforehand. One study, for instance, 00:12:38.659 --> 00:12:41.200 trained models that achieved strong AUCs, around 00:12:41.200 --> 00:12:44.480 .81, in predicting knee -away dissatisfaction 00:12:44.480 --> 00:12:47.409 two years after surgery. This means we can potentially 00:12:47.409 --> 00:12:49.590 intervene with targeted preoperative education, 00:12:50.289 --> 00:12:52.629 perhaps health optimization programs, or even 00:12:52.629 --> 00:12:54.610 guide patients towards non -surgical options 00:12:54.610 --> 00:12:57.429 earlier. That could reduce unnecessary treatments 00:12:57.429 --> 00:13:00.330 and significant costs later on. It's about tailoring 00:13:00.330 --> 00:13:02.190 the care pathway much more to the individual. 00:13:02.429 --> 00:13:04.850 That sounds much more patient -centric. It aims 00:13:04.850 --> 00:13:07.990 to be. And AI also excels at risk stratification 00:13:07.990 --> 00:13:10.330 of various complications, making surgery safer 00:13:10.330 --> 00:13:12.950 overall. For instance, predicting the need for 00:13:12.950 --> 00:13:15.730 blood transfusion after TKA1 and ML model did 00:13:15.730 --> 00:13:19.750 this with an excellent AUC of 0 .880 using just 00:13:19.750 --> 00:13:22.929 six readily available preoperative factors. Or 00:13:22.929 --> 00:13:25.730 for periprosthetic joint infection, PJI, which 00:13:25.730 --> 00:13:28.669 is a really severe complication. An artificial 00:13:28.669 --> 00:13:31.009 neural network identified obesity as a stronger 00:13:31.009 --> 00:13:33.750 predictor than previously thought. And ML systems 00:13:33.750 --> 00:13:35.850 have even outperformed traditional fixed criteria 00:13:35.850 --> 00:13:38.529 for diagnosis, offering more adaptive, transparent 00:13:38.529 --> 00:13:40.909 models, helping clinicians diagnose faster and 00:13:40.909 --> 00:13:43.509 more accurately. Length of stay, LOS, that's 00:13:43.509 --> 00:13:46.090 a major cost driver in hospitals. AI models can 00:13:46.090 --> 00:13:48.850 identify key predictors age, BMI, operation time, 00:13:48.990 --> 00:13:51.230 blood results. This allows for better resource 00:13:51.230 --> 00:13:53.129 allocation, more efficient hospital management. 00:13:53.570 --> 00:13:55.539 Better for everyone, really, but... And this 00:13:55.539 --> 00:13:57.740 is really important, but a significant ethical 00:13:57.740 --> 00:14:00.019 concern arises here. These powerful predictive 00:14:00.019 --> 00:14:02.100 methods could inadvertently marginalize patients 00:14:02.100 --> 00:14:04.879 flagged as too high risk. Ah yes, the risk of 00:14:04.879 --> 00:14:07.240 creating a sort of uninsurable category but for 00:14:07.240 --> 00:14:10.659 surgery. Potentially yes. We must be incredibly 00:14:10.659 --> 00:14:13.019 vigilant to ensure these tools are applied equitably, 00:14:13.340 --> 00:14:15.639 that no vulnerable individual is excluded from 00:14:15.639 --> 00:14:18.200 necessary treatment just because of an algorithm's 00:14:18.200 --> 00:14:21.080 prediction. The goal must always be to improve 00:14:21.080 --> 00:14:23.480 access and outcomes for all. That's a really 00:14:23.480 --> 00:14:25.620 crucial point about equitable access, ensuring 00:14:25.620 --> 00:14:28.200 these powerful tools enhance rather than limit 00:14:28.200 --> 00:14:31.639 patient care. Okay, moving beyond diagnostics 00:14:31.639 --> 00:14:35.240 and risk, how is AI influencing these sort of 00:14:35.240 --> 00:14:37.480 nuts and bolts of preparing for surgery? Things 00:14:37.480 --> 00:14:39.720 like surgical planning and even implant design. 00:14:40.059 --> 00:14:42.759 Yes, this is another area where AI is driving 00:14:42.759 --> 00:14:45.899 some incredible innovation. AI software is now 00:14:45.899 --> 00:14:48.399 directly informing implant design, positioning, 00:14:48.799 --> 00:14:52.019 and 3D templating. All critical for successful 00:14:52.019 --> 00:14:54.840 outcomes. For example, researchers have applied 00:14:54.840 --> 00:14:57.240 machine learning to optimize short stem implant 00:14:57.240 --> 00:14:59.600 design and hip replacements, aiming to reduce 00:14:59.600 --> 00:15:01.620 something called stress shielding, which can 00:15:01.620 --> 00:15:04.299 weaken the bone over time. Others have used deep 00:15:04.299 --> 00:15:06.379 learning to estimate patient -specific adjustments 00:15:06.379 --> 00:15:09.279 for optimal landmark use in total knee arthroplasty 00:15:09.279 --> 00:15:11.639 planning, ensuring a much more precise fit for 00:15:11.639 --> 00:15:14.840 that individual patient. There's also AI -based 00:15:14.840 --> 00:15:18.419 3D preoperative planning software, like AI -HYPE 00:15:18.419 --> 00:15:21.080 for total hip replacement, THA. Studies found 00:15:21.080 --> 00:15:23.179 it offered significantly more accurate surgical 00:15:23.179 --> 00:15:25.759 planning and component sizing compared to traditional 00:15:25.759 --> 00:15:28.200 2D templating. And it saves surgeons valuable 00:15:28.200 --> 00:15:30.559 time by giving these highly precise visualizations 00:15:30.559 --> 00:15:34.500 of the anatomy. usually need CT scans, which 00:15:34.500 --> 00:15:37.159 means more radiation and cost. That's often been 00:15:37.159 --> 00:15:39.940 the case, yes, but future AI applications promise 00:15:39.940 --> 00:15:42.519 to mitigate this. Some researchers, for instance, 00:15:42.779 --> 00:15:44.919 have developed an innovative AI program that 00:15:44.919 --> 00:15:48.139 can convert standard 2D X -rays into highly accurate 00:15:48.139 --> 00:15:50.860 3D bone models. This could potentially reduce 00:15:50.860 --> 00:15:53.659 the need for those costly higher radiation CT 00:15:53.659 --> 00:15:56.259 scans, making this advanced planning much more 00:15:56.259 --> 00:15:59.419 accessible. Another fascinating application is 00:15:59.419 --> 00:16:02.370 automatic implant identification. With so many 00:16:02.370 --> 00:16:04.889 THA procedures done now, correctly identifying 00:16:04.889 --> 00:16:07.149 implant types for potential revisions is vital, 00:16:07.629 --> 00:16:09.330 but it can be tricky and time consuming with 00:16:09.330 --> 00:16:11.029 the sheer number of different implants out there. 