The Prosthetics and Orthotics Podcast
The Prosthetics and Orthotics Podcast is a deep dive into what 3D printing and Additive Manufacturing mean for prosthetics and orthotics. We’re Brent and Joris both passionate about 3D printing and Additive Manufacturing. We’re on a journey together to explore the digitization of prostheses and orthoses together. Join us! Have a question, suggestion or guest for us? Reach out. Or have a listen to the podcast here. The Prosthetic and Orthotic field is experiencing a revolution where manufacturing is being digitized. 3D scanning, CAD software, machine learning, automation software, apps, the internet, new materials and Additive Manufacturing are all impactful in and of themselves. These developments are now, in concert, collectively reshaping orthotics and prosthetics right now. We want to be on the cutting edge of these developments and understand them as they happen. We’ve decided to do a podcast to learn, understand and explore the revolution in prosthetics and orthotics.
The Prosthetics and Orthotics Podcast
Prosthetic Socket Rectification at Scale with Josh Steer
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Can rapid prototyping truly save the day in critical situations? Discover how a simple 3D-printed component prevented a costly machine breakdown, and learn why this story exemplifies the incredible potential of additive manufacturing. We also welcome Josh Steer from Radiide Devices, who shares his insights on the challenges of adopting CAD technology in the prosthetics and orthotics field. Josh provides valuable perspectives on improving clinician's efficiency without overshadowing their expertise, offering a glimpse into the future of prosthetic design.
Our exploration continues with the intricate world of prosthetic socket design, highlighting the evolution from academic research to practical implementation. We delve into the artistry and skill required for precise socket fitting, and the supportive role of data from systems like TracerCAD and Omega. Through Josh's insights and our discussion, you'll gain a deeper understanding of how artificial intelligence is reshaping the industry, while still emphasizing the irreplaceable human touch necessary for addressing unique patient needs.
Finally, we examine the collaborative journey in creating advanced prosthetic solutions. With a focus on the intersection of technology and human ingenuity, we explore how scanning and digital tools are empowering prosthetists. Josh Steer shares his team's dedication to enhancing patient outcomes through innovative software solutions, highlighting how automation and customization can work hand-in-hand. This episode dives into the promising future of prosthetics, where technological advancements enable clinicians to maintain their pivotal role in patient care.
Special thanks to Advanced 3D for sponsoring this episode of the podcast.
Prosthetics and Orthotics Podcast Highlights
Speaker 1Welcome to Season 10 of the Prosthetics and Orthotics Podcast. This is where we chat with experts in the field, patients who use these devices, physical therapists and the vendors who make it all happen. Our goal To share stories, tips and insights that ultimately help our patients get the best possible outcomes. Tune in and join the conversation. We are thrilled you are here and hope it is the highlight of your day.
Speaker 2Hello everyone, my name is Joris Pules and this is another edition of the Prosthetics and Orthotics Podcast with Brent Wright. How are you doing, brent?
Speaker 1Hey man, I'm doing well. Joris, I've got a crazy additive story and it's not the one that you're thinking of, or the one that I'm thinking of, or the one.
Speaker 2I don't want to talk about that one dude. Talk about the other one.
Speaker 1Yes. So a buddy of mine who is in I'm not going to say what it is because I mean it is working, but I don't want him to get in trouble or whatever but anyway, he's got a, he's got a robot that does something very important and he calls me in a panic. He goes brent, I've got this small part and it's keeping my whole robot from working. Can you scan it, reverse engineer, whatever you do, can you do that? So I said, dude, I you know it takes a little bit of time to get multi-jit fusion parts, whatever ever, out, and I'm actually not at my house right now, but if you drop it off, I'm going to get home in a couple hours and I promise I'll get right to it. He goes oh, that'd be great, cause I just called the company. It's going to be like the end of the week, maybe next week, before they can come and they don't have the part. They don't know when the part's going to be available. And I said, no, yeah, no problem. So I literally get home, I caliper it up. I have my little bamboo A1 mini $199 printer. Had PLA loaded up, because that's all you know. Like he wanted something fast and I just wanted to make sure it fit.
Speaker 1I printed one. It took like 35 minutes and I looked at it, kind of compared some stuff and I'm like, okay, I need to just tweak a little bit and look like everything lined up well. So I printed two because, honestly, there's one pin that had to go through and I thought, man, he's never really done like stuff with 3D printing. But I told him, hey, you're going to have to use a needle or something to get this pin in right. And I gave him two because I frankly I thought he was going to screw up the first one and maybe screw up the second one.
Speaker 1So I actually printed him a third one too after he picked these up. But he texted me at 7.30 this morning, said the line was up and going and it's going to make a big difference for a lot of people, and so I mean it's just crazy, you know, so it's. I mean it's just, it's just crazy, you know, so it's. You know, fusion 360, a set of calipers, a bamboo printer and literally there's a multi hundred thousand dollar machine down because of this little part, and and he's up and going.
Speaker 2And that happens so often and as often in large companies. You see a lot of times when it's very difficult to make the business case for additive, until something like this happens and then all of a sudden it's like okay, well, that paid for itself, and then everybody's kind of happy. You know, Exigent stuff is really what. We're really really amazing with that stuff. So great example, dude, Great example. So who are we joining? Who's joining us today on our podcast?
