Season 2, Episode #3 - AI for Orthopedic Surgery - Dr Michael Murphy

Dr Michael Murphy, founder of Mortho, joins the show.

Mortho’s mission is clear: to democratize orthopedic intelligence and make AI-powered surgical planning more accessible to the surgeons and patients who need it most.

With a rare background spanning engineering, software development, machine learning, and orthopedic surgery, Michael is building technology from the inside out. His perspective is especially compelling because he’s not just building for clinicians — he is one.

We discuss -

• How Michael’s early background in engineering and machine learning led him into medicine and ultimately orthopedic surgery
• The problem of implant identification and why revision surgery has historically involved a surprising amount of guesswork
• How Mortho uses AI, imaging and workflow automation to support surgeons before, during and after surgery
• Why Michael made the platform free for most users, and his mission to democratize access rather than build a closed moat
• The challenge of winning trust from skeptical surgeons in an industry flooded with new AI tools
• Why security, compliance and real clinical understanding matter just as much as the technology itself
• The differences between building for healthcare versus simply building cool technology
• How Michael thinks about compute costs, scalability, and keeping AI tools affordable over time
• Why he believes even failed startups can still push an industry forward

A fascinating conversation blending frontline clinical insight, technical depth, and an unusually thoughtful vision for the future of AI in orthopedics.

https://www.linkedin.com/in/charlie-webb-a5605768/
https://www.linkedin.com/in/sumair-gill-086a151a1/

Transcript

SPEAKER_01 0:00

Welcome back to the Smart Health podcast. After a brief hiatus, we're delighted to return with Michael Murphy, the founder of Mortho. Now, Mortho is an AI platform for orthopedics that helps surgeons identify implants, streamline surgical planning, and support better patient care. They're essentially helping make surgical planning faster, smarter, and more accessible. Now, Michael is a really impressive guy. He's a practicing orthopedic surgeon, currently at Illinois Bone and Joint Institute, with a background in machine learning and software engineering. He's incredibly passionate about bringing truly value add AI into what is a sceptical industry, and he really understands the practicalities of what's possible and what's actually useful on the front line. We discuss the realities of building AI in healthcare from a clinician's perspective, what clinicians are looking for in technology, how to build trust in that skeptical environment, the crucial role of security, which was really interesting, and of course, much more. With that said, let's get into it. Hi Michael, thanks so much for joining us on the Smart Health Podcast. Great to have you.

SPEAKER_03 1:15

Thanks, Charlie. It's good to be here.

SPEAKER_01 1:18

It's a bit of a different niche for us, orthopedic surgery, not something we've touched so far. So exciting to hear how AI and technology can add value in that sector. But as always, it'd be great to hear a bit more about your background. People do like to hear what was the background leading up to becoming uh a health tech entrepreneur. So please, wherever you see fit, take it away about your background leading up to founding the company.

SPEAKER_03 1:46

Yeah, appreciate it. So um I always kind of start off like most people in my family are engineers. So I always kind of grew up being told you could be anything that you want to be, but you're gonna be a great engineer. So it was kind of a decision of what kind of engineer I was gonna be. So when I initially went to school, I wanted to do chemical engineering and I wanted to study uh nuclear power plants and kind of solve nuclear fusion. Um, unfortunately, where I went, the program closed, so I had to pivot. And I went into biomedical engineering with the goal of doing biofuels. Since then, I kind of fell in love with implant design, and then when I was starting that, I learned that the orthopedic surgeons have the cool job. So along the way, I was starting to do a lot of software development, and I was kind of doing machine learning before it was really cool to say that you were doing machine learning. It was a newer technology. Uh, the people who I worked with were critical of it, which was a really good experience for me because I had to prove that it worked, both to them and to myself, right? Um, that's what that's what all this is about. We got to see the data first, and I'm a huge proponent of that, you know. Seeing is believing. So since then I transitioned to medical school uh knowing that I wanted to do orthopedics, and I began merging that technology into orthopedics.

SPEAKER_01 3:26

Nice, that's a really interesting background. So it was almost a bit of fate that got you into it with that course being pulled. And I like that you said you were convincing people early on of the merits of technology because there is certain Luddite behavior in the industry, uh, or indeed just general skepticism, which I'm sure we'll discuss further further down the line. Um, so you went into that niche. What was it within orthopedic surgery that inspired you to build the company? You had that machine learning background before it was cool, but what problem did you see that inspired you to found Morpho?

SPEAKER_03 4:09

Yeah, um, really good question. So when I was doing software development, some of the things that I really enjoyed was an area called signal processing. Okay, so you basically you have some sort of data, some signal, okay, and you process it and you interpret it. So the things that I was working on were things like an EKG interpretation system. That way you could have a live feed of an EKG, which is a signal, and you interpret it uh through some computer software, you feed it back to the user, and so on, um, which was very exciting for me. Um some of the other things that I did were early on, like uh computer vision, and that is basically just a two-dimensional version of an EKG, right? You have numbers and data for an EKG, and then you have a second set, and now you have an image. Okay. So when you apply this to orthopedics, what I would do is, you know, I kind of got told in residency, hey, if you're gonna excel in orthopedics, you kind of have to publish a lot. You need to do a lot of you need to do a lot of literature and publish, get your name out there, get recognized. So some of the early publications that I was working on were in the computer vision area. So things like implant identification. I was one of the first people to uh identify implants from an X-ray, which is important. If you're planning to do a revision surgery, then you have to know what implant is there. So I was um one of the first people to publish on that. Some of the other things were using uh machine learning, artificial intelligence, computer vision to determine implant orientation. So it kind of helps the surgeons. They are deciding all the time is this a patient who I need to do surgery on, or is this a patient who you know we can do non-operative care? Part of that assessment is deciding, hey, are the implants in good alignment? Are they safe? And you can use computer vision to do all that stuff. So this has been like a passion of mine uh for a long time, and I've really been interested in the in the technology. And then, you know, orthopedics is a great space to do vision tasks. Um if you're doing vision type tasks, you're kind of you're a little bit more limited in terms of the specialties that you would want to go into. Radiology, of course, right? Orthopedics is very heavy on vision, pathology, dermatology. Um, those ones are very heavy on vision. Um so orthopedic surgeons are are all the time, you know, they're saying, okay, did we get an x-ray? You know, uh CT scan, MRI. That stuff is is very regular for them. So it was it was a space that I thought that I could have a big impact on, and I was already seeing even in medical school, I was having that impact and was getting a lot of attention.

