0:00 Just want to kind of start off, my name’s Nate Shirley, I work a part of a team with an HP that, that helps with application as a consultancy. We work with our customers to help automate their applications for geometry you might be asking, why are HP and radi on the stage at the same time, you know, to very different entities. But we found that, when it comes to Applications, we have our domain of expertise, we need to partner with, you know, companies that really know the application space and are pushing the boundaries there. So in this context, Orthotics and Prosthetics is, is a very viable application space, but the scale that a company like HP needs to be able to make it viable for us, we have to, we have to look into the automation space, like Brent was alluding to, that you need both the, the clinical side and the geometry side to be robust and automatable. And so, just as a segue, radii is doing an amazing job with that, and we’ve been working with them on a project with Veterans Affairs for the past year, on how to help them digitize their Prosthetics process. So, just as a intro for Josh, oh, that’s what I just said. Thanks.

1:23 N hi Von, I’m Josh, I’m the founder and CEO of radio devices, so we’re fairly early stage startup using software and data to improve the fitting of all sorts of custom devices that kind of interface with the Skins, it could be medical devices, but one of the kind of key areas we’ve been focused on is in Prosthetics. So I’m going to be talking about the socket fitting stage, and you know, Brent gave just the most incredible instruction, and the work that he’s doing is, is just mind-blowing, and he’s, you know, really personally driving the industry forward. So he’s given a lot of this, this background, but I’m really going to talk about why, this, this fitting stage, briefly referenced it in the kind of the clinical aspect, but this is really where the Protist comes in.

2:07 So from the limb, here, the bit at the top, the socket, the bit that BR was talking about, if that doesn’t fit right, that’s where the kind of the pain, injury, and immobility will come from, much of the time. These can take multiple refittings in order to get right, and in Prosthetics as a whole we see some fairly High rejection rates of, of these devices. So kind of again, Brent spoken about this, but the kind of traditional aspect of how you go through and create a prosthetic socket, starting with this clinical scanning or shape acquisition through plaster cast, and then the, the fitting aspect, we then go into fabrication of this, what’s called a check socket, or a diagnostic test socket. So this is the, essentially, what’s then used to see, okay, how well is this fit going to be, before we move on to the final device, the definitive, that the, the patient’s going to take home and, and use.

3:00 So my background is from mechanical engineer, I’ve kind of moved across, and my PhD was looking at fitting of these devices, and I think the thing that actually really grabbed me in this whole Space wasn’t just a kind of what can we do with additive manufacturing and device design, but really actually the fitting of these devices is probably one of the most interesting geometry problems I’ve ever come across, because you might just think, okay, we’ve got the shape of the limb, we can take a scan, why, why is that not enough. And so, room full of many designers and Engineers, I’m going to show you with a finite element model to demonstrate why.

3:35 So if the limb and the socket were basically the same shape, so if we just took a direct scan, the, the fit would be unwearable, you wouldn’t have any contact or sheare around the main bulk of the limb, and you just get all of this loading at the end of the limb, and that’s where the scar has been, that’s where there’s going to be lots of pain. So what people like Brent do, the prosthetists, they do this just incred able process, where they take the anatomy of the limb and they apply a press fit around the main bulk of the limb, and this kind of provides the suspension, this provides the security of the, of the limb within the socket, to enable someone to ambulate. This then removes the pressure on the distal end of the limb, but the question is, where do you apply the pressure, and again this is then where we start to get into really interesting areas of, of really what’s compensational design.

4:25 So what we have, we’re essentially applying a deoration field onto that initial, initial shape of the limb. So kind of from the literature you can see that there are these areas of the limb, so this is someone that’s had a baloney amputation. So if you kind of feel around your knee, you’ve got all of these kind of bone structures or different soft tissues, and on some of those areas you can put pressure, and you can put that support to provide the suspension, but on other areas you want to remove pressure, because that’s where it’s going to cause pain.

4:54 And so what kind of through the literature and through kind of Decades of prosthetic experience have come around, is that we’ve kind of got these two designs of prosthetic sockets, one which is called a PTV design, which aims to put pressure on specific areas and then offload it on others, and the other one, which is called a total surface bearing design, so that aims to, you know, minimize pressure gradients across the surface, in order to get a good fit.

5:22 So in a Prosthetic Clinic, the way that I look at it, from kind of an engineering perspective, is that there’s this huge, complicated, multiobjective optimization problem that’s happening every day. So when Brent showing, kind of, you know, that plaster, that, you know, what looks like quite a traditional process, the actual kind of optimization that’s happening in there is incredibly Advanced, because all they’re working with is just being able to change the shape of that Surface, by providing loading in some areas, offloading others, to try and get this, this good fit.