Speaker 1Well, I'm super excited. We actually interviewed Josh Steer from Radiide Devices at AOPA last year on our was it last year? I don't know. It all blends together. No, it wasn't last year, it was in September.
Speaker 1So we had a really good chat with him and we were like, man, hey, we need to get him back on more of a full length to do a little bit more of a deep dive into what they're up to, and I think we're going to have a really good conversation. It's kind of neat. They've been working with us here at Advanced 3D a little bit talking about, hey, how do we take these concepts and actually scale to production? And one of the things that you have always said and I think this is where we're going to have, you know, a really great discussion is the bottleneck really is the CAD side for adoption for our field, for orthotics and prosthetics. And the reality is we have always said that it's not really in most clinicians' best interest, it's not their best use of time to do the CAD stuff.
Speaker 1Well, josh and his team really have some great insights into some of that using their software, and so we'll hop into some of that too. So I just really excited. They've been on a long journey and the technology definitely has matured a lot. You know it was and Josh don't hurt me for this but when I first saw some of the stuff and I was like man, I think they've got some good stuff going, but I just don't know. But now I get it. So I'm really looking forward to diving into that with Josh as well.
Speaker 2Welcome to the show, Josh.
Speaker 3Thanks Jero, Thanks Brent. It's a real pleasure to be on. As you said, I had a great preview over a few months ago, so really looking forward to kind of diving into the details. And I think, Brent, you know, I think when you first saw what we were working on, I think we took a bit of pride that it was early, because I think that's been a key part of our development from day one is kind of show it early, get feedback and then develop and iterate from there.
Speaker 2That's cool, and so tell us a little bit first about Radii.
Speaker 3What is it exactly? So? Radii Devices we're a startup from the UK. We do software, primarily focused for prosthetic and orthotic fitting. Our kind of real focus and kind of what we've been doing from day one is how we can use data and how we can use evidence within the fitting process to really kind of support the prosthetist or the orthotist in getting the best outcomes for the patient. And I think, brent, what you're just touching on there about that as a bit of a bottleneck, I think we're really looking to say how we can use that data and support that process to really achieve that scale and make the process as straightforward as we can.
Evolution of Prosthetic Socket Design
Speaker 2So, if we're looking at that, I mean I think so. First of all, why did you make this thing? Well, why make a tool like this for well, for OMB, and not for, I don't know, fitting glasses or something like that, Because it's kind of a similar type of thing, right?
Speaker 3Yeah, I think so. The background is that I've kind of been in and around the prosthetics and orthotics industries since about 2015 when I started a PhD. So, as a mechanical engineer by background, always had a real interest in the biomechanical side of things and was offered a PhD project looking actually kind of very unrelated to socket design, but looking at an implant for a bologna empty. The idea is that you'd have a bone capping implant at the cut end of the tibia to try and improve some of the stress distribution. So that was the kind of original project. But we were working with a prosthetist over in in germany when we were kind of doing some of the initial design for this implant and I was in charge of the modeling, so doing finite element analysis, modeling, looking at how this implant would perform, and we're working with the prosthetist and we had a couple of participants coming because we wanted to kind of understand, okay, what's going on inside the socket so we can then apply those forces to the soft tissue. And the prosthetist, who's kind of an amazing prosthetist, has done loads of the sockets for the German Paralympic team. He bought out three different sockets for for the same patient and we kind of scanned these in. We looked at them and they just looked completely different. I I couldn't believe these were all these, first of all, fit to the same person and were designed for the same person, just completely different design philosophies.
Speaker 3And I think from there really kind of that idea of okay, what is happening within socket fitting, this actual process of trying to optimize the load distribution in between the limb and the socket there's so much complexity in here and so kind of my PhD diverted a bit from there into looking at trying to understand a bit more about this relationship between the design and initially looking at lots of pressure and then by the end of the PhD, phd. So this was in kind of 2020. We had, you know, developed some technology and we kind of said, okay, we think this might actually potentially be useful within a, within a clinic, and so that was kind of the. You know, we really did start in prosthetics and orthotics and I think that is kind of where it's been. But yes, as you say, I think there are similarities in many of these other custom devices. But I think the difference that you have from something like OMP compared to glasses is really just the complexity of that fitting process. How much goes into trying to get the best fit for an individual.
Speaker 2And, of course, you used AI and all these really cool things, or how does it actually work?
Speaker 3so I I it's an interesting term and it's one I kind of have a you know bit of a love-hate relationship with sometimes. I think, um, for us what it really came down to was, first of all, um, the data collection. So really fortunate to have some fantastic clinical partners in the uk that had been actually using cad cam within prosthetics for a long time, using systems such as TracerCAD and Omega, and so they had this database of pairs of limbs and sockets so we're able to see, okay, what were the kind of the local changes made to that original shape in order to create the socket fit. And so then from there, what we kind of look to say is okay, we've got all of these different strategies. I think there's lots of interesting nuances within prosthetic fitting which mean that you can actually really kind of constrain the problem. So we know that you know more than one socket design strategy can be successful for an individual.