SPEAKER_01 7:13

Nice. So for people who know nothing about the space, me included, before you, and it's really impressive that you were the first, one of the first to publish on the difference between implants and and normal joints, before that happened, what was the process? That's always kind of a good question before and after um scenario to and fully understands the impact of your eventual products. What was the process in identifying implants before? Was it guess what? Do they look really similar on an X-ray image to the naked eye?

SPEAKER_03 7:50

Yeah. Um, so typically the process is uh keep in mind there are thousands of implants. We've been making implants for uh quite some time, putting implants and knees pretty regularly around 1950 and up. Um, so there have been many, many iterations of implant designs for the hip and for the knee, um, thousands of them. So typically, what the process is is you have a patient who comes in and the surgeon decides, hey, there's something wrong here, we need to change it up through surgery. With that, you need to decide what tools do I need to get out the current implant, and then of course, what tools do I need to put in the new one. So the the old process would be you take an X-ray, you look at it, and from the morphology, you hopefully have a good idea as to what it is. Now, with thousands of implants, there's a ton of overlap, right? You have these companies that are trying to do two sets of things, right? You have the patent sector, and they're saying, Hey, uh, our implant is very unique. And then you have the FDA standpoint, which says, Hey, it's actually the same thing. And then you have the marketing standpoint that says, Hey, we have a very similar version, it looks very, very similar, but uh it's actually just a little bit different, and now it's cheaper. Okay, that's an easy marketing strategy. So, with that, you can obviously see that there are many iterations that look very similar and many iterations that look very different. So, when a surgeon makes that decision, often what they do is they take a picture and then they start asking their friends. And it's literally just like looking at it, and you're like, well, it could be one of these, like you know, maybe we narrow it down to 10, or we're a little bit of guesswork. I think the surgeon was using this company at this time. Uh, you know, that sort of thing. We look at the operative report. Hopefully they wrote it down. Not always. Um they're they're not always that descript. Um so it's kind of all those things. Today, with uh with the technology that I produced, uh, I was actually the first person to make it openly available. Okay, so I I was the first person to do the research. Someone beat me to the punch for getting it into a uh a journal, but I certainly was the first person to make it openly available. So today you can use AI and it gives you its recommendation on what it thinks the implant is, and it also gives you a confidence score. So it says, hey, I'm like pretty confident, or look, I'm seriously just guessing here. You you gotta double check me.

SPEAKER_01 10:39

Nice. And not to get into the technical weeds too early, but it's a good off-ramp to touch on it. How did the AI learn to predict and give itself confidence? Is it trained on past surgeries, lots of imaging data, or something else?

SPEAKER_03 10:55

Yeah, yeah, exactly. So we have a data set and a growing data set as part of um the Mortho platform. So it has its own data set and uh and a growing data set as well. That's part of the reason why I wanted to build Mortho the way that it is, is the way that it works, it's actually able to iterate on itself, which is very unique. You'll see that there since Mortho came out, there are these other players that are starting to emerge. Uh but um, you know, Mortho is unique in that the data that it also captures is very unique, and I'm I'm very confident that it will secure itself as the best because of that.

SPEAKER_01 11:40

Nice. I like I like the confidence. So I guess going back to the initial problem that you're solving, there's a lot of guesswork without this technology. What does that guesswork lead to down the line? Like what problems is that causing being a little bit in the dark?

SPEAKER_03 11:57

Yeah, so that's a really good question, also. Um I can tell you're well versed in this area.

SPEAKER_01 12:05

Don't know about that. I've done a lot of podcasts. Appreciate it.

SPEAKER_03 12:08

Yeah. Um, so a lot of times what happens if you don't have the right uh tools to extract the implant, you kind of have to go bigger. Okay. And you can always kind of cut out the bone and remove the implant that way. But that's that has its own complications, right? So if you cut out the bone, there's always the chance that the bone doesn't heal. Okay. You the more work that you're doing in the area, the higher risk for things like infection, you know, damaging the muscles could be permanent damage as well. Uh, you could damage nerves, uh, so on. So all those risks increase as you're as you're beginning to do this stuff.

SPEAKER_01 12:55

Interesting. So from a purely hard-nosed capitalist perspective, which we have to talk about when you know selling technology, hospitals in the US tend to care about the bottom line, that safety issue could lead to insurance issues, legal issues, and further surgery and further treatment, not to mention reputational issues, if it's left untouched.

SPEAKER_03 13:19

Yeah. Um maybe. So with regards to the bottom line, since you brought it up, the the way that the surgery works from the physician standpoint is we get paid for specific uh specific surgeries, right? So even if we have like a really long surgery, uh, but for example, a regular total hip replacement. Let's say we're doing a regular total hip replacement in someone who has some really complex anatomy, it's still the same billing code as someone who has a regular uh, you know, kind of more normal or more something that we're more comfortable with, um, anatomy. So we would get paid the same with that. Now everyone else might get paid more or less, but from the surgeon's standpoint, we would get paid the same because it's the same, the same billing code. So from the surgeon's standpoint, I would say the primary thing that they're thinking about and why they're so incentivized to prepare is that that type of surgery could easily turn from a two-hour surgery to a six-hour surgery or an eight-hour surgery. And now the patient's very unhappy because they are, you know, understandably so. They could be permanently damaged because of the things that you had to do because you were more or less unprepared.

SPEAKER_01 14:43

Got it. So when you're going in and selling the product, it's pretty it's an obvious return on investment in the sense that this could become X, which will be a two-hour surgery turning into an eight-hour surgery. And they've probably seen this before, they've already got the past pain from not having this in place. So it's quite an easy sell in that respect when you're going and speaking to surgeons. Or is there is there quite a lot of skepticism? Maybe that's a good uh time to touch on the Luddite skepticism side of the industry. How have you found speaking with your fellow surgeons about it so far?