00:16:11.129 --> 00:16:13.429 So, researchers developed a deep learning tool 00:16:13.429 --> 00:16:16.149 for automatic THA implant detection, trained 00:16:16.149 --> 00:16:19.529 on over 240 ,000 radiographs. It achieved impressive 00:16:19.529 --> 00:16:23.129 accuracy, 98 .9 % internal, 97 % external for 00:16:23.129 --> 00:16:25.549 femoral components across different x -ray views. 00:16:26.009 --> 00:16:27.769 And a unique, really vital aspect of their model 00:16:27.769 --> 00:16:29.690 is it incorporates uncertainty quantification, 00:16:29.789 --> 00:16:32.110 basically. It gauges its own confidence level 00:16:32.110 --> 00:16:34.009 in outlier detection, flagging data points that 00:16:34.009 --> 00:16:36.149 might skew results. These features are absolutely 00:16:36.149 --> 00:16:38.149 vital for enhancing trust in the model and supporting 00:16:38.149 --> 00:16:40.700 its potential for widespread clinical use. gives 00:16:40.700 --> 00:16:43.259 surgeons confidence in what the AI is telling 00:16:43.259 --> 00:16:46.639 them. OK, so AI is clearly reshaping the before 00:16:46.639 --> 00:16:49.940 picture with remarkable precision. But what happens 00:16:49.940 --> 00:16:52.980 during surgery? And then crucially, in that often 00:16:52.980 --> 00:16:56.179 overlooked recovery phase, how is AI stepping 00:16:56.179 --> 00:16:58.500 in there? That's an excellent question because 00:16:58.500 --> 00:17:00.820 AI's utility definitely extends significantly 00:17:00.820 --> 00:17:03.039 into the intraoperative and postoperative periods 00:17:03.039 --> 00:17:05.720 too. During surgery, we're seeing this rapid 00:17:05.720 --> 00:17:08.720 integration of AI into quite sophisticated tools. 00:17:09.140 --> 00:17:11.779 Think augmented reality, AR -based navigation 00:17:11.779 --> 00:17:14.599 systems, AI -assisted arthroscopic procedures, 00:17:15.180 --> 00:17:18.160 and increasingly advanced robotic surgery. Robotic 00:17:18.160 --> 00:17:20.220 surgery, already known for its minimally invasive 00:17:20.220 --> 00:17:22.640 approach, brings greater precision. That can 00:17:22.640 --> 00:17:25.039 lead to less blood loss, less pain for the patient, 00:17:25.420 --> 00:17:28.160 and typically shorter recovery times. AI's role 00:17:28.160 --> 00:17:30.000 here isn't just controlling the robot's arm. 00:17:30.240 --> 00:17:32.859 It refines surgical planning by analyzing huge 00:17:32.859 --> 00:17:35.339 amounts of data from previous successful surgeries. 00:17:35.559 --> 00:17:38.000 And it provides real -time feedback during delicate 00:17:38.000 --> 00:17:40.839 procedures like bone cutting or soft tissue resection. 00:17:41.039 --> 00:17:44.220 This helps ensure remarkably consistent, reproducible 00:17:44.220 --> 00:17:46.730 results, which is key for patient safety. getting 00:17:46.730 --> 00:17:48.849 the best possible outcome. It sounds incredible, 00:17:48.990 --> 00:17:50.450 but there must be challenges to adopting this 00:17:50.450 --> 00:17:53.170 everywhere. Oh absolutely. As with any advanced 00:17:53.170 --> 00:17:55.869 tech, there are hurdles. The financial investment 00:17:55.869 --> 00:17:58.589 for robotic gear and the consumables is substantial. 00:17:58.809 --> 00:18:01.849 System compatibility can be limited to specific 00:18:01.849 --> 00:18:05.710 implant types, and crucially, extensive ongoing 00:18:05.710 --> 00:18:08.250 training for surgeons and the whole theater staff 00:18:08.250 --> 00:18:11.250 is absolutely essential for safe and effective 00:18:11.250 --> 00:18:13.730 integration. It really requires a dedicated commitment 00:18:13.730 --> 00:18:15.769 from healthcare systems. It's not just plug and 00:18:15.769 --> 00:18:17.750 play. It sounds like incredible progress, but 00:18:17.750 --> 00:18:20.289 the idea of AI assisting during surgery, it does 00:18:20.289 --> 00:18:22.769 raise crucial questions, doesn't it? About human 00:18:22.769 --> 00:18:25.710 expertise versus relying on the machine. How 00:18:25.710 --> 00:18:27.950 do we ensure human intuition and skill remain 00:18:27.950 --> 00:18:30.410 paramount? You've hit on a really critical point 00:18:30.410 --> 00:18:32.930 there. A very valid concern is what we sometimes 00:18:32.930 --> 00:18:35.950 call automation bias or over -reliance on AI. 00:18:36.710 --> 00:18:39.630 If surgeons become overly dependent on the technology, 00:18:40.190 --> 00:18:42.269 there's a genuine risk that their foundational 00:18:42.269 --> 00:18:45.329 human intuition, their clinical judgment, even 00:18:45.329 --> 00:18:47.589 their manual skills could diminish over time. 00:18:47.849 --> 00:18:50.609 And if the technology fails or encounters an 00:18:50.609 --> 00:18:53.009 unexpected situation the AI hasn't seen before, 00:18:53.630 --> 00:18:56.190 that over -reliance could lead to severe consequences. 00:18:56.430 --> 00:18:59.009 We can draw parallels perhaps with the airline 00:18:59.009 --> 00:19:01.750 industry's use of autopilot. It's incredibly 00:19:01.750 --> 00:19:04.549 efficient for routine flight, yes, but pilots 00:19:04.549 --> 00:19:07.049 must remain highly skilled, capable of taking 00:19:07.049 --> 00:19:10.240 manual control in emergencies. This also fundamentally 00:19:10.240 --> 00:19:12.720 changes surgical training. It needs to evolve 00:19:12.720 --> 00:19:15.500 to include not just manual skill, but also proficiency 00:19:15.500 --> 00:19:18.000 in operating, interpreting, and critically understanding 00:19:18.000 --> 00:19:19.960 these advanced machines. It's about building 00:19:19.960 --> 00:19:22.720 a robust human AI team, not just delegating tasks 00:19:22.720 --> 00:19:25.200 to a black box. That makes sense. Building a 00:19:25.200 --> 00:19:27.380 team, not just relying on a tool. And what about 00:19:27.380 --> 00:19:29.819 after the surgery, in the recovery phase? Right. 