Speaker 3We don't really have kind of clear biomechanical metrics as to what we're targeting for. So as an engineer, it's not like I can say I want to get 20 kilopascals of pressure on this part of the limb. So the model really starts by going, just trying to understand what what prosthetists do in certain situations so say, okay, for this particular presentation of limb, what sort of strategy is is being taken, and really kind of using that framework of understanding what going into fitting. And this thing comes kind of why we've always been, rather than just kind of doing all of the r&d you know on, in a lab on our own, we've always so much of our development has been getting constant feedback from prosthetists and orthodisks because we knew we needed to create a model that not only kind of really understood the process of trying to fit socket, but then also something that as a prosthetist they can control, they can override, they can say actually, that's completely not the approach that I want to take.
Speaker 2I'm going to take a wholly different one and how do you like create value for your, for your customers? Is it really just time saved? Is that it? Is it better devices? And how do you explain that to them really? Because I think you're going to be really invasive if you said to people actually you can make a really good device with this, implying that I cannot.
Speaker 3Yeah, and I think again, that comes down to the. I think AI is a really interesting topic. In that area, I think there is a lot of the kind of the media framing around. Ai um brings kind of like people have ideas of what that means when you come into it. But I think what we're really trying to do is just, you know, moving from fitting in plaster to then kind of doing everything from in cavern day one. What we're trying to do is to really give that starting point, so that idea of the 80 20. So we're using the data.
Speaker 3We're actually looking towards the model in terms of saying, actually, this is going to tweak to what you do as a prosthetist and what you feel comfortable with and then applying that automatically.
Speaker 3So, yes, it's trying to save some time, it's trying to get those outcomes are kind of the big thing that we really strive for is trying to reduce refits, but really just providing this data and this information to the prosthetist to help them make whatever decision that they want.
Speaker 3It's something that you've got to really take care of. I think there's a very different way that we could have designed the system and just said, okay, press a button, here's the socket design. This is what we think is best. That's the end of it, and I think we would really push back against that. It really is about saying how can you make these tools useful for the, for the prosthetist as a user, so they can access them, so they can have more information, but they've are in complete control and they can override them across the whole process, because you are dealing with the fitting, which is so much the value that the prosthetist brings and ultimately, no matter what you do with the that the prosthetist brings and ultimately, no matter what you do with the model, the prosthetist is there with the patient in the room, and that's something that no amount of data is ever going to be able to replicate.
Speaker 1Yeah, I'd like you to lean into that just a little bit more. Like, what is the role of the prosthetist? Is it important? Is the software going to take care of it? Will the robots take the jobs of prosthetists? I mean, you've been studying this stuff for man. It's almost 10 years now. What are you finding as the relationship of the? You know where does the prosthetist fit in, where does software fit in? And you know maybe, what does future look like?
Challenges in Fitting Prosthetic Sockets
Speaker 1I know there's a lot there, but I think for for our listeners, I think it's important for people to hear yeah for sure.
Speaker 3I think at the end of my phd, having run of all of you know, hundreds of simulations looking at different socket designs for different individuals, looking at the pressure, the main kind of outcome that I had was that I don't really understand how any socket manages to fit um. A colleague of mine in the same lab did some testing where they apply were applying pressures onto why I was one of the participants onto my calf and that's of others and the pressures that um, the residual limb is able to tolerate within a socket and for then it's actually people for it to be comfortable and to do kind of unbelievable feats on it from almost from a biomechanical perspective. You look at the numbers and you think this shouldn't, this shouldn't work, but it really does. So I think, if I look at as a kind of an engineering problem, I think that what you know prosthetists and orthotists are trying to achieve in, you know, putting so much load through, you know different parts of the limb that were never really designed to tolerate that load, it's just an incredible feat of knowledge. There is so much knowledge built into that and I think for us it's really about saying, okay, fitting a socket is really, is that really difficult. I described that as a really difficult technical, scientific problem and you want to almost be throwing everything you possibly can at that. So that's how we've always viewed it. We've always said we want to give prosthetists and orthotists the best possible tools to help them achieve this really difficult outcome, which is taking, you know, all of someone's body weight and putting it through soft tissues in in the lower limb that were never designed to take that load, and so I think it always has.
Speaker 3For us, the way that we see it, is that it's always led by the prosthesis. It's all about creating tools to support them and, ultimately, to make their life easier, allow them to see more patients, allow them to access more information about what strategies have worked before, what have worked for them, meaning they don't have to go through so many clicks in the software. It always comes down to giving the prosthetist more support within within that process and on the model itself. We uh, we've done a clinical study in the uk on this where we actually looked at how the model would perform on its own, because we really wanted to see, you know, how the model would break and where its issues were, and a study went really well. You know really great data from that.
Speaker 3But my favorite outcome from that study was one participant who had a really sore area, um on their limb, and the software had no idea that there was a sore area there, so it loaded pressure on there. It said, okay, right, we'll do that. And so the socket wasn't comfortable in that particular area. And the reason that's my favorite is because it just shows that you always need the prosthetist leading this process and it's always about them being in front of the patient and just giving them having this as a tool to enable them to achieve the outcome that they want for their patient.