SPEAKER_03 15:20

Yeah, I think that you know artificial intelligence is growing so fast that um there is a healthy amount of skepticism. And uh, you know, I am the biggest skeptic of myself, right? Uh, but I have a background where I'm you know very comfortable with math and statistics and so on. So um I've learned to trust the things that I do and and prove that it works. That's kind of how that happens. Um, you mentioned uh the cost associated with it. So the platform is free for surgeons and for even for most people. Um, I want people to see and begin to trust it. Uh that's that's how it works. And then um just to highlight uh some of the advantages of Mortho to show you how this data is um kind of it it comes for full circle. So what Mortho does is it's it's primarily a logistics platform. Okay, so as an orthoplasty surgeon, someone who does total hips and total knees, what I do is I see a patient in clinic, okay, and Mortho has an AI chatbot. Okay, so right from seeing the patient in clinic, I can start using Mortho, and it has a a free AI chatbot. Some of the things that Mortho does to that chatbot that's unique is it's the first AI chatbot, at least that I know of, where you can throw in an image. We already talked about how important that is for physicians. Uh, you can throw in an image, you can throw in an intake form for the patient. If they're a new patient, you can have multiple practitioners uh enter and exit, and that is very unique. If you're to do that in any other platform, you typically have to pay per seat. So uh other platforms charge you you know hundreds of dollars a month, and they say per you know, medical assistant person you charge per seat. And I'm comfortable with the compute on this, so I was able to lower those costs a lot. That's how I'm able to offer it for free. So we see the patient, we sign them up, right? Now, Mortho has uh automated messages that go directly to the patient. We can say pre-operative communication to the patient, stop this at this time, whatever the surgeon wants. You have uh the surgery day, right? So from the information that we collect from the patient, the x-rays, Mortho, what it will do is it will predict what implants it thinks you are going to use. Day of surgery, the sales rep or the surgeon will provide feedback to Mortho. Hey, this is what we actually used, and now the algorithm is learning. If you upload post op x-rays, now we have a picture of what that implant looks like. Now the implant identification algorithm is learning, and it's verified, right? That's the unique thing compared to other implant identification systems. And then we have the post-op recovery. So from there, Mortho has videos on post-op recovery, physical therapy exercises for the patient, um, and all those things are associated with, of course, better patient satisfaction. Uh, patients like it a lot. They get their therapy exercises, they can do them from home if they like. Uh, lower rate of them going to the emergency room because they have all this communication and a better understanding, and it's digestible because it's not just a dump of 50 pages that they have to read at one time. It's sent to them in small little snippets, you know, days at a time.

SPEAKER_01 19:11

Nice. So there's a patient facing side to the platform as well, and the chat bot that the how have you found, just briefly on the surgery operating room side, how have you found integrating with the workflow of the surgeon? We hear from radiology that it's very difficult to integrate with the archaic, sometimes a lot of red tape, hospital system. How's it been on the hospital on the surgeon side?

SPEAKER_03 19:37

Yeah, so um that's part of the reason why Mortho is structured the way that it is. So I I totally agree. Like hospital systems, there is a lot of red tape for good reason. You know, we need to be safe about things. It certainly slows things down a lot. Um what Mortho when when I was first designing Mortho, and I I built Mortho with my dad, uh, who is very familiar with financial technology, which is actually very similar to uh medical technology. So he helped build uh the Illinois State taxes, uh, the database for it. So he's very he's very comfortable with you know money movements and and so on, which is uh incredibly similar to HIPAA compliance and security with uh sending numerical data, right? It's exactly the same thing, okay? So he really helped from the security end. What we decided to do is we wanted to have this overarching security web-based platform. Okay, now if you work on the web, you can work anywhere from there. So we have a phone app, and that makes it easily accessible for surgeons, but that uh phone app is has this um has this really strong uh threshold around it, which is the web architecture. Um and that's really how it works, and it's able to stay secure. And we're up to date on all the security standards. Obviously, right, my dad does this stuff for Illinois State taxes, and we don't want anyone uh getting past that. So same thing with Mortho.

SPEAKER_01 21:27

No, that's really interesting. It makes a lot of sense. We've been working a role recently for kind of a Helltech insurance play, and they have said that people from fintech would be really relevant uh because of that security, that regulatory similarity. So so from a user perspective, what all they need to do is have a phone with the app or log in using whatever browser they have so it doesn't have to be a new application integrated into the old school workflow. It's pretty smart.

SPEAKER_03 21:58

Yeah, no, that's exactly right. You can you can sign up and start using it in seconds, and it's as easy as that.

SPEAKER_01 22:06

And you say it's free to use and that's smart, kind of like a Trojan horse play and the model's learning as it goes on. But unless I missed it earlier, you said something about you get paid when an operation happens using the platform. So is it free until and how does that work, or is it free until you get to a certain level where all surgeons want it, and then you're going to be a bit more serious about uh like a mental payment or what have you?

SPEAKER_03 22:32

Yeah, so that's that's a really good question, too. So in the US, um 60%, somewhere around 60% of total hip and knee replacements are performed by surgeons who do around 200 or less a year. Okay, there's and then there's the other half of surgeons who do you know upwards of a thousand, six hundred to a thousand. Okay. So what Mortho does is is there will be a study coming out soon. It's currently in uh a journal, but what one of the things that I showed was that when you when the surgeon implanted a size, an implant size, because they come in different sizes to match patient anatomy, if they put in a size that was closer to the morto-predicted size, there was a lower complication rate. Okay, so let's talk about um hip replacements real quick, and then I'll I'll get back to your question. Um so if you put in, we have a femur, which is uh kind of like a tube, okay, a PVC pipe, if you will. If you put in something in the center of that tube and it's too big, the tube will crack. Okay, so you fracture the patient. If you put in something too small in that tube, then you'll get windshield wipering, it'll loosen, patients don't like it, it's painful. Okay. So we did a study. How was Mortho doing? Okay, you know, I think that Mortho is the most accurate system out there. Let's prove it. If we look at surgeons who put in sizes closer to the mortho predicted size, what were those outcomes? And we saw that the surgeons who were implanting sizes closest to the mortho predicted size had the lowest rate of fracture, breaking the bone, too big, right? Too big it breaks the bone, and implant loosening. So uh it kind of proves a little bit that mortho is working, which was great and very exciting for me. When I saw that, I thought to myself, I shouldn't really keep this um selective. The people who need this the most are the people who are doing the surgery less frequently. So I didn't want to kind of uh bottleneck who was using it. As you can imagine, the people who would want this the most are the people who are doing the surgery all the time. But the people who probably need it more are the people who are doing it the least. So what Mortho does is, like you mentioned before, is it will allow you to do uh 10 cases a week for free. So that's 520 cases a year, which will capture the uh the majority of physicians in the US. That's why I that's why I did it that way.