00:19:30.039 --> 00:19:32.880 Postoperatively, AI is revolutionizing rehabilitation 00:19:32.880 --> 00:19:35.420 and monitoring, primarily through remote patient 00:19:35.420 --> 00:19:38.490 monitoring. The Internet of Things, IoT, and 00:19:38.490 --> 00:19:40.690 mHealth mobile health solutions. They integrate 00:19:40.690 --> 00:19:43.369 AI into wearable devices, sophisticated sensors, 00:19:43.630 --> 00:19:46.009 smartphone apps, collecting vast amounts of real 00:19:46.009 --> 00:19:49.269 -time patient data from their own home. Clinicians 00:19:49.269 --> 00:19:51.869 can then remotely monitor a patient's rehab progress, 00:19:52.450 --> 00:19:54.470 stepping in quickly if milestones aren't met 00:19:54.470 --> 00:19:57.329 or if complications seem to be rising. For instance, 00:19:57.509 --> 00:19:59.910 these innovations have shown real value in reducing 00:19:59.910 --> 00:20:02.549 opioid use after surgery and improving adherence 00:20:02.549 --> 00:20:04.829 to anticoagulation drugs, which are vital for 00:20:04.829 --> 00:20:07.569 preventing serious blood clots. There was a study 00:20:07.569 --> 00:20:09.670 developing motion -based machine learning software 00:20:09.670 --> 00:20:12.250 for remote total knee arthroplasty monitoring. 00:20:12.930 --> 00:20:15.130 It used a simple knee sleeve with Bluetooth sensors 00:20:15.130 --> 00:20:18.049 transmitting data to a phone app. It gave clinicians 00:20:18.049 --> 00:20:20.210 invaluable insights into the patient's mobility, 00:20:20.589 --> 00:20:22.309 how well they were doing their exercises, and 00:20:22.309 --> 00:20:24.349 could even flag up early signs of complications. 00:20:24.990 --> 00:20:27.109 And crucially, patients found it really motivating, 00:20:27.410 --> 00:20:30.309 engaging. It transforms how we oversee recovery 00:20:30.309 --> 00:20:33.750 from just reactive clinic visits to a more proactive 00:20:33.750 --> 00:20:36.190 continuous feedback loop directly from their 00:20:36.190 --> 00:20:38.490 home. That sounds much more empowering for the 00:20:38.490 --> 00:20:41.599 patient too. It really can be. And finally, AI 00:20:41.599 --> 00:20:43.980 extends its powerful image analysis skills to 00:20:43.980 --> 00:20:47.079 complication screening via imaging post -operatively 00:20:47.079 --> 00:20:49.799 as well. It can detect early signs of implant 00:20:49.799 --> 00:20:52.359 loosening or determine dislocation risk, often 00:20:52.359 --> 00:20:54.640 with far greater accuracy and efficiency than 00:20:54.640 --> 00:20:57.460 traditional methods. That allows for prompt intervention 00:20:57.460 --> 00:21:00.299 before a problem really escalates. Prosthetic 00:21:00.299 --> 00:21:02.140 loosening, for example, it's a leading cause 00:21:02.140 --> 00:21:04.720 of arthroplasty failure, accounts for 30 % of 00:21:04.720 --> 00:21:07.700 knee revisions, a huge 55 % of hip revisions. 00:21:07.900 --> 00:21:09.880 It's a massive clinical and economic burden. 00:21:10.109 --> 00:21:12.990 One group developed a fully automated CNN algorithm 00:21:12.990 --> 00:21:15.430 to detect mechanical loosening of hip implants 00:21:15.430 --> 00:21:17.910 from plain x -rays, reporting high sensitivity 00:21:17.910 --> 00:21:21.609 around 91 .6%. This doesn't just supplement the 00:21:21.609 --> 00:21:23.569 surgeon's decision -making, it also frees up 00:21:23.569 --> 00:21:25.170 their valuable time for more patient -focused 00:21:25.170 --> 00:21:27.970 activities. Similarly, hip dislocation after 00:21:27.970 --> 00:21:29.950 replacement is a significant challenge, costs 00:21:29.950 --> 00:21:32.869 a lot too. Another group developed a CNN, trained 00:21:32.869 --> 00:21:35.670 on over 92 ,000 x -rays, to classify patients 00:21:35.670 --> 00:21:38.410 for dislocation risk with about 89 % sensitivity. 00:21:38.569 --> 00:21:40.809 Identifying low -risk patients with high confidence 00:21:40.809 --> 00:21:42.950 could mean fewer post -operative restrictions, 00:21:43.410 --> 00:21:45.849 fewer follow -up visits. That eases the burden 00:21:45.849 --> 00:21:49.109 significantly on both patients and the healthcare 00:21:49.109 --> 00:21:52.250 system. So, okay, broadening out from the clinical 00:21:52.250 --> 00:21:54.869 setting. What does all this mean for athletes 00:21:54.869 --> 00:21:58.609 from the weekend warrior maybe right up to the 00:21:58.609 --> 00:22:01.750 elite professional? How is AI truly enabling 00:22:01.750 --> 00:22:04.250 this shift from just reacting to injuries to 00:22:04.250 --> 00:22:06.710 actively preventing them, transforming sports 00:22:06.710 --> 00:22:09.490 medicine? Yes, this is a truly transformative 00:22:09.490 --> 00:22:12.809 area. AI fundamentally shifts the whole paradigm 00:22:12.809 --> 00:22:15.289 in sports medicine from reactive treatment to 00:22:15.289 --> 00:22:18.369 proactive prevention. Musculoskeletal injuries 00:22:18.369 --> 00:22:20.730 globally are a huge burden in sports, aren't 00:22:20.730 --> 00:22:24.539 they, impacting careers, well -being. AI, mainly 00:22:24.539 --> 00:22:26.279 through advanced machine learning and deep learning, 00:22:26.420 --> 00:22:28.720 is designed to analyze these incredibly complex 00:22:28.720 --> 00:22:31.019 multi -dimensional data sets. We're talking data 00:22:31.019 --> 00:22:33.279 from high -tech wearables, detailed biomechanics 00:22:33.279 --> 00:22:36.119 