Speaker 1Can you also dive into like I mean you mentioned it a little bit the importance of the data, but the data doesn't tell the whole story. But, as you know and I can say this because I'm a process right, I mean we are notoriously bad for getting data right, so we don't necessarily like to try to quantify it. Data right, so we don't necessarily like to try to quantify it. But at the end of the day, it is a compilation of data, whether we like it or not. It could be subjective objective that creates this prosthesis, and so what you guys are doing is really trying to help the prosthetist with some science and data to go along with expertise. Is that a fair summary?
Speaker 3Yeah, I think that's a really good summary. It's just about providing that as another piece of information that goes in, and also that information is embedded within a useful tool as well. So there's two parts of it. First of all, we can collect the data, but one of the challenges in the industry as a whole is that the data is kind of very complex. It's not a nice Excel spreadsheet. You've got pairs of these complicated 3D scans. They're all wildly different shapes across the population, and so what we see our role as is. We say, okay, we're able to look at this data, we're able to extract useful information out of it and kind of simplify that into a form then, such that it can go into the decision-making process and we can use that to ultimately kind of yeah, as I keep saying, support the prosthetist. That's what it's all about.
Optimizing Prosthetic Socket Design
Speaker 1So what is one thing that's been surprising to you in this journey about the data? I mean, I know you mentioned that you know, like the very original one, you said three different, very different sockets, but they all actually fit the same patient. I mean, what are you, what are you finding in your data sets? Is there truly that big of a Delta? Is there a best kind of fit, like rule of thumb type of thing that you're finding, or there's just a lot of room?
Speaker 3So I think one of the clearest signals that we've seen in the data and it's something which has been really I think almost every prosthetist I've spoken to about this will kind of say oh yeah, that's what happens is that you know we've got and brent jump, jump in, if I may speak, because you know I'm not a prosthetist by training, but broadly, if you you know, and so for me, someone that was looking at the literature, coming into this, and you'd look at research that was comparing between different socket designs for someone with the bologna amputation and you'd have these two different philosophies the total surface bearing, of kind of distributing pressure evenly across the limb, versus the patella tendon bearing, where you'd be really using the load tolerant and load sensitive areas, and these are kind of really pitched as a kind of almost two very distinct design philosophies that you can so you can compare between them in a study. And when we had our original data set they were actually all labeled as either patella tendon bearing or total surface bearing. So one of the first things that we tried to do was say, great, can we get the model to automatically identify which one is a patella tendon bearing and which one is a total surface bearing socket. We thought this is going to be really really easy to do because you know, we look at the literature and they should all be very different. This, the model, was terrible. It was, I, unbelievably bad. I couldn't actually believe how badly it performed.
Speaker 3And then we really went and looked into the data and kind of looked at the original data and it was what was really clear.
Speaker 3Was that actually what was happening in the vast majority of the time was this idea of the kind of the hybrid socket where, you know, prosthetists were taking the best bits of both of those philosophies that they learned in school and applying them to get the best outcome for the, for the patient in front of them. And you know, speaking to prosthetists, that's, you know exactly what they say. They say, yeah, we learn these kind of philosophies at school, but then, as we're going to practice, we then really kind of learn what works in different cases and ultimately kind of borrow from different parts in order to get that socket design. And you know, that's something which has been quite anecdotal, but something that's just incredibly clear within the data. And why that's quite exciting is that you don't have to create a model which kind of deals with these two sockets as completely separate entities. You can have it something that more looks across the whole of baloney socket design in order to try and get the best outcome.
Speaker 1Yeah, and just out of curiosity, are always taught that we need to unload specific areas and load specific areas in school. Are you finding that in your model itself, or are you finding that actually almost everything can take a little bit of load, or what? What are you finding, I think?
Optimizing Prosthetic Socket Design Factors
Speaker 3we kind of find that both can be true. I think you do see that those certain areas, you know, if you create a socket where certain areas can take a bit of load and you then get, you can get one particular outcome and but then if you want to distribute the pressure kind of over everything, you get another outcome. And actually the way I see it is that you're potentially kind of optimizing that for different things and in different use cases one might be more appropriate than the other. So I think what the model and the data has really showed is that all of these approaches are valid. It's just about using the data to say, okay, kind of a wisdom of the crowd approach.
Speaker 3So what do most prosthetists do in this situation? What's the strategy that they most often use? And then, once you kind of pick that strategy, how can you really optimize the, the data within that? How can you really look to try and hold the outcome that you that you get?
Speaker 3So I think it almost goes right back to that first idea of those three sockets that there are different approaches to making what is a good socket, because that's not a, you know, if we look at other areas of medicine, if you're trying to diagnose something, you've got very much got this idea of, like you know, someone has the disease or they haven't, whereas in prosthetics you've got far more subjective measures that you're constantly dealing with.
Speaker 3You're doing it based upon kind of outcomes in terms of how the patient perceives the limb, and so therefore you don't just have these really kind of clear yes or no answers. There are kind of multiple ways to get the same thing. I think that the way that I always like to look at the problem is a bit more kind of like maybe Google maps or something, where you're trying to get to a destination which is a good socket, and there are multiple possible ways that you can get to that destination. You can take one road or you can take the other, but if you take one road, you then want to try and get the best route going in in that direction. So that's kind of the frame of reference that we often use for thinking about socket design, rather than something purely from a kind of a medical environment so I mean, but making this model is is fairly intensive.