SPEAKER_01 25:31

No, but that's really benevolent of you. I was gonna say when I heard the fracture horror story, that kind of puts me off ever having surgery unless something like this is in is in place. Like they make out like it's safe, but who knows? If they if they get I don't have implants, so maybe I'll I'll be alright. Uh anyway, I feel like I've talked far too much as always. So, Samare, I'll hand over to you to ask some of some of your questions.

SPEAKER_02 25:55

Yeah, thanks. Um, I'll be honest, Michael. Like um I think we we've covered a lot of the basics of what Morpho does. One thing I'm really interested in, though, is I think you're one of the few people who we've had on who has got very much uh frontline uh view of you know a clinician, a physician, someone who's working with patients and in and yet has built something from that viewpoint. So when you first conceived of the idea of morpho, did you find that you were using uh any aspect of your surgical understanding and background when it came to producing Mortho? Did it sort of uh uncover any new abilities which you had to rely on, which you previously might have done not done before?

SPEAKER_03 26:49

Yeah. Um yeah, that's that's a really interesting question. I mean, I've always been um incredibly interested in the vision learning space. Um, and you know, since then there has been this just wide adoption of large language models, which are actually very, very similar. So um certainly I've learned a lot about that and all these different compute techniques, which I think is really exciting. Uh, if you ever see me on LinkedIn, I'm always doing these posts that uh that um I don't always know how interested people are, but it's like, you know, I really have just a few people to talk to about that. So it's like, ah, I might as well just throw it out there and maybe I'll maybe I'll make a new friend. Um but yeah, I've I've learned a lot. I'd say, you know, one of the things that like really pushes me to do all this stuff is um I I get a lot of inspiration from the CEO of Anthropic, Dario Amade. He worked for OpenAI, right? The creators of ChatGPT. And when he separated from OpenAI, what he kind of set out to do is he's like, look, I'm not trying to make a ton of money, just like me. I'm not trying to make a ton of money. But what he saw was that he was scared that players like OpenAI and whoever else were going in a direction that maybe wasn't as beneficial as where he thought they needed to go. So he separated, created a separate company, and he talks about this all the time, where he says, you know, the goal here is not to make like a really cool product and to take over another company, it's to push the field in the right direction. So he talks about safety all the time and so on. So Mortho has the same goal. This is the thing that I talk about all the time is what what my goal with Mortho is the phrase that I use is to democratize the technology, okay, to make it accessible to all the implant companies, to all the surgeons. And you know, I'm seeing that there are complication benefits with it. So that's why I just want this available. One of the things that has happened, especially in the US, is companies get kind of like a moat on their uh product market fit and where they are in the market, their um their market capture. So it incentivizes them not to change as much. And that's really the goal of Mortho. If I democratize this technology, then all everyone has to keep moving forward, right? So I have an AI chatbot that uses uh vision, x-rays, and so on. There, I don't know of another AI chatbot that can do that. So I can guarantee you that when other companies find out about that, they're gonna push to include that. It also is um is significantly cheaper. So that's gonna push other companies to make it cheaper. I don't charge for additional, well, first off, it's free, but the way that it works, it's at no additional expense to me to charge for additional seats, a PA and MA or so on. So other companies are gonna go that route. The other things it does is you know communication between patients and physicians and automating communication between physicians and the sales rep and the hospital. So that's something that is historically very expensive and very challenging. Mortho largely makes that for free. Over 90% of its users use the free platform. So this is you know, this is why I wanted to do it to democratize the technology and kind of push the image, uh, the industry forward.

SPEAKER_02 30:57

See, this is I that is really uh a key aspect which I wanted to get to the bottom of because I think when you are someone who probably you know is working with these patients, is utilizing this sort of uh technology and can see the benefit benefits of it yourself, it does change your approach in helping to push it forward, expand it, and get it out there into uh general usage. Um how difficult has it been? Because I imagine you're getting more people who are utilizing the platform, and as a result, you're getting people who are using X-rays, and and we know what X-rays are like. They, you know, people have different imaging protocols, different KVs that are utilized, uh the patient differentiation. You might have patients who are slightly more obese when they're being scanned, and that can affect image quality. Is that something that you had to take into account when it came to building Moto? And how much of an effect does that play on the actual model itself?

SPEAKER_03 32:03

Yeah, you know, that's that's a really great, uh really great question. When I first launched the implant identification uh system that was freely available, still is freely available. Um, one of the first images that I got from someone was a picture of a cat. Which is not actually an implant. So uh it worked uh not quite well for that type of system. But luckily, I have a lot of very creative users who will uh send you know images of all levels of quality. So the system's always learning and it gets better from that. So yeah, that that was um that's it's been a fun experience, you know, both from people who use the platform and the feedback that I get as well. You know, the thing that has been very interesting in in this process has been honestly when I talk to surgeons. So I I love surgeons, I'm a surgeon, ton of ton of respect for them. Um it's very hard as a surgeon to change the way that you do things. You know, we all even as you know, non-surgeons, people, you know, just in the community, there is just a fire hose of AI opportunities coming at us all the time. Right? Google has their thing, Microsoft has their thing, OpenAI, Anthropic, whatever it is, like there's a new flavor of the day. Okay, it's impossible to try all of them. And when I talk to surgeons, they're like, why would I try yours? And it's I think to myself, it's so obvious, isn't it? Uh, but it's not to them. So um that has been the really fun part for me, uh, because I get to prove it to them, and this is the thing that I love. I know that it I know that it works, and it's just like, okay, how can I word this in a different way to show it to you?

SPEAKER_02 34:21

Okay, yeah, it's it must be a strange feeling because you have inadvertent it's like you said, Morto is something fairly unique in the orthopedic space when it comes to AI uh itself. So, in a way that you've become sort of the poster boy for pushing um the benefits of AI in the space, and also trying to build that trust, which, as you mentioned before, isn't necessarily as strong uh when it comes to AI in the healthcare space. I mean, does that add on any added pressure to yourself in the role that you have? Like you people are coming to you because you've built a platform, but also because you have got the credentials to back up um you know what you've built. So, how how does that change the way you're approaching people? Not only as like, okay, I'm trying to get this product to you, but I'm also a clinician who's got that understanding of it.