5:53 But we do see that there’s, you know, a large number of challenges in Prosthetics, so these devices are kind of near the limit, or at the limit, of innovation with what could be done with carbon fiber and traditional fabrication. Adequate fit is, is a constant challenge, High rates of Abandonment, and a really heavy Reliance on a incredibly skilled but relatively small Workforce, for the, the ultimately the increasing demands that we’re seeing for these devices, particularly in lower middle- inome countries, or, or conflict zones, and there’s a real kind of lack of objective data on best practi, where you are reliant on these, you know, highly skilled individuals.

6:32 So that’s kind of then where we came in, from a, from an engineering perspective, and said, okay, can we look at this problem as though it was like that optimization problem, can we start to say, can we really understand what’s happening within prosthetic socket fitting, and then use data and machine learning to be able to support that process.

6:51 So we started with a relatively small training data set, within, you know, data science as a whole, 162 data set, you know, in prothetics this is pretty big, bringing together these kind of data sets together is, is a real challenge. But at the, on the top row we have, these are all of the original scans of the limbs, on the second row we then have those sockets that have been made, the third row, what you can then see is the process that went into the fitting, so the red areas are where it’s being pushed in, blue areas are where we’re offloading pressure. And then what we were able to do is we were able to create a parametric model of socket design, so we can actually start to look at, for each of these areas, what’s happening, what are the decisions that the, the clinicians are making, in order to get the, the best outcome for, for patients.

7:39 And what I’ll then kind of move on to talking about, is that we then want to take this and wrap this within the clinical process. So what we’ve done, we’ve kind of gone from starting at a very clinical standpoint, then kind of bringing in into an engineering context, but how do we kind of take that back into the clinical context, in order to make useful, valuable tools for, for clinicians.

8:02 But what we can see is that, even just by looking at this data, we can start to get loads of insights into how design worked that weren’t there before. So as I kind of mentioned, there were these, this idea, if you look in, in kind of, you speak to in Academia, what people get taught in schools, is that got these two very distinct socket designs, and it’s, it’s either kind of a comparison of one or the other, like you’re taking two different Products off the shelf. But what you actually see is that the data suggest that’s, it’s a real hybrid design, so when it comes to the clinic, prosthetists are taking inspiration from both aspects, they’re using designed variables from kind of one principle and matching them to another principle, to get the, to get that output, so really kind of part of that, that optimization.

8:47 So what we’ve then been able to do is we did a, a clinical feasibility study, looking at the model that we’ve built and comparing it to prosthetists. So we, we have, would have the same patient, they were try on one prosthetic socket that was fitted using our machine learning model, versus another one that had then been fitted by the, the prosthetist. So we PR this as a non-inferiority study, we want to say, okay, we, you know, we always want to put the proces at the heart of the decision- making process, but if it was used in what we would describe a state of misuse, where someone just said, yep, okay, I’ll accept the output from the model and I won’t think about it, are we still going to give a safe results.

9:28 And so, when we kind of did, did these side-by-side comparisons, we saw that the model was actually able to not just kind of get the non-inferiority condition that we were looking for, but actually, in some cases, really improve on the, on the kind of the socket that have been made by the, the prostatitis. So I won’t go into loads of detail on this, but I, we can kind of split it up between which of the sockets were more comfortable and which were less comfortable, and we found that the, the model actually, you know, it’s, didn’t get as many kind of tens and nines, so this, it’s measured in a socket Comfort score out to 10, as the prosthetist. But those kind of those sockets which weren’t fitting so well, from the Protist, it brought them consistently up to about an eight out of 10.

10:11 And I think we’ve already spoken a bit today about the 8020 principle, and I think you really see this here, the idea that we can use this model to generate kind of 80% of the fit, and all allow the prosthetist to come in and really optimize that final 20% of, of getting the, the device to fit well.

10:30 And so this is then kind of our software, so this is reform, which is used, so this is a software for clinicians. The idea is that this gets used within clinical practice, the, the prst can come in, they can change all of these different variables about socket design, in order to get the outcome that they’re looking for, because the interesting thing about this, is a kind of optimization problem, is that you’re not just dealing with the scan that came in, yeah, you can take measurements and then decide on a socket design, but all of the different design variables into play with each other as well. So if I offload pressure in one place, I’m probably going to have to put it on in somewhere else, in order to get that balance, again, multiobjective optimization, what we’re kind of dealing with in all sorts of different domains here, but this is happening every day in a clinic.

11:17 So just a few kind of quotes from our patients, so on the right hand side, these are these test sockets, so that’s what we used in the study. Nate, in a second, is about to start talking about all of the really cool stuff that we can then do with additive manufacturing on top of this. But really good feedback from patients about how comfortable the sockets were, really, you know, engagement from the clinicians as well, kind of saying that, okay, this can help get kind of 90, 80% of the way there, really helping them to dial in for their fitting, and so, you know, really kind of positive feedback there.