Speaker 1It's not just one set of data, right, because it sounds like okay, so you've got the input data, so, whether it's scan data or a cast data, then you have the data of the rectification, and then you have the data of the rectification and then you have the data. Did it work for the patient? And I'm sure I'm missing a whole lot of other things. But I mean, that's really the lift that you guys are trying to lift and that's a lot of work to get all that together.
Speaker 3Yeah, the initial data in particular, and you know, we were very fortunate to have a incredible partner in the UK that had been curating this data set for a long time. Without that, we wouldn't really have been able to build this in anywhere close to the time that it's actually taken us, because this, you know, although we've been in the space since 2015, this effort really only started in kind of end of 2021, beginning of 22, to really train up the model. Yeah, without that initial data set, it would have taken a lot longer to kind of really get to the point of where we are. But actually, now we've got a framework, now we've built that initial model, we understand the architecture. It becomes a lot more straightforward in terms of expanding the model, training it, testing it.
Speaker 3A big thing that we really want to do is understand kind of different approaches, regional approaches as well. What are these different strategies? How are they used for? Can we actually really tweaking the model?
Speaker 3So it's then providing very much personalized suggestions for the individual prosthetist, because if you've got one prosthetist who really goes towards one type of design and they're absolutely brilliant at that type of design, they understand all of the nuances inside and out and they always get a good outcome with it. Even though the statistical data looking across all prosthetists might say that actually a different approach is better for that patient, you're probably not going to get a better outcome for that prosthetist because you're trying to make them do a socket design that they're not comfortable with. So I think that's why yeah, really understanding you know what is what the divergence is in between different prosthetists and different regions and different strategies, different where they've been to school, all of these aspects, that's that kind of next level of what we're now starting to see within the data, and that's the kind of data you can only collect with scale.
Speaker 1Really, so also I mean you kind of mentioned it Did you guys look at some of the historical documents you know, going back in history just a little bit to read up and glean some of that just a little?
Advancements in Prosthetic Socket Design
Speaker 3bit to read up and glean some of that. Yeah, and I think again, that was kind of particularly during my PhD. That was a huge part of actually getting to where we've got to now and it really informed the selection of the model and how we kind of approached it. I think there was definitely a way where we could have said, okay, there is some amazing kind of AI models out there, amazing technologies and what these models are very good at, a way where we could have said, okay, there is some amazing kind of ai models out there, amazing technologies and what they, these models, are very good at. I would describe as saying you can give them quite messy input data, but if you give them enough, they'll kind of like just brute force through it and they'll find the signal and they'll give you an outcome but really when?
Speaker 3we're looking back into the literature and I think particularly the papers that were coming around, the kind of the late 80s, the early 90s, from people such as David Boone, david Reynolds, ed LaMere, joan Sanders, lots of work happening in between, seattle University of Washington University, college, london. There was these really kind of strong themes emerging of that when they were talking about using CAD because this is what they were developing. This is where they were developing the first CAD systems that would become things such as ShapeMaker. They were not just saying we're going to just take a process which is currently done in Plaster and make it digital and that's the end of it.
Speaker 3It was always about that data. It was always about building an evidence set and really starting to use that to understand what is going to give a good outcome. That was always what the end goal was and I think you know, looking at that, we then kind of said, oh, actually maybe CAD hasn't quite lived up to that those kind of original goals of this work. But some of the kind of the ideas and the foresight within those papers was really instructive in terms of us like thinking about this as a problem and looking at what are the strategies that we can use to break down this data what hasn't worked previously? So, yeah, I think we were that. There's, yeah, just this really rich vein of about 10 years worth of papers where you know the ideas behind what they were trying to do were just way beyond what could be achieved with scanning technology and data processing at the time.
Speaker 2And is there a way for you to also get like stuff, like data, like how happy patients are, how they feel with it, because that's really important, I think, in the relationship usually? Is that a way to get in your model how you know how robust it feels or how comfortable it is, that kind of thing? Or do you want that even, or is that something for later on?
Speaker 3It's definitely something that we've won, and I think again, when we designed the model, one of the key things was that we wanted to be a model that we could build on top of, so we could plug in new pieces of data and we wouldn't have to completely overhaul how the model worked in the architecture. We would just be able to add that data in and that would then kind of, I guess, change the weightings within within the model, and so that's definitely that next step. I think the way that we see it is we're kind of almost starting from a space of there's no real structured way of collecting any of this data around socket fitting at the moment. There are kind of pools of data here, here and there, but there's no consistent way of looking at it and analyzing it. So I think the model that we've got at the moment is kind of our foundation model.
Speaker 3It's that it's that first step, but there's so much more that you can look into in terms of what else can be integrated, and I think you know patient outcomes are a huge part of that.