SPEAKER_03 35:25

Yeah, I mean, it's been it's been exciting for sure. There's been a lot of exciting stuff. Um, it's certainly a different role. As an engineer, I've always kind of been a little bit more reserved, and I am gaining this uh more and more uh influential role. Um, it's also something that I very strongly believe in, right? So part of the reason why I initially wanted to go into energy, nuclear fusion, biotechnology, biofuels, and so on, is because when I was younger, I heard this phrase that um I still believe that there's two things in the world that really helps everyone in the world. There's two things that when they advance, uh, it helps everyone in the world and it really moves the needle. Those two things are lower cost of energy, whatever the energy is. So that nuclear fusion, biofuels want to solve that problem. And the other one is advancing technology. When you advance technology, the needle moves up, the quality of life as a whole moves up. So as I was kind of deciding to pivot into medicine, I'm like, you know, my passion is gonna be in one of those two fields. And I kind of, as I was doing research, I started falling into this technology area. And I was, you know, to my surprise, happily seeing complication benefits and so on. And, you know, there's there's other obvious things, right? It makes quality of life better for the surgeon, right? Because we no longer have to spend as much time identifying implants, we can communicate more easily, lower expenses, moving implants and so on. Uh patients are better educated. And then there's like this beautiful separate thing that happens with all of this technology, which is uh the complication benefits that that we're seeing as well. Um so it's like and I was kind of surprised to see that. I I was really hoping that would be true, um, but I was very happy when when I saw it. Um so yeah, I mean that's when I think about advancing technology, when I was younger, I always I used to think to myself, like people often say the phrase um work smarter, but not harder. And when I was younger, I always used to think to myself, like, work smarter but hard work smarter, not harder, but you still have to work hard. Like I would I would add on that phrase, work smarter, not harder, but you still have to work hard. And as we're seeing this push for large language models, I actually am starting to believe it that I think you can with this advancing technology, you can actually work smarter and not as hard. And it's beautiful because what these language models are doing are the same thing that Mortho's trying to do. They're trying to democratize uh intelligence. So you push intelligence, readily accessible, smart uh tools at the fingertips of everyone. And people love it. It's so easy, it's so fast, and it's so good. So that's what I'm trying to do with Mortho, just in a different realm, the healthcare sector. And I think to myself, like, you know, there's a decent chance Mortho won't work, right? I hear from CEOs all the time because I'm regularly trying to interface with them. What should I do? Where should I go? What am I doing wrong? What am I doing right? And they always tell me, hey, 90% of startups fail. It's like, okay, great, let's let's do that. Like, let's do that. That's that's what I want to do. Um, but I fall back on Dario Amadei, CEO of Anthropic, who says, yeah, 90% fail, but we pushed everyone else forward along the way. So even if I fail, I I hope that it made an impact.

SPEAKER_02 39:52

Yeah, without a doubt. And it's it's funny just through that answer, you you mentioned there the the in implant assessment that uh Mortho utilizes. I was just wondering because the one thing is not only do you have implants that are all different sizes that you know may play a big part in um what the surgeon's going to use? Um you also have different brands of certain implants as well. How does Morpho assess something as almost as esoteric as like, okay, then this brand compared to this brand compared to this brand, and also taking into account okay, this size compared to this size compared to this size? Does that generate some difficulty? I mean, does that change the way that you're approaching the data? Because one hospital might favor one certain brand and they're sending you this data, and another hospital's sending something completely different.

SPEAKER_03 40:50

You know, you guys uh impressed me so much. You guys asked some incredibly insightful questions. I I'm not gonna lie. So that is uh a great question, and it is partly what makes Mortho uh so like recognized. Okay. So there is one implant uh platform that can work with any implant company, and that is Mortho. Mortho works with any implant company in the US, and it is you know largely free. So that is some of the magic that I was able to figure out with Mortho is how can you get a machine learning algorithm to learn the morphology of the different implants without stealing their intellectual property, right? Because I have no desire to design implants, okay, or uh to man to work to get into manufacturing. I actually I did that for a little bit. I helped uh design spinal implants, and it was pretty cool. Uh as an engineer, I did this a while back. It was pretty cool, but it also showed me that um it's incredibly competitive, and it's so hard for manufacturing companies to succeed. So I thought that I could have a bigger impact in the software space, and that's kind of when I pivoted. It's so hard for a manufacturing company to start up and succeed. You have to kind of wedge your way into hospitals and in physicians' lives, and in it, and it's incredibly hard.

SPEAKER_02 42:34

The only question I can put uh finish with for my part, anyway, is then where do you see yourself going with Morso? Because it's it's doing such a key aspect of surgical planning and in a specific key area. Do you want to expand on that? Do you want to introduce other potentially other modalities of imaging? Uh, you know, surgeons that are utilizing maybe more CT scans before they make judgments on uh what implants to use, or do you want to look at other aspects of orthopedics, not just hip and joint replacement, but maybe something else?

SPEAKER_03 43:10

Yeah. Um so where is Mortho? I'll I'll answer it kind of piecemeal. Where is Mortho going? Um Mortho, its goal is again just trying to democratize this technology. So if I enter the space or I see a space that I feel like there is a need to push that to all the different implant companies, Mortho can do that pretty easily. It already has its arm for the implant companies, and then if I see that there's something that needs to happen that where the technology needs to be democratized, that's probably where it will go. That is my goal with Mortho. In terms of the other areas, um I get Like people ask me all the time, when is this going to be out for shoulder replacements? When is this going to be out for spine? We need it for this, that, and the other. And what I'll say is Morpho is trying to succeed in the hip and knee space. And the way that I'm building the platform is such that it is very easily scalable. So a lot of these algorithms they work on their problem only. You train an algorithm and it's like, hey, this is an apple, this is not an apple, this is a cat, this is not a cat. And then you apply it to somewhere else and it like works terribly. And then there are all these new techniques that make that an advantage. Uh, where you know it'll work better elsewhere. The way that I'm designing more, though, is I'm designing it with the future in mind, right? Because the same thing, while I am in the hip and knee replacement space, the exact same thing is happening everywhere else, too. And I care about them also. Some of my like really close friends are in that space and they're like begging me, like, hey man, where are we at with this? Um, so I'm certainly planning for that. Um, and I wouldn't say it's as easy as a simple click of the button, but what I'd say is that I believe that it's gonna be way easier for me than than anyone else because of the way that I designed it with that future in mind, where it can very easily expand. The way that I'm building these models are with that approach in mind. So um yeah, that that's kind of where I'm going. But ultimately, like the goal for Mortho is just to democratize the technology, make it accessible for everyone.