11:51 So that is the kind of introduction into the importance of, of fitting within, within it, I’m talking about, about Prosthetics specifically, but any of these kind of devices where you’re relying on this kind of device human interface, and it’s applying pressure, getting that fit and getting the geometry right, is just absolutely critical in terms of the outcome. You know, you, you could go put all of the technology you wanted into additive manufacturing, if that fit and that interface isn’t right at the beginning, you’re going to get a failed device.

12:22 But we see that digital methods enable us to create models which aim to improve outcomes and support the kind of challenge resources in the Pio Workforce. So I think, and this where I’m going to take our for tonight, but up to date, these sockets that have been produced through either kind of carbon fiber lamination have all looked broadly similar, if you go to kind of any technician within Prosthetics you’ll get a similar outcome. But additive manufacturer really opens up avenues for product identity, Innovation, and improve outcomes.

12:57 But in order for it to scale, there needs to be this kind of consistency in the fitting process, there needs to be consistency in the processing as well, if clinicians using all sorts of different devices, and it’s not cleaned during that software process, that’s going to be a really challenging problem to scale. But the kind of the most important aspect, and it’s something that we constantly grapple with, is ensure that these methods can be embedded clinically, because it’s very easy for us to kind of go into an engineering context and then, you know, talk about all of these things, but ultimately these about tools, they get are being used in the clinics every day. So that kind of balance of how do we take this data and these new technologies and make them accessible and appropriate for, for clinical practice.

13:42 I’m not going to hand over to Nate, to then talk about the next stage of the process, which is, once you’ve got that fit, what are you then able to do with these, with these devices.

13:57 Yeah, so, so that fit, I mean, just to be able to have that to show to the clinicians, seeing what they’re doing represented in, in that way, is, is already quite impactful. Without this Foundation, you can’t build the house on top of it, it has any function, you’ll just, you’re just kind of in a, a state of, you know, useless Innovation, if you don’t have that interface worked out.

14:20 So from there, we see that people struggle, companies struggle a lot, and that’s in terms of how to automate the geometry. What Brent just showed is, you know, probably a lot of work to come up with a system that can automate a 3D printable file, you can only, you can probably imagine that from there the complexity that we could innovate into these products is, you know, is immense. And, you, we want to have clinicians doing clinical work, and you want to have Engineers doing engineering work, and so we want to bake that into the model and, and be able to do that.

14:57 Now the software landscape is a little bit scary today, in terms of how you would go about doing this, and it’s not comprehensive in terms of the possibilities, it’s probably three times as many things that a, that a company would have to look through to decide. So one of my roles is to help them make the right choices, because sometimes it’ll be different based on what they want to, to be able to do. So they’re GNA, they’re asking, what scanners do I use, sometimes it matters, like for a class one device, if they’re doing like a cranial helmet, it might matter more than another device, but they have this, they know they want a 3D print, but they just simply can’t get there through this mess.

15:36 Now radii is solving the rectification, is what they call it clinically, that aspect, in a way that is completely different from all these other programs that actually ask you to basically sculpt digitally, it’s doing that for you and then giving you controls over it. And so that’s why we’ve seen them as a, a forrun, a front runner in the space, because they’re taking that lift away, because, as Brent mentioned, it’s actually fairly efficient to do this by hand, it’s just not scalable in terms of talent. So if you go to a prosthetist and you say, do this digitally, and he says it takes, or, or she says, it takes me longer, I’m not going to do this, you have to actually provide benefits over the traditional process.

16:16 Now I have CAD design in there, but that’s actually not part of the, the production workflow, that’s part of the R&D workflow, the development. And so people are wondering, how do I design the product in the first place, what product, what, what, or if they’re doing one-offs, how do I do that. If we’re getting mesh data in, then going into traditional CAD is kind of a mess, so what’s the better way of working with that. And you, you, there’s lots of different ways to go from there, in term into automation, all with different benefits, and there are, you know, a number of softwares that could complete what I’m going to kind of show you, but they have different various advantages, and it may be better in one case or another.

16:53 So I’m just going to walk you through, because I think it might be kind of interesting thing, the actual step process, and, and designing this part. So, so this is kind of the level of, you know, what we’re helping automate today, but we want to see our customers move into something with more complexity, but with purpose. This is just a show car, to say, Hey, you could do this, or this, or this, but when clinicians to tell us that’s a horrible idea, you should do this instead. So this is kind of like a question to the, the market, what should we be doing, and I got really good feedback at the last conference to say, you know, why this doesn’t work, so now I can improve it.

17:31 But as far as the geometry process, I’ll just kind of, you know, go through the, the tree, and you can see kind of how it’s very different than modeling in a cad context, when you’re doing voxal and meshes to try and get consistency. And the reason we’re in voxels and messages, messages, is because they’re robust, particularly voxels, they don’t break, they can do offsets, they can do fillets all day long, and they just give you an answer instead of saying no, which is painful.