assessments, performance metrics, speed, power, 00:22:36.200 --> 00:22:38.299 and even subtle psychological factors like sleep 00:22:38.299 --> 00:22:40.839 quality or stress levels. This rich data lets 00:22:40.839 --> 00:22:43.680 AI identify subtle patterns, risk factors that 00:22:43.680 --> 00:22:46.140 traditional methods just relying on human observation 00:22:46.140 --> 00:22:48.539 might easily miss. It's about getting ahead of 00:22:48.539 --> 00:22:50.660 the curve, predicting vulnerability before an 00:22:50.660 --> 00:22:52.339 injury actually occurs. So it's like seeing the 00:22:52.339 --> 00:22:56.000 invisible signs. In a way, yes. In team sports, 00:22:56.519 --> 00:22:58.799 AI models are already widely used to predict 00:22:58.799 --> 00:23:01.400 muscle injuries, meticulously manage workloads 00:23:01.400 --> 00:23:03.779 in demanding sports like football, basketball. 00:23:04.619 --> 00:23:07.039 Think about high -contact sports rugby American 00:23:07.039 --> 00:23:10.599 football. Here, AI precisely monitors collision 00:23:10.599 --> 00:23:13.220 data to track an athlete's recovery from concussions, 00:23:13.819 --> 00:23:15.720 providing early warnings based on cumulative 00:23:15.720 --> 00:23:18.589 impacts over a season. Studies clearly highlight 00:23:18.589 --> 00:23:21.250 AI's role as an early warning system, spotting 00:23:21.250 --> 00:23:23.470 injury risks before they even manifest as a serious 00:23:23.470 --> 00:23:26.289 problem. This proactive approach lets coaches, 00:23:26.549 --> 00:23:28.569 medical staff, implement timely interventions, 00:23:28.869 --> 00:23:30.950 adjust training loads, rest an athlete, potentially 00:23:30.950 --> 00:23:32.990 saving their season, maybe even their career. 00:23:33.210 --> 00:23:35.269 And what about individual sports? Where the demands 00:23:35.269 --> 00:23:37.630 might be different? The application is equally 00:23:37.630 --> 00:23:40.549 groundbreaking there. AI monitors real -time 00:23:40.549 --> 00:23:43.210 biomechanical data, often from sensors embedded 00:23:43.210 --> 00:23:45.950 right into clothing or equipment. It detects 00:23:45.950 --> 00:23:48.789 subtle signs of fatigue or poor technique that 00:23:48.789 --> 00:23:51.329 could predispose an athlete to injury. There 00:23:51.329 --> 00:23:53.630 was the first prospective longitudinal cohort 00:23:53.630 --> 00:23:56.789 study using machine learning to identify specific 00:23:56.789 --> 00:23:59.529 risk factors for running -related injuries that's 00:23:59.529 --> 00:24:01.730 revolutionizing prevention strategies for runners. 00:24:02.650 --> 00:24:04.809 We're also seeing truly adaptive technologies. 00:24:05.240 --> 00:24:08.059 Like mechatronic ski bindings using AI for real 00:24:08.059 --> 00:24:10.299 -time knee protection in alpine skiing, they 00:24:10.299 --> 00:24:12.700 instantly adjust based on forces to prevent injury. 00:24:12.920 --> 00:24:15.259 That's incredible stuff. In sports like cycling, 00:24:15.440 --> 00:24:18.759 swimming, tennis, gymnastics, AI analyses precise 00:24:18.759 --> 00:24:21.720 biomechanics to prevent overuse injuries. CNN's 00:24:21.720 --> 00:24:23.779 analyzing high dimensional wearable sensor data 00:24:23.779 --> 00:24:25.940 can detect risky movements that would otherwise 00:24:25.940 --> 00:24:27.839 just go completely unnoticed. And it helps with 00:24:27.839 --> 00:24:30.039 getting back after an injury too. Absolutely. 00:24:30.400 --> 00:24:33.000 AI plays a critical role in optimizing return 00:24:33.000 --> 00:24:35.460 to play protocols. It doesn't just predict, it 00:24:35.460 --> 00:24:38.420 prescribes. By creating personalized rehabilitation 00:24:38.420 --> 00:24:40.599 plans based on an individual athlete's unique 00:24:40.599 --> 00:24:44.000 data, AI acts as both a predictive and prescriptive 00:24:44.000 --> 00:24:46.880 tool. It meticulously monitors recovery metrics, 00:24:47.440 --> 00:24:49.839 tracks cumulative impacts, especially vital and 00:24:49.839 --> 00:24:52.259 contact sports, to minimize the risk of re -injury 00:24:52.259 --> 00:24:54.940 and support long -term athlete health. A prime 00:24:54.940 --> 00:24:57.799 example in action is MAI Motion's AI applications. 00:24:57.799 --> 00:25:00.000 They offer precise diagnostics, customized rehab 00:25:00.000 --> 00:25:02.480 pathways, predictive analytics for athlete recovery, 00:25:02.779 --> 00:25:04.460 and they've showed significant reductions in 00:25:04.460 --> 00:25:06.460 recovery time, getting athletes back to peak 00:25:06.460 --> 00:25:09.160 performance faster, safer. Similarly, Health 00:25:09.160 --> 00:25:11.759 First, using closed -loop tech, deployed the 00:25:11.529 --> 00:25:14.049 those 17 predictive models, nearly a thousand 00:25:14.049 --> 00:25:16.450 customized machine learning features. It just 00:25:16.450 --> 00:25:18.650 shows the immense value AI can drive by speeding 00:25:18.650 --> 00:25:20.869 up insights into clinical workflows, improving 00:25:20.869 --> 00:25:23.230 health outcomes across a population. It truly 00:25:23.230 --> 00:25:25.670 heralds a new era of data -driven athlete recovery 00:25:25.670 --> 00:25:27.769 and health management, making prevention not 00:25:27.769 --> 00:25:30.539 just a goal, but a tangible reality. It's clearly 00:25:30.539 --> 00:25:33.839 transformative. But as you've hinted, no technology 00:25:33.839 --> 00:25:36.019 like this comes without challenges. While the 00:25:36.019 --> 00:25:38.640 promise is clear, what are the significant hurdles 00:25:38.640 --> 00:25:41.180 that really need addressing for AI to embed itself 00:25:41.180 --> 00:25:43.900 seamlessly and responsibly into health care and 00:25:43.900 --> 00:25:46.079 sports, especially thinking about those ethical 00:25:46.079 --> 00:25:48.599 nuances? Indeed, there are significant hurdles. 