SPEAKER_02 46:04

The the little hint there that there's uh a little gleam for the future, something to uh to potentially look forward to. Um I'll uh I'll hand back to you, Charlie, because I think uh um we've we've really got into the nitty-gritty of the science there.

SPEAKER_01 46:23

No, I appreciate it. I mean great conversation. I mean it's really honourable intentions, which is refreshing to hear. Not that we've had loads of greedy guests who only care about money before, but putting putting the benefit of patients and the wider industry first is refreshing and will probably ironically see you scale more than if you had if you had other intentions. Just on the AI side of things, because you still are like a full-time position, position entrepreneur. How difficult is it to continue building as a small company, particularly on the AI side? You mentioned the compute costs briefly earlier. As you as it keeps learning, are those compute costs going to increase to a point where you need to scale, involve outside investments, or indeed like grow the team, or do you think you can keep it lean and keep working full-time? Don't even know your hours, full-time, part-time as a physician while still in the company and keeping it lean.

SPEAKER_03 47:23

Yeah, uh, really good question. So with as you expand on data, then uh certainly the training time goes up, uh, just because you know you're computing more things. But when I was a software engineer, one of the things that I would do is I would try to make the models as small as possible to make sure that the compute cost stays really low. So training time might get a little bit higher, but but I'm very uh familiar with a lot of the techniques to keep the inference cost really low. So inference being meaning like once someone clicks submit uh or whatever that action is, then that'll stay really low. Those costs will stay really low. And that's why I'm able to offer uh Mortho for free to over 90% of users, is because the costs are so low on my end that it's offloaded by those uh select people who are noticing its really high value for their workflows. So, like even even for those people uh who are using it at high numbers, I'm charging$20 a month. Okay, so this is not like breaking the bank. You know, my inspiration for that number is actually ChatGPT. So ChatGPT, if they're able to conquer the world, you know, I'm they're obviously not, but if they're able to do that for$20 a month, certainly I can, right? I I certainly don't feel like I'm better than them. They're they're doing some pretty incredible things. So what how am I supposed to charge more than$20 a month? So that's that's why I keep it at$20 a month. And then I also like ChatGPT has a free version too, which is also super impressive. So Morpho has a free version. Go ahead, use it, try it out. You know, see if you like it. If it doesn't, don't worry about it. It's not for you. It'll be for someone else.

SPEAKER_00 49:26

I also have an expensive version.

SPEAKER_01 49:27

Is it$200 a month?

SPEAKER_03 49:29

Yeah, I don't have an expensive version. I'm not, I I guess I don't uh yeah. Maybe, maybe one day.

SPEAKER_01 49:36

Who who's maybe when you get more advanced, when you've got voice mode and other fancy, fancy tools in there.

SPEAKER_03 49:43

Yeah, yeah, maybe. I mean, the whole goal of Morthho is to push the industry forward and then do it in a way that is sustainable. So unfortunately, all this stuff, the security, as much as I want to offer all of it for free, as well. I would absolutely love that. In fact, the first app that I made was free, completely free. I didn't even have a billing platform. It wouldn't be possible for me to bill for it. Um, I gave it away for free, and people didn't like it. So I had to make these workflows easier, right? So, like uh Mortho has AI templating. So you have you upload an image and it it automatically kind of orients the the implant on the image to kind of plan ahead of surgery. Now it does it's not FDA approved for planning, so it's just kind of like a visual representation. The surgeon still has to do all that themselves. But physicians really love that stuff, but that kind of costs money to do that. Initially, I was trying to do all that stuff for free, like on the device, so that's how small the models are. You know, if you can do it on a cell phone, that's a really tiny model, and it was it was pretty accurate. But then you have to add in all the security stuff, and you have a chatbot, and you have this implant identification, and you have this authorization and authentication of the user, uh, and so on. Well, that starts to begin to go from zero cents to maybe a hundredth of a penny, and now you scale it by a thousand, and now you're you know, at more. So um you offload it a little bit, but I still try to keep the goals in mind. The goal here to democratize the technology, push the industry forward. I'm not trying to retire off of this, you know. I'm trying to keep my mentors in mind. Um, Dario Amade doesn't even know that he that I that I look up to him. Uh, same thing for you know the leaders at ChatGPT and Google. Google is historically very famous for keeping costs really low, so I look to them a lot for that. Um my best friend works in this space a lot too, so I'm bugging him all the time. But I give him free medical advice.

SPEAKER_01 52:10

So there we go. Yeah. Well, you never know. Darrier might be a secret listener of the Smart Health Podcast. Hopefully. That's delusion, that's delusion right there. I think I've got something like that listening. So, do you think this, just lastly on that, do you think this can grow purely by word of mouth, contagion in a sense that it's so value ad, you're in you're a surgeon, you're in the industry, you know people who know people, that it's just going to it will grow naturally from there. And because you're not looking to retire from it, you're not looking to scale, scale, scale desperately, you wait, you're willing to wait, keep adding value, keep training the model, and it will just naturally get into the places. So, what is is there a strategy where you're outreaching to hospitals who might not know about it? Are you attending events? Like, how are you getting the word out of that sort of thing?

SPEAKER_03 53:00

Yeah, yeah. So the marketing area of Mortho, uh a whole new experience for me. Um, because I've always been in the tech field, right? I've always been kind of designing the software, and then you just kind of like it. It's funny, when I was an engineer, I always kind of thought to myself, hey, if I just build something that's really cool, then it's gonna succeed. It turns out marketing is actually really important. Um, and Mortho, building Mortho kind of taught me that. Um, but I don't spend a dollar on marketing uh purposely. Um ChatGPT, when they first came out, their initial presentation, uh, and I don't know this for certain, but this is kind of what I've heard. The website ChatGPT initially was supposed to be a demo, a demo of the technology to investors to say, hey, here's what we think it could do. And then, you know, people heard about it and they thought it was really cool, and then it spread. So the point is that their marketing arm initially was very low. And then they became world famous in a matter of days, weeks, and months. Um, so I kind of look for inspiration in that. You know, if my product is good enough, then I think people will use it. And I know that there is a huge benefit to marketing, but I still try to fall back on uh pressure on myself. Like, hey, if people aren't adopting this without the marketing, then the product's probably not good enough. Let's make it better.