17:59 So you know we get this, this file coming in from radii, in this case, and it’s just the single surface mesh, right, what do we do with it. You know, so we’ll start to do a few things around the edges, just to try and increase Comfort, but then we need to create our Boolean tools, so we need to start creating ribbons, and, and this is going to be for the subtraction tool later. And then, in this case, I wanted to provide some, you know, hint at the limb shape, so it’ll fill out the pant leg in a nice way and show that’s a possibility.

18:32 So where do you, you take the other side, you know, if they have one sound limb, mirror it and bring that in, but then how do you combine those two models, they have really no correlation besides General Anatomy. So we have the shape of the interior of the socket versus the limb shape, I’m just showing two sides of the same limb, and you smash them together, and that’s not really helpful, right.

18:55 So what do you do from there, there’s, there’s processes that you can automate, to say we need to get to a, you know, first we’re going to do some smoothing, but even then that’s not respecting the geometry right at the r, where we need them to come back together, right. So we actually paint an attribute that says, in this location respect mesh a, and everywhere else respect to mesh B, and that gets us to something that has this constriction around the The Edge. You can imagine trying to do that in like solid works, good luck, right.

19:35 And then I’m just truncating it, because it’s, you know, it’s not going to be the full foot, as well we’re subtracting that internal volume, and we come up with a shape that is different on the inside than it is on the outside, and that’s your volume that you’re going to start working with from there.

19:50 I, I Incorporated some perforations, so we have a surface that we then scatter points across, and we C to point cylinders for each of the, the perforations we’re going to subtract. From there, there’s some Design Elements where we want to not have perforations along the top, or along the, the shin, or along the bottom, so I’m tracing out some key curves that allow me to then subtract from the surface mesh, to, to create a distribution of points for aoro fracture of the part. So now I have my lattice mapping for, for this exterior pattern.

20:31 So, so now, you, but you have all these parts disparately, the various balloon tools you want to, to work with, and they’re all overbuilt outside of the surface, because you’re going to trim it later. So you get some very odd looking midst steps, right, where you’ve, you’ve overbuilt the geometry, you start to combine in those lates, but then you start trimming away the outer surfaces, and it starts to look more like the end product, as well as now removing that interior volume and Perforating, just so you can, can see a little bit more of the detail, and, and you can see perforated part.

21:15 So the nice thing is that there, early, early adopters in the space, so i’, I’d love to get a picture of one of your products up there, brand, and terms Advanced Digital, you know, doing bespoke Solutions and now moving into scale. But there are products on the market using HPS technology mjf, where they are at various levels of automation, offering these, these services, but having talked to most of these companies, I can tell you that sometimes they’re spending four to eight hours to design a part, and that’s not scalable either. So we’re helping them figure out how to make this happen in seconds, instead of, you know, having an engineer constant, you know, working on everyone.

21:59 Just so you can kind of see the process here of what a more advanced prosthetic looks like, you can incorporate all kinds of different details. This is with a company called Corum Prosthetics, and they are incorporating these pressure pads using a boa system, kind of like a ski, you know, for ski boots and Snowboard snowboarding and stuff like that, but it allows someone to, to have a different level of activity, based on, you know, being able to control the fit throughout the day, or throughout their, their process of, of their activities.

22:30 And then we’re seeing it grow into other areas of orthotics as well, you know, cranial orthotics for babies, insoles for, for feet, and also ankle foot Orthotics. So it’s a quite an interesting space, great application of additive, it’s one of our key strategic Focus areas at HP, where we think that, you know, there’s huge amount of growth potential.

22:53 And just kind of as a general invitation to, to this crowd, you know, additive, computational design, and AI have all been accused of being Solutions without problems at various points, you know, it’s kind of like a hammer you walk around with looking for a nail. But helping UTS is a real problem to solve, with very rewarding problem to solve, and we need more people with your skill sets in that space to, to drive that forward. So there’s a, there’s kind of a open area of opportunity there, to, to innovate and, and have great impact, and then if you are going to do that, don’t start by yourself, we can help you kind of get a jump start in that area.

23:37 So as far as closing out, so we mentioned that part of the reason that Josh and I are connected is that we’re working with Veterans Affairs, on their, with their Advance office of advanced manufacturing for athetics and Athletics and Prosthetics, to deliver their digital solution, or their change over to, to printing their Prosthetics. And it involves a number of parties, as you see up there, all working together, under, you know, under a government contract, to be able to deliver Care at scale with meaningful improvements, and it takes that kind of effort to really, you know, move things forward, a lot of collaboration. So yeah, thanks for your time, and appreciate it, the.