Speaker 3Not just patient outcomes there and then during the kind of the diagnostic socket fitting, but actually how's the patient doing after three months how they're doing. After six months are we able to see that some sockets maybe are, you know, influencing volume change, for example, or long-term outcomes. They're really interesting looking at as well, kind of data for someone that's having their first socket so mainly only had an amputation, say three months ago, looking at how their limb is kind of rapidly changing over time and saying what sorts of sockets are for that. I you know there's enormous amounts of potential data that we can start to include. But it all starts from that kind of base of just understanding, creating a language to really understand socket design and say here are the output parameters that ultimately go into creating a socket, and then the more data that you put in, the more kind of categories of data, that's what's going to give more personalized models, kind of down the line.
Speaker 2And we often hear like kind of talk about, like you know, we kind of contrast the full-on CAD. Should you learn full-on CAD, or should I have like a push button wizard which will kind of own all the data and I won't learn anything from it? You know, is there a third way as far as you're concerned in this kind of element, and then doing this in a very, very different way than those two extremes?
Speaker 3Yeah, I think so and I think this is, but it needs thought and it kind of almost goes into the AI question. It's not just trying to throw a really powerful brute force at it. It's really bringing that care and that design into the architecture and making. We spent a lot of time working on the model architecture to really say how is a prosthetist going to be able to use this and actually drive this? And when it makes you know when they make different decisions, how is the model going to be able to update based upon that, based upon that new information? I think one of the kind of one of our outcomes is the kind of it would be really nice if this was an outcome was that actually, by presenting socket design in this way, by kind of going through that process of fitting, actually it will open up more conversations about the causality and why different socket designs being picked, because suddenly there's a kind of a common language that everyone can use and there's a framework. So, yeah, I think to me that kind of that balance of where do you want to bring in some automations versus where do you want to kind of allow that full customization, that balance and working through that process again, is a really big part of what we've been working on. I think optimization, that balance and working through that process again is a really big part of what we've been working on.
Speaker 3I think, and even I think the term cad within a prosthetics and orthotics space is probably maybe one that needs kind of updating or splitting within additive manufacturing, because you know, traditionally when cad cam was talked about within prosthetics it would be, the output would then be a cnc carver and then you know the sockets then being laid up over over the top of that mold where, so you know, kind of the output would always be very similar. There was a standard manufacturing process thought for that. But additive has really kind of opened up this whole other design window in terms of what I'd almost call the product design aspect. You can get a socket from one place that looks completely different to the other ones based upon the design decisions that have been made within, how you kind of shape certain things, and that then goes into making sure that the socket is strong enough, making sure that it's lightweight enough and performance is really much more of that kind of design optimization, and I think that's where, from a CAD perspective, you're looking at those really powerful kind of high-end CAD programs that you need in order to create that engineered part.
Speaker 3And then there's the kind of what has traditionally been known as CAD CAM within the space, which is doing rectification within a digital environment based upon a 3D scan, and those two elements you're both manipulating the shape in there, but actually there's a very different skill set that goes into them. They've got very different kind of outcomes, potentially different users, unless you're someone like Brent Wright that's kind of mastered both aspects across the the whole process. But you know the work that goes into that and the fact that different softwares are awfully needed means there is this kind of like interesting Passover point in that whole kind of digital process, and I think that's only going to get on greater as additive manufacturing kind of continues to scale within the industry. The software, what can be done gets more advanced. All of these different features such as lattices, variable materials that you're seeing, which can really be used to drive those improvements in patient outcomes from a socket design side of things.
Speaker 2And if we're looking at this, I mean, how do you see about like, for example, scan data right and the input of that, you know? Do you expect that to get better all the time?
Speaker 3Do you want to go to iphones? Do you want to uh, or do you see that as somebody else's challenge? Yeah, I think for us, scan data has been where we spend most of our time. That's a lot of our work in terms of algorithm development definitely goes into taking those 3d scans and actually trying to get the best outcomes with them and really normalizing them so that we're not reliant on one particular scanner in order to run it into a model. We can accept scans from different places.
Speaker 3I mean, it depends on what device specifically we're talking about within the prosthetics and orthotics space, but I think we're definitely getting to a kind of key inflection point where I think in this space you can get too good a scan and sometimes actually, if you get a really high density scan of someone's limb and you know the limb, it's soft tissue, it changes shape, it swells up and down during the day, kind of, there isn't really a kind, it's not a machined part. There isn't an absolute truth of what the limb size and shape is. It is variable. So a scanner that's going to be up in the high tens of megabytes as an STL file and has hundreds of thousands of points and 0.1 of a millimeter accuracy. Potentially you're adding in a lot of computational power but you're not necessarily hugely improving the outcomes.
Speaker 3Now I think a few years ago some of the phone scanners, they definitely weren't at that level of accuracy that you'd be able to come from one of these devices. I'm seeing it, you're really seeing it now starting to come to a point where actually it's getting within that window. I'm not the person to be able to say exactly what the window is. I think that's something which each prosthetist kind of makes that judgment themselves in terms of what they get good outcomes but that kind of that increasing focus on low cost scanners is definitely something that's massively kind of benefiting the industry and what we're trying to do in terms of increasing that accessibility. And I think many of these scanners are kind of very close to being that or, frankly, are already there. I know huge numbers that using scanners like home and having great outcomes within the space.