SPEAKER_01 54:36

Yeah, it makes you wonder how many products out there aren't that good, but get by just on huge marketing costs and subconsciously brainwashing people to use them. That's interesting.

SPEAKER_03 54:48

Uh not only um sorry, let me let me just add because this is like kind of like a personal uh thing that I uh focus on a bit, there's there's a whole separate arm, especially in the medical space, where you get these contractual agreements. So these like electronic medical records systems are historically famous for this. If you sign a contract with an electronic health record system, the average contract, the duration of it, is over 10 years long. That means you're not using another electronic health record system, you're not considering another health record system for an additional 10 years. Imagine the drive for inner for innovation when you know that you don't that you're locked in for 10 years. There's far lower pressure. So that's another thing with Mortho that I really try to uh try to think about is you know I'm not locking people in on things. I just want to have a better product. And and my goal here is to have a better product to democratize the technology. So I don't need marketing to tell me that. I need users to find out about it and love it, and if that happens, then I know that I have a better product. But marketing will just uh make me think that I have something better than I do.

SPEAKER_01 56:23

Yeah, you're preaching to the choir there. Uh we have year-long contracts with some of our vendors, and I get I realize I bought the wrong technology three or four months in, and I'm just stuck with this product for the rest of the year. Um LinkedIn are famous for being almost a monopoly, and their service is pretty terrible because of that, and there's no innovation. They're like they're hardly doing anything with the AI tools, which again is probably because you say they're not forced to innovate because they know they've got you uh they've got you over a barrel. Uh it's interesting, it kind of uh tees me up for the question I was going to ask because being a surgeon yourself, being on the front line, you are being sold to. You uh touched upon it earlier. So, where what's that experience like? It sounds like you're getting bombarded. Where do people go wrong when trying to sell you as a surgeon technology?

SPEAKER_03 57:26

Um I think yeah, so I'd say where do people go wrong? I'd say most commonly, people go wrong in trying to sell a technology that they don't understand. Um so that becomes very apparent, uh, maybe you know, maybe more so to me because I'm more familiar with it, but there's a big risk to selling healthcare technology, right? You are uh, you know, as the company, you are accepting healthcare information. And um there are many platforms out there that advertise themselves for healthcare use, but they're not they're not secure, okay, and that is the biggest risk, uh, both on the company's end, but also from the physician's end. As a physician, if you are encouraging a patient to use a platform and it's not secure, that is a little bit on your on you. It's a little bit your fault for uh pushing the patient toward that uh non-secure platform. So there have been like you know a ton of right, a ton of technology. People are coming to you know, me especially and and others all the time about new technology, hey, try this out, what do you think? And um some of them don't even have uh a security um aspect to their platform, they're just using kind of like the API of ChatGPT, but they don't do it in a way that's secure. And it's like, oh my gosh, you're you're accepting healthcare information and you're sending that healthcare information directly to Chat GPT. Uh that's really dangerous. So, in my opinion. Um I'd say that's probably the thing that I don't like about the industry uh so much, is you know, with that in mind, I very much respect and I want and I appreciate the concept of what they're doing. They're doing the same thing as me, right? They're trying to democratize the test the technology. You say, hey, there's a need out there, we're not seeing the needle move fast enough, let's do it. But you always have to keep in mind the the users and the benefit and harm ratio. Okay. So when Mortho was first getting developed, we built the security arm first. It was it was a painful process. It it took so long to build that authorization and authentication platform to the specs of financial technology, right? My dad's best friend does this for banking systems. Let me tell you, they care a lot about their money.

unknown 1:00:16

Okay.

SPEAKER_03 1:00:17

So um we use the same technology uh for security. Um so that data is getting sent in a very, very secure fashion. Um, there's like all these regulations and and uh areas for improvement and recommendations on how to secure this stuff. HIPAA, right? Um there's a separate one called high tech, and they have all these recommendations on on how to move, how to log uh this type of information. Um we're like so far beyond that. So um we're you know really at the front lines for the security stuff. And I think it's really important to move the needle forward, but I really hope that everyone does it with benefits first and risks as minimized as possible. Um I think that you know, when I was speaking at the American Orthopedic Association, um that was one of the things that I uh spoke a lot about, actually. Is uh you know, while I I certainly do believe in Mortho and the vision of Mortho, but I also encouraged everyone who is listening to be the biggest critic possible. Right. We need to be more critical. If someone comes to you with a free lunch, it's probably not so free. Right. Um the way that you can keep something cheap is by cutting corners sometimes. Um you know, Mortho tries to do it by pushing the industry forward, keeping the costs as low as possible, and follow that Google type model, right? Where Google is able to keep the compute so low, and because of that, their search engine is is free, right? But they have advertisements. You can't really do advertisements in the medical space. It makes you feel like you're not as trustworthy. So you have to do a subscription model, and that's what Mortho does.

SPEAKER_01 1:02:37

Yeah, that's a really interesting perspective uh I'm learning as I go with this podcast. And one of the big mistakes that I've heard companies make is not having one of the other healthcare or technology expertise. I think healthcare comes first. There's so many kind of tech gros who think they can shake up healthcare just with their computer science and machine learning masters. Uh maybe they can, but without being in the weeds and truly understanding the workflow, the problems, the problems within the problems, it's very, very, very difficult. And they're relying on advisors more than anything else. Uh so it seems like the perfect storm with your own backgrounds, the fact that you had that machine learning, the software development mindset, the engineer mindset early, then combined with actual experience as a surgeon. And on top of that, and this is the angle that I've just learned, even though I knew it loosely that security is important. You've got your dad, your dad's friends with the finance, the taxes, that level of rigor, engineering rigor, and security protocol being brought into your product. So those three things are ticked off versus other companies. Maybe they don't know the healthcare system, maybe they don't know the technology, and the technology is not so good, or maybe they ignore security. So that's quite that's a big lesson for anyone listening. Kind of need to tick all three to do things properly. So thank you for sharing it. It's really, really useful.