00:25:49.140 --> 00:25:52.119 They mainly cluster around data, ethical considerations, 00:25:52.779 --> 00:25:55.960 and, well, operational realities. Firstly, data 00:25:55.960 --> 00:25:59.160 quality diversity. As I mentioned earlier, AI's 00:25:59.160 --> 00:26:01.140 effectiveness is entirely dependent on having 00:26:01.140 --> 00:26:03.579 substantial, diverse, high -quality datasets. 00:26:04.480 --> 00:26:06.880 The reality is, much healthcare data remains 00:26:06.880 --> 00:26:10.059 uncaptured, unstructured, varied, stuck in silos. 00:26:10.779 --> 00:26:13.039 It's hugely costly and time -consuming to clean, 00:26:13.259 --> 00:26:16.099 standardize, integrate it for AI. Crucially, 00:26:16.440 --> 00:26:18.779 comprehensive data, including those social determinants 00:26:18.779 --> 00:26:20.859 of health, lifestyle choices, daily activities 00:26:20.859 --> 00:26:23.880 vital for accurate prediction, are often systematically 00:26:23.880 --> 00:26:26.700 lacking. Our own RU case survey found I think 00:26:26.700 --> 00:26:29.359 76 % of the MSK community agrees it's difficult 00:26:29.359 --> 00:26:31.900 to access and use patient data at scale, often 00:26:31.900 --> 00:26:33.799 because of how data protection rules are interpreted. 00:26:34.339 --> 00:26:36.079 For instance, researchers have been told to get 00:26:36.079 --> 00:26:38.119 updated written consent from every single patient 00:26:38.119 --> 00:26:40.420 for long -running datasets, even if generic consent 00:26:40.420 --> 00:26:42.779 was given and data anonymized. That can render 00:26:42.779 --> 00:26:44.880 vital research completely unviable. That sounds 00:26:44.880 --> 00:26:47.180 like a major bottleneck. It really is. And it 00:26:47.180 --> 00:26:49.440 leads directly into the complex ethical and legal 00:26:49.440 --> 00:26:52.079 considerations. A major worry is algorithmic 00:26:52.079 --> 00:26:54.720 bias. If AI systems are trained on data that 00:26:54.720 --> 00:26:57.000 isn't representative or reflects historical biases, 00:26:57.500 --> 00:26:59.519 they can inadvertently generate unequal outcomes 00:26:59.519 --> 00:27:02.470 based on race, gender, ethnicity. This could, 00:27:02.470 --> 00:27:04.609 unfortunately, lead to disparities in care quality, 00:27:04.930 --> 00:27:07.329 accessibility for certain patient groups. There 00:27:07.329 --> 00:27:09.430 are urgent calls for much greater transparency 00:27:09.430 --> 00:27:11.769 in how these algorithms are built and robust 00:27:11.769 --> 00:27:14.210 oversight to tackle this threat. Data privacy, 00:27:14.750 --> 00:27:17.430 patient confidentiality are, of course, paramount. 00:27:17.789 --> 00:27:20.210 Entering protected or identifiable patient info 00:27:20.210 --> 00:27:22.930 into public AI systems, that's a clear violation 00:27:22.930 --> 00:27:25.509 of ethical legal obligations. We need robust 00:27:25.509 --> 00:27:28.069 ethical frameworks to guide responsible AI deployment, 00:27:28.450 --> 00:27:30.349 always prioritizing patient dignity, autonomy, 00:27:30.589 --> 00:27:33.119 privacy rights, Transparent, accessible, informed 00:27:33.119 --> 00:27:35.420 consent processes are also vital, especially 00:27:35.420 --> 00:27:37.259 for people with limited English proficiency, 00:27:37.599 --> 00:27:39.339 ensuring patients fully understand how their 00:27:39.339 --> 00:27:41.480 data will be used. That has to align with rules 00:27:41.480 --> 00:27:44.920 like GDPR and NDP. The transparency aspect feels 00:27:44.920 --> 00:27:47.460 particularly crucial doesn't it? If a doctor 00:27:47.460 --> 00:27:50.460 can't actually explain how an AI reached a conclusion, 00:27:50.960 --> 00:27:53.680 That surely arraigns trust. Absolutely. There's 00:27:53.680 --> 00:27:57.200 a real risk of this automation bias or over -reliance 00:27:57.200 --> 00:28:00.000 on AI we talked about, where over -dependence 00:28:00.000 --> 00:28:02.220 could diminish human intuition, clinical skills. 00:28:02.740 --> 00:28:05.019 If the tech fails or hits a unique situation 00:28:05.019 --> 00:28:07.599 the AI wasn't trained for, that over -reliance 00:28:07.599 --> 00:28:10.220 could have severe consequences. Clinicians need 00:28:10.220 --> 00:28:12.539 to understand the underlying logic, not just 00:28:12.539 --> 00:28:15.529 blindly accept if the model says so. This lack 00:28:15.529 --> 00:28:17.950 of transparency, the black box problem, is a 00:28:17.950 --> 00:28:20.289 significant barrier to trust and widespread adoption. 00:28:20.630 --> 00:28:22.549 Clinicians must be able to critically evaluate 00:28:22.549 --> 00:28:25.170 AI recommendations, just like any other diagnostic 00:28:25.170 --> 00:28:28.069 test. And the operational side, beyond data and 00:28:28.069 --> 00:28:30.869 ethics. Yes, several operational challenges too. 00:28:31.509 --> 00:28:34.089 The inconsistent application of AI tech, due 00:28:34.089 --> 00:28:37.009 to a lack of clear unified regulations, creates 00:28:37.009 --> 00:28:39.410 uncertainty for developers and healthcare providers 00:28:39.410 --> 00:28:42.849 alike. While AI can create new jobs, there's 00:28:42.849 --> 00:28:45.369 also potential for job displacement as AI takes 00:28:45.369 --> 00:28:48.369 on tasks humans used to do. That needs proactive 00:28:48.369 --> 00:28:51.269 planning, retraining initiatives. There's a pressing 00:28:51.269 --> 00:28:53.910 need for systemic oversight in AI research and 00:28:53.910 --> 00:28:56.170 development generally to prevent exploitation, 00:28:56.670 --> 00:28:59.109 ensure equitable outcomes for all patient demographics, 00:28:59.730 --> 00:29:01.630 and fundamentally, even the most sophisticated 00:29:01.630 --> 00:29:04.549 AI can't function without high -quality ground 00:29:04.549 --> 00:29:07.460 truth data. That means accurate labeling, tagging 00:29:07.460 --> 00:29:10.559 of orthopedic images by human experts is absolutely 00:29:10.559 --> 00:29:12.779 fundamental for training reliable AI models. 