Innovations in Prosthetic Socket Design
Speaker 1So yeah, that was my, that was going to be my question, yours you stole my thunder there a little bit of the scanning and the inputs of scanning, but you know, I'll forgive you this time, I guess. But you know, I think it's interesting that you say that you can't have too much data, and I mean we definitely see that not only on the intake side of things, especially like when you're trying to manipulate large areas with a massive, you know, massive mesh. It can make it very much. So I can't even imagine when it when you actually try to bring it into some sort of algorithm like what you guys have. I mean the other part is you can have too much data on the other side, when it comes to the print, the output. You know we you're giving so much data that the printer can't even print what you're asking it to and I. So I think that's all stuff that we've got to. You know, get some data around, but also be you know, as the industry matures, we'll understand those limitations much more as it comes around.
Speaker 1So yeah, I'm glad you touched on the scanning side of things and shout out to Comb, you know if John or Alex is listening, because it sounds like they're doing a really nice job. We're actually doing some stuff with the CombScan, their iPhone 12 mini, which was the smallest handheld phone for scanning. I think that's been around and we've just been having great results with that, even though it's what? Four generations old, but it actually has the best hardware, which is kind of which is kind of goofy, but I think that that's neat. So so, after scanning, though, what is the most important part that you feel is kind of? Next, when the clinician starts interacting with your model.
Speaker 3Yeah, I think it does really all start with the scan. Um actually, on that too much data to find a scan point we actually run when someone imports a scan into a big part of our data pipeline, is that we have a remeshing tool that brings it into a mesh density, a mesh structure that we know works and we're and we're happy with, and I can tell you for many of the high-end scanners it's it's not up sampling, it's down sampling. After that then, and where we've spent an enormous amount of so much of our time, is then really the, that process of taking the, the rectification, the modification, that fitting and bringing it into a digital environment such that people can access the benefits of data. Because I think for me that is when you kind of look at what are the kind of classic motivations to go to digital in any industry. One of the biggest ones is always around where you've got this data now and you can learn from it. And I just don't think that's just not really been there until now. It's just been how can we replicate our plaster process in software, trying to do exactly the same thing rather than saying, ok, this is different. You don't have that idea of being able to kind of hands-on with the plaster and shape it. You've lost that, you're always going to have lost that by moving to digital. But what can we gain instead? And are those kind of benefits actually able to kind of, you know, make up for the kind of the lack of the hands-on on the plaster?
Speaker 3So really then, going into the rectification, the fitting, and saying how can we make sure that this process is delivered in a way where the prosthetist is in control, is able to get the outcome that they want, but they're not having to go around and kind of making these fine sculpts and trying to use their mouse to get exactly what they want? It's about making it so, once they've got their intention of saying, okay, I want to do this sort of modification, to come up with this sort of socket, the actual kind of the process of moving those points is then done for them. So they're making the clinical decisions, but we're making it very easy within the software for them to replicate that. So then, understanding how we go through that fitting and get to the outcome, I think is is the next bit. And from from there, you, you then get into the kind of rapidly changing world of fabrication which I think is for a long time, was kind of very standard. This is how you make a test socket. This is how then you go into a common fiber definitive.
Speaker 3But I you know, I really think we're just scratching the surface in terms of what these sockets are going to look like in the next five or ten years as people really start to push the bounds and exploit what can be done with additive. But creating those sockets at scale is probably going to need a whole different approach to what's being, what's being done before in terms of automation, because you're then trying to automate something into a to go okay, how can I take this socket bit shape and make it into a full product as a socket with its own kind of design, language and life, depending on where it's come from? So I think that, to me, is the thing where we're going to see huge amounts of change and progress. You know something like the Quorum socket, for example, really just pushing the bounds with what can be done with socket technology.
Speaker 1Yeah, with what are you finding as far as people kind of leaning your way? Do you feel like it's some of the younger clinicians? Do you feel like actually there's some clinicians that have some gray hair kind of in between, or does it run the gamut? I'm always curious to know, like, who is really looking at this.
Speaker 3You know we really and you asked a question about this earlier about the kind of you know how it's perceived and I think we've really been kind of very positively surprised. I think we were maybe expecting a case where we'd get a lot of pushback, but I think because we're trying to do so much engagement on understanding ultimately what makes a good socket fit and how can we make that more consistent and how can we share and distribute that knowledge, I think we've really had a lot of engagement from a very wide variety of individuals and you know all sorts of different experiences, time in the industry. You know I think it's almost easy to forget that. You know there'll be prosthetists out there that have been doing cad cam and you know we're working with some of them for 20, 25 years and they've trained a huge number of people on these 25 years and they've trained a huge number of people on these technologies. Actually it is a lot more kind of accepted than maybe we expected coming into it and often it's just around kind of not just how do you create the right software tool for that, but how do you actually put that into a really valuable end-to-end process to get those consistently good outcomes?
Speaker 3But yeah, it's not just the kind of those that are just graduating and they're the only ones that are talking to us. It really has been a very wide range of different experiences, different backgrounds, that have engaged with the technology, which, for us, has been fantastic, because one of the things that we think about is that you've got these incredibly skilled prosthetists that have been in the industry for several decades. All of that knowledge gets lost and actually, if there's a way that can be retained and understanding what they've done that makes really good sockets and can that be shared and taught, that for us, is a really exciting potential outcome.