SPEAKER_03 1:04:04

Yeah. Those are the things that kind of uh scare me. When I, you know, when I was doing the American Orthopedic Association talk, um, one of the co-panelists was the senior vice president of Microsoft Healthcare, and he said that he was uh often uh staying up late at night because he was concerned about the fact that the technology is moving so fast that people are going to use it in the wrong way. And it's funny because I would talk to him and I would say, you know, it's funny that I have the same concern. And I often would stay up at night concerned that people are gonna use it in the wrong way. So what I decided to do was to offer it to people so that way they could use it in the right way. And it's like you have a problem in front of you and you just gotta figure out the solution, and then later you can figure out um you know how to minimize the costs and so on. Because there's a ton of different techniques that you can do. Um that's that's one of the beautiful things about AI, actually, is there is naturally there's inflation. Um so the the dollar or the British pound is valued today uh less than it will be valued at tomorrow, and so on and so on. So, how do you offer something for twenty dollars a month when you know that the value of twenty dollars a month is going down in the future? Um there's an interesting thing that happens with artificial intelligence is that when What we're seeing is that the compute costs actually go down over time. So that$20 a month while today I might be losing money. Tomorrow I will be gaining money. We see that all the time, and uh there's several companies that kind of follow that model. That way you can gain uh you know your uh customers and your moat uh in the market space um and then later on knowing that you're gonna eventually make money off of it.

SPEAKER_00 1:06:38

Interesting. I didn't I didn't know that either. Interesting perspective.

SPEAKER_03 1:06:43

Yeah, certainly I mean they advertise these big companies, Meta, Facebook Meta, um, advertise like all these crazy costs, and certainly there are. And they certainly are getting more costly with more data. But if you keep the data the same, or if you get smarter data, then the costs begin to go down. And if you if you uh keep the training time the same with the same amount of data, but you know, ideally better data, then the costs of training it actually go down over time. That's that's the trends that people are seeing with with technology. And that's why there's these so many companies trying to invest in AI, because they know that once they build a world model, a model that can understand everything, and I'm not exactly using that term correctly, but uh, you know, world model on blah blah blah. So uh we can get into that if you want, because I love world models. But um, you know, if you build a model that is successful at everything and the perfect big brain model, then those compute costs will go down over time, and now you have your market capture so you can stay smart. That's how OpenAI and Anthropic and Google and Microsoft are able to succeed in meta. Meta's done a lot in this space. Meta gives away their models for free. How are they able to do that? That's really impressive stuff.

SPEAKER_01 1:08:24

So a world model is a is it an industry model, so you can build one. A world model is learning everything within that industry rather than the world itself, holistically.

SPEAKER_03 1:08:35

Yes. Yeah, so a world model, um, or some people call it like a digital twin, if you've ever heard those.

SPEAKER_01 1:08:42

Um the time these days, aren't we, Samar?

SPEAKER_03 1:08:46

Yeah, it's I mean the difference between the two, at least from what I've seen and my understanding, is that they are largely just a marketing difference. So you have the IT realm, the technology, ChatGPT, Google, they are trying to advertise this concept of a world model. Okay. So you have the world model, and what both of them do, world model and digital twin, is they are able to understand the three-dimensional environment. So they kind of build and simulate and act on a three-dimensional environment, a vision space, two-dimensional environment, and so on. So uh the whole you get a whole other realm of um, you know, problem space and solution space when you get that. You can still work in the one-dimensional space, which is text. And then you can work on the two-dimensional space, which is images, and then you can work in the three-dimensional space like MRI scans and CT scans or the operating world, or uh, you know, robotic technology. That's where world models come in. It's very exciting. Um but uh like I mentioned CT scan and MRI. So the healthcare sector is trying to differentiate themselves from these, you know, unsuccessful robots. And the way that they, at least from what I've seen, the way that they have strategically tried to differentiate themselves is by coming up with a different name for the same thing. So they call it a digital twin. We're creating a twin of your anatomy, a digital twin.

SPEAKER_01 1:10:37

Yeah, and you can ask questions, what would happen if I did this, what would happen if I did that, what are the consequences without causing chaos in reality? That's the idea. Which is a really nice idea, but I imagine it's very complex to build properly to really understand the different permutations that could happen in reality. You have to have a lot of data. I'm guessing. Yeah.

SPEAKER_03 1:10:58

Yeah, for sure. And very expensive data. Yeah.

SPEAKER_01 1:11:02

Yeah, we hear a lot about medical image data shortages. There's been a company on here actually who essentially sell it, act as a middleman.

SPEAKER_03 1:11:09

Uh yeah, it's it's it's a crazy expensive uh industry. Luckily, you know, um as a physician, you're treated totally different when you when you begin to ask for data. So they yeah. So um it's it's been a huge advantage for me as I had this background in machine learning and was able to merge it with with my uh career as a physician. And then my dad having this financial technology background, my best friend who does computer vision also, so I'm able to bounce ideas off of him. Um my dad's best friend who has his financial technology guy. So it's it's been a uh a really uh lucky coincidental uh experience. Um and it's been super exciting because I I feel like I'm making an impact, and that's really the main goal is to push the industry forward.

SPEAKER_01 1:12:15

Can't think of a better note to close the show. Uh been an absolute pleasure, Michael. Uh not sure if you've got anything else to add or some air.

SPEAKER_02 1:12:24

Um otherwise we will uh I I'm just amazed I managed to go through the entire podcast without making one surgeon joke. I think I had a million uh surgeons just filming how to cut jokes going uh in the back pocket, but didn't get to use them once.

SPEAKER_01 1:12:42

Is that a thing, is it? Surgeon jokes, or like dad jokes.

SPEAKER_02 1:12:47

A little bit, a little bit.

SPEAKER_01 1:12:51

Well, probably probably not the best setting to start uh taking the Mickey, as we say here in the UK. Um but yeah, pleasure, Michael. Um, thanks again for coming on. Um maybe we could do this again in a year and see where you're up to.

SPEAKER_03 1:13:05

Yeah, I would love that. Thank you both so much. I mean, this was so nice of you both. I I really very much appreciate you, Samir. You are an absolute gem. I I just have so much enjoyed getting to know you. Uh Charlie, it's been an absolute honor uh getting to know you, and I I am so grateful for both of your time. Thank you very much. Absolute pleasure.

SPEAKER_01 1:13:29

Thank you, Michael.

SPEAKER_02 1:13:31

Thanks.

SPEAKER_01 1:13:33

Thank you for listening to another episode of the Smart Health Podcast. We hope you enjoyed it. Uh, if you've got any questions or would like to appear on the show yourself, please reach out to Charlie Webb or Samar Gill at WalterSmart. See you again.