00:29:13.259 --> 00:29:15.099 That often requires significant human effort 00:29:15.099 --> 00:29:17.220 and expertise right at the start. These are definitely 00:29:17.220 --> 00:29:18.960 significant challenges, but you also mentioned 00:29:18.960 --> 00:29:21.619 orthopedics seems uniquely positioned somehow 00:29:21.619 --> 00:29:24.059 to overcome them. What makes this field such 00:29:24.059 --> 00:29:26.660 fertile ground for AI and what are the key steps 00:29:26.660 --> 00:29:28.619 needed to move forward and really unlock its 00:29:28.619 --> 00:29:31.529 potential? Orthopedics is indeed uniquely fertile 00:29:31.529 --> 00:29:33.769 ground for this kind of technological innovation. 00:29:34.309 --> 00:29:36.650 It tends to have clear, well -defined pathways 00:29:36.650 --> 00:29:39.349 for common diseases, which helps with data analysis, 00:29:39.710 --> 00:29:41.769 and it's inherently a highly technical field, 00:29:42.289 --> 00:29:45.009 always embracing innovation in devices, procedures, 00:29:45.700 --> 00:29:48.779 As Professor Xavier Griffin from Queen Mary University 00:29:48.779 --> 00:29:51.279 of London noted, orthopedics is always at the 00:29:51.279 --> 00:29:53.400 vanguard when it comes to using large data sets 00:29:53.400 --> 00:29:56.140 and embracing new tech. Yet, despite this natural 00:29:56.140 --> 00:29:59.039 fit, it often gets overlooked in broader AI health 00:29:59.039 --> 00:30:01.859 care discussions. A landmark NHSX report, for 00:30:01.859 --> 00:30:04.319 instance, didn't even mention orthopedics. Cancer, 00:30:04.599 --> 00:30:06.720 drug development often get the headlines, the 00:30:06.720 --> 00:30:08.900 investment. And that, I believe, is a missed 00:30:08.900 --> 00:30:11.420 opportunity because even small gains in orthopedics 00:30:11.420 --> 00:30:13.599 can lead to better outcomes for huge patient 00:30:13.599 --> 00:30:16.279 populations globally. just given the sheer volume 00:30:16.279 --> 00:30:19.160 of musculoskeletal conditions. So how do we capitalize 00:30:19.160 --> 00:30:21.839 on that potential? What's being done? Well, current 00:30:21.839 --> 00:30:24.099 initiatives are crucial. Orthopedic Research 00:30:24.099 --> 00:30:27.819 UK, ORUK, is playing a key role here, encouraging 00:30:27.819 --> 00:30:30.720 AI application through education, strategic research 00:30:30.720 --> 00:30:32.920 investment, pushing for data standardization, 00:30:33.180 --> 00:30:35.400 fostering networking across the field, really 00:30:35.400 --> 00:30:37.680 calling on the whole MSK community to get involved. 00:30:38.220 --> 00:30:40.700 A shining example of data -driven success already 00:30:40.700 --> 00:30:43.819 is the Getting it Right First Time program, GRLFT. 00:30:43.960 --> 00:30:47.019 It leverages data sharing across orthopedic departments 00:30:47.019 --> 00:30:50.539 to identify best practices, improve care, and 00:30:50.539 --> 00:30:53.180 GRFT has led to remarkable results, reduced revision 00:30:53.180 --> 00:30:55.759 rates, shorter hospital stays by a fifth, releasing 00:30:55.759 --> 00:30:59.039 over 368 ,000 bed days, and has delivered staggering 00:30:59.039 --> 00:31:02.079 NHS savings 696 million over just five years. 00:31:02.660 --> 00:31:04.519 AI can further inform decision -making within 00:31:04.519 --> 00:31:07.319 programs like GRFT by spotting those nuanced 00:31:07.319 --> 00:31:10.119 patterns, helping remove existing biases in resource 00:31:10.119 --> 00:31:13.579 allocation or patient care pathways. give us 00:31:13.579 --> 00:31:15.980 some more concrete examples how AI is already 00:31:15.980 --> 00:31:17.940 making a difference perhaps beyond the research 00:31:17.940 --> 00:31:20.500 labs in everyday orthopedic practice patient 00:31:20.500 --> 00:31:23.400 lives. Certainly. predictive modeling, for instance. 00:31:23.799 --> 00:31:25.880 It's already identifying hip replacements at 00:31:25.880 --> 00:31:28.660 risk of failure. And this isn't just stats. AI 00:31:28.660 --> 00:31:30.960 is spotting subtle changes on x -rays missed 00:31:30.960 --> 00:31:33.279 by the human eye, proving more accurate even 00:31:33.279 --> 00:31:36.240 than experienced clinicians sometimes. This helps 00:31:36.240 --> 00:31:38.160 flag problems early, preventing catastrophic 00:31:38.160 --> 00:31:40.920 failures, yes, but also potentially saving the 00:31:40.920 --> 00:31:43.779 NHS tens of millions by reducing the need for 00:31:43.779 --> 00:31:45.920 routine screening of joints that are highly unlikely 00:31:45.920 --> 00:31:49.160 to fail anyway. AI is also assisting with swift, 00:31:49.539 --> 00:31:51.700 accurate classification of fractures. That's 00:31:51.700 --> 00:31:54.039 vital given the significant backlog of unreported 00:31:54.039 --> 00:31:56.940 x -rays in the UK, estimated over 300 ,000 waiting 00:31:56.940 --> 00:32:00.000 more than 30 days. AI can help triage these, 00:32:00.119 --> 00:32:02.279 ensure urgent ones get reviewed quickly, reducing 00:32:02.279 --> 00:32:04.819 delays in care. That backlog figure is shocking. 