Speaker 2Yeah, and how about you know we're looking at different technologies. You know the main ones we always end up talking about, like MJF, centering and then FDM. Right, you know for you guys, are you looking on the output side much more to one of these technologies and doing more calculations for it, making it for a particular technology, or do you see that changing in the future, that you're going to have to work with all of them?
Speaker 3Yeah, I think that we see ourselves as a real enabler. I think you know we aren't as a company, we're not 3D printing experts. You know we only do software. We don't have 3D printers, we don't send out sockets. That's kind of a decision that we made. We're really going to focus on our bits of the process and do that really well, because you know, one company trying to do everything end-to-end across digital workflow is an incredible number of different skill sets. That's needed different. It's so complicated.
Speaker 3You know there's so much that goes into making a socket through a digital process and but I think where we're really seeing is that because of all the work that goes into our the data cleaning of scans at the start, the consistent structure that we have to use because we have to get it into a form where we can run it through our data model, it actually means that when it comes out of our system it's in a very consistent structure. You know exactly the file is going to be always pretty much the same size. It's always going to be aligned in a particular configuration, exactly where everything's going to be, and so that consistency then makes the next step of the process in terms of that automation for actually kind of getting ready for 3D printing. So, whether it's a basic level thickening, adding in a distal end connector but on the advanced side, adding in all of these different lattice structures, variables, different materials they are all based upon having a really good input data. It's that kind of garbage in, garbage out, and what we're trying to do is we're doing a lot of cleaning of the original scan so that when it goes into that process and it can really start to be automated for these different technologies.
Speaker 3Actually, they know that, okay, if we've got a file that's coming from a radii system, we know that it's going to look like this and that's going to enable us to automate and that's ultimately what's going to be able to. Then the next stage of the process takes up and and scale up. But I think, yeah, we, we very much see it as all of these, the kind of the selection of the 3d printing technology and the ultimate design of the device that comes from that 3d printing technology. Because depending on, obviously, on that technology, you're going to design your device in a different way and it's going to have a different purpose. We see it as it's our job to make sure that when it goes into that process, it's set up, it's optimized and that we can easily integrate with those. So from the prosthetist's point of view it becomes a really seamless experience.
Speaker 1So we talk a lot about intellectual property, who's who's is what and that sort of thing, and it sounds like I mean where you're going is more the intellectual property is is what belongs to or what makes a good. Fitting right With the clinician, and your tool is the helper to help them make good decisions. What do you say to the like? What is intellectual property? Maybe what isn't, and then what is for the greater good as well.
Speaker 3Yeah, it's a really interesting question as you kind of look across that whole span.
Speaker 3I think the way that we've always seen it is that the input data so the scans, the socket designs, those raw ones themselves they're not ours to own a big part of what you know.
Speaker 3If you look inside our data model, it's not just got loads of kind of limbs and sockets floating around. What we do is we take those, we analyze those scans and we extract the measurements and the kind of bits of data that we need from them and that's what's then used to train the model. But then the original files get kind of released, released back, so that for us that data model is what we're building and ultimately we want that to become a collaborative thing so that everyone's benefiting and learning from what different kind of places are doing. But those raw limb scans, those raw socket designs, the the final design that gets made, we see that very much as the kind of that is the data of the prosthetists. Really, that's what's going through, that's their patient data. So we try and take a very light touch approach in terms of learning from that data but then having a model which doesn't just rely on, you know, storing all of these kind of these raw scans.
Speaker 2All right, that's really interesting. So, josh, thank you so much for being on the show today thanks very much for having me.
Speaker 3It was a real pleasure to be able to get into the details of this stuff and uh, I bet you really really like that, didn't you rent?
Speaker 1uh, we got into the weeds there in a little bit. You were quiet there for a little bit there, oh my god, it was so over my head.
Speaker 2I was like, oh, I'm gonna just sit in a corner here and just hope nobody asked me anything.
Speaker 1No, it was great and, josh, thank you for sharing your journey and really being an advocate for our industry profession as well. I find it refreshing to see like hey, this is really an interdisciplinary type of thing. It's not an engineer that's looking to replace prosthetists and then prosthetists how do we work together to get these good models and really create great outcomes for our patients? I think that's what it's all about. I just really appreciate that's the direction that Josh and his team are looking at as well.
Speaker 3Yeah, for sure, brent. I think from an engineering perspective I think prosthetics and orthotics is the best way to describe it is as humbling, particularly when you get into the socket design. It really kind of challenges your assumptions of how you look at engineering problems and it is that kind of real interdisciplinary work that is what kind of continually attracts people to look at it and say, ok, how can we try and solve these incredibly difficult problems by working together and coming up with these different approaches to get those great outcomes.
Speaker 2Awesome stuff. Hey, thank you, Josh, for being on the show today, Thank you Brent as well, and thank you for listening to another episode of the Prosthetics and Orthotics Podcast.
Speaker 1Have a great day, and that's it for this episode of the Prosthetics and Orthotics Podcast. A special thanks to Josh Steer for sharing about Radii. If you want to know more about his company, check him out online. Follow him on LinkedIn. If you enjoyed this episode, please share it with your friends and leave us a review. It would mean the world to us and until next time, we'll see you on the next episode.