00:32:05.160 --> 00:32:08.259 It really highlights the need. Oro UK has even 00:32:08.259 --> 00:32:11.440 taken an equity stake in radii devices, a promising 00:32:11.440 --> 00:32:14.839 startup using AI and biomechanical modeling to 00:32:14.839 --> 00:32:17.259 design better fitting prosthetic sockets for 00:32:17.259 --> 00:32:20.119 amputees. The aim is to drastically reduce the 00:32:20.119 --> 00:32:22.279 number of clinical visits needed for a comfortable 00:32:22.279 --> 00:32:24.660 fit, potentially from up to nine visits down 00:32:24.660 --> 00:32:27.240 to far fewer. That dramatically improves patient 00:32:27.240 --> 00:32:29.900 comfort, speeds up their rehab. And then there's 00:32:29.900 --> 00:32:32.079 the widespread use of wearables in patient apps, 00:32:32.119 --> 00:32:34.559 things like Get You Better, Good Boost, the Versus 00:32:34.559 --> 00:32:37.880 Arthritis AVA Chatbot. They harness AI to provide 00:32:37.880 --> 00:32:39.799 personalized self -management support at home. 00:32:40.240 --> 00:32:42.920 This reduces the need for GP appointments, physio 00:32:42.920 --> 00:32:45.559 referrals, even medication by empowering patients 00:32:45.559 --> 00:32:48.880 with tailored advice exercises. Our Oreo case 00:32:48.880 --> 00:32:51.660 survey showed an overwhelming 91 % of the MSK 00:32:51.660 --> 00:32:54.279 community agrees passive continuous monitoring 00:32:54.279 --> 00:32:57.019 can significantly aid post -treatment recovery, 00:32:57.680 --> 00:32:59.480 underscores the value of these AI tools. Those 00:32:59.480 --> 00:33:01.720 are really impressive practical applications 00:33:01.720 --> 00:33:04.240 showing direct benefits. So looking ahead now, 00:33:04.240 --> 00:33:06.539 what are the key ingredients needed to accelerate 00:33:06.539 --> 00:33:08.700 AI's integration, maximize its impact across 00:33:08.700 --> 00:33:11.400 orthopedics, sports medicine? Several crucial 00:33:11.400 --> 00:33:14.000 elements are needed for us to really unlock AI's 00:33:14.000 --> 00:33:17.400 full potential, I think. Firstly, Continued enhancements 00:33:17.400 --> 00:33:19.420 in machine learning and natural language processing. 00:33:19.859 --> 00:33:22.900 Advancements in ML algorithms mean faster, more 00:33:22.900 --> 00:33:25.460 accurate diagnoses from complex medical images, 00:33:25.740 --> 00:33:29.180 X -rays, MRIs, CTs. NLP will streamline data 00:33:29.180 --> 00:33:31.200 access, translate vast amounts of unstructured 00:33:31.200 --> 00:33:34.039 medical jargon into actionable insights, significantly 00:33:34.039 --> 00:33:36.160 improve patient management through speech recognition, 00:33:36.720 --> 00:33:39.319 efficient record processing, huge operational 00:33:39.319 --> 00:33:42.140 efficiencies there. Secondly, interdisciplinary 00:33:42.140 --> 00:33:45.240 collaboration. Absolutely crucial. We need seamless 00:33:45.240 --> 00:33:47.420 cooperation between clinicians, data scientists, 00:33:47.700 --> 00:33:50.380 engineers, policymakers, and importantly, patients 00:33:50.380 --> 00:33:53.259 themselves. As Dr. Justin Green in Orthopedic 00:33:53.259 --> 00:33:55.759 Registrar rightly said, we need to involve genuine 00:33:55.759 --> 00:33:58.099 data scientists. They bring skills clinicians 00:33:58.099 --> 00:34:00.660 no matter how tech savvy cannot replicate. This 00:34:00.660 --> 00:34:03.000 collaboration ensures AI solutions are both technically 00:34:03.000 --> 00:34:05.039 sound and clinically relevant. Makes sense. You 00:34:05.039 --> 00:34:07.960 need both sides of the coin. Exactly. Thirdly, 00:34:08.119 --> 00:34:10.599 there's an urgent need for AI education. Our 00:34:10.599 --> 00:34:13.400 ORR UK survey highlighted that startling figure, 00:34:13.860 --> 00:34:16.460 75 % of MSK professionals feel they lack the 00:34:16.460 --> 00:34:18.699 knowledge to apply AI and big data effectively. 00:34:19.320 --> 00:34:21.519 We need structured, comprehensive training tailored 00:34:21.519 --> 00:34:24.780 to MSK professionals. ORR UK is actively working 00:34:24.780 --> 00:34:27.340 on developing such a program to upskill the workforce. 00:34:27.960 --> 00:34:30.550 Fourth, research investment. critically lacking. 00:34:31.269 --> 00:34:33.929 71 % of the MSK community agrees there's insufficient 00:34:33.929 --> 00:34:36.909 funding for AI research in MSK health. We must 00:34:36.909 --> 00:34:38.889 strategically fund innovative projects using 00:34:38.889 --> 00:34:41.670 AML to reduce MSK pain, improve function, cut 00:34:41.670 --> 00:34:43.570 costs, and fundamentally enhance quality of life 00:34:43.570 --> 00:34:47.469 for millions. Fifth, data standardization. Paramount, 00:34:47.570 --> 00:34:49.769 championing better data standards for MSK research, 00:34:50.150 --> 00:34:52.250 including improved reporting of AI studies, addressing 00:34:52.250 --> 00:34:54.570 eligibility criteria, ground truth data, data 00:34:54.570 --> 00:34:56.989 set diversity model details, the need for rigorous 00:34:56.989 --> 00:34:59.550 external validation, larger sample sizes. This 00:34:59.550 --> 00:35:01.449 will help counter biases from narrow data sources, 00:35:01.630 --> 00:35:03.530 making AI models more robust, more equitable. 00:35:04.289 --> 00:35:06.489 So better data, better reporting. Precisely. 00:35:07.050 --> 00:35:09.989 And finally, we need proactive regulatory frameworks, 00:35:10.489 --> 00:35:12.469 ones that strike that crucial balance between 00:35:12.469 --> 00:35:14.789 fostering innovation and ensuring patient protection. 00:35:15.039 --> 00:35:18.340 They must ensure AI advancements contribute positively 00:35:18.340 --> 00:35:20.940 to public health without infringing on individual 00:35:20.940 --> 00:35:23.920 rights, and crucially, longitudinal studies and 00:35:23.920 --> 00:35:25.599 validation across different sports environments, 00:35:25.860 --> 00:35:28.500 diverse athlete groups. That's absolutely essential 00:35:28.500 --> 00:35:31.179 to improve accuracy, generalizability, and finally 00:35:31.179 --> 00:35:33.739 bring these groundbreaking innovations from promising 00:35:33.739 --> 00:35:36.699 research into widespread, impactful clinical 00:35:36.699 --> 00:35:38.869 practice. Right, now for our lightning round. 00:35:39.050 --> 00:35:41.590 Quick, sharp insights. What's one AI technology 00:35:41.590 --> 00:35:44.449 you believe is poised to make the biggest immediate 00:35:44.449 --> 00:35:47.909 impact in orthopedics in, say, the next 12 to 00:35:47.909 --> 00:35:51.280 18 months? I'd have to say AI -powered diagnostic 00:35:51.280 --> 00:35:53.679 imaging for faster, more accurate detection, 00:35:54.320 --> 00:35:56.760 coupled with predictive analytics for early intervention, 00:35:57.159 --> 00:35:59.139 particularly around complications or predicting 00:35:59.139 --> 00:36:01.360 outcomes. That's where the immediate wins are. 00:36:01.460 --> 00:36:03.820 Okay. And for an aspiring athlete, maybe even 00:36:03.820 --> 00:36:06.119 a weekend warrior like me, looking to leverage 00:36:06.119 --> 00:36:09.059 AI for personal injury prevention, what's one 00:36:09.059 --> 00:36:11.340 practical, actionable step they could take today? 00:36:11.579 --> 00:36:14.840 I'd say start by embracing AI -powered wearable 00:36:14.840 --> 00:36:17.519 technology if it's accessible. Use devices that 00:36:17.519 --> 00:36:20.400 monitor your real -time biomechanics, your workload. 00:36:20.900 --> 00:36:23.380 Look for apps or platforms offering personalized 00:36:23.380 --> 00:36:26.800 feedback based on that data. Companies like MAI 00:36:26.800 --> 00:36:28.960 Motion show how using this data for customized 00:36:28.960 --> 00:36:32.079 rehab, predictive analytics can really cut recovery 00:36:32.079 --> 00:36:34.400 time and prevent injuries. It gives you a powerful 00:36:34.400 --> 00:36:37.119 edge. Good advice. Finally, for any healthcare 00:36:37.119 --> 00:36:39.460 professional or sports scientist out there feeling, 00:36:39.460 --> 00:36:42.159 well, maybe a bit overwhelmed by the sheer pace 00:36:42.159 --> 00:36:44.940 of AI change, what single piece of advice would 00:36:44.940 --> 00:36:48.239 you offer them? I'd say Embrace AI as augmented 00:36:48.239 --> 00:36:50.880 intelligence, not as a replacement. Focus on 00:36:50.880 --> 00:36:52.840 fostering that interdisciplinary collaboration 00:36:52.840 --> 00:36:55.340 with data scientists and commit to continuous 00:36:55.340 --> 00:36:57.599 learning, especially understanding AI ethics 00:36:57.599 --> 00:36:59.760 and how to critically interpret AI -generated 00:36:59.760 --> 00:37:02.159 data. Your human expertise remains invaluable. 00:37:02.340 --> 00:37:04.840 AI just makes it potentially more powerful. OK, 00:37:04.840 --> 00:37:07.699 brilliant. So let's try and crystallize our key 00:37:07.699 --> 00:37:11.320 takeaways from today's Deal Dive. Firstly, artificial 00:37:11.320 --> 00:37:13.920 intelligence is fundamentally an augmentative 00:37:13.920 --> 00:37:16.989 tool. It's designed to... enhance not replace 00:37:16.989 --> 00:37:19.650 the expertise of clinicians and sports scientists 00:37:19.650 --> 00:37:22.650 leading to greater precision, greater efficiency 00:37:22.650 --> 00:37:26.469 in care. Secondly, AI's applications are vast. 00:37:26.829 --> 00:37:29.010 They span the entire patient journey, the athlete 00:37:29.010 --> 00:37:31.449 life cycle from predictive diagnostics, personalized 00:37:31.449 --> 00:37:33.670 treatment planning, right through to real -time 00:37:33.670 --> 00:37:36.690 monitoring and advanced rehab. Thirdly, while 00:37:36.690 --> 00:37:38.730 the benefits are immense, successful integration 00:37:38.730 --> 00:37:41.030 really hinges on diligently tackling those crucial 00:37:41.030 --> 00:37:43.690 challenges, ensuring access to high -quality, 00:37:43.750 --> 00:37:46.510 diverse data, navigating complex ethical and 00:37:46.510 --> 00:37:49.510 privacy concerns, and establishing clear, transparent 00:37:49.510 --> 00:37:52.190 regulatory frameworks. Fourthly, orthopedics 00:37:52.190 --> 00:37:54.170 in particular stands out as uniquely fertile 00:37:54.170 --> 00:37:57.070 ground for AI innovation. It offers immense potential 00:37:57.070 --> 00:37:59.449 for significant cost savings and vastly improved 00:37:59.449 --> 00:38:01.809 outcomes for huge patient populations, provided 00:38:01.809 --> 00:38:03.650 there's continued strategic investment and targeted 00:38:03.650 --> 00:38:06.170 education. And finally, the future of AI in medicine 00:38:06.170 --> 00:38:08.429 and sports demands unwavering interdisciplinary 00:38:08.429 --> 00:38:11.369 collaboration and a steadfast commitment to integrating 00:38:11.369 --> 00:38:14.329 this technology ethically, responsibly, ensuring 00:38:14.329 --> 00:38:16.730 it always complements and elevates human expertise. 00:38:17.179 --> 00:38:19.679 If you found this deep dive valuable today and 00:38:19.679 --> 00:38:21.780 gained some new insights, please do consider 00:38:21.780 --> 00:38:23.900 rating and sharing it with a colleague or a friend 00:38:23.900 --> 00:38:25.659 who might also benefit from these discussions 00:38:25.659 --> 00:38:28.820 on AI's impact in healthcare and sports. Thank 00:38:28.820 --> 00:38:30.659 you so much for illuminating this incredibly 00:38:30.659 --> 00:38:33.019 dynamic field for us today. It's been truly insightful. 00:38:33.519 --> 00:38:35.420 And thank you, our listener, for joining us on 00:38:35.420 --> 00:38:37.460 the deep dive. Do subscribe to make sure you 00:38:37.460 --> 00:38:39.340 don't miss our next exploration into the fascinating 00:38:39.340 --> 00:38:42.360 world of knowledge and insight. Now, as AI becomes 00:38:42.360 --> 00:38:44.559 ever more intertwined with our health, just think. 00:38:45.119 --> 00:38:47.519 How might our understanding of human performance, 00:38:47.599 --> 00:38:50.599 of recovery, evolve when every subtle change, 00:38:50.739 --> 00:38:53.699 every nuanced response can be quantified, understood 00:38:53.699 --> 00:38:56.619 by algorithms? What new frontiers will this precise 00:38:56.619 --> 00:38:58.539 understanding unlock for human potential?