A New Take on Prototyping that Could Save You Millions

A New Take on Prototyping that Could Save You Millions

17 min reading time

A New Take on Prototyping that Could Save You Millions

Presented by Doug Fankell, PhD, Structural Integrity Associates, Inc – October 12, 2018

Reading Time: 17 minutes


Dr. Fankell: Arlen is much more in the energy tissue interaction world. So hitting it with lasers, hitting it with ultrasound, things like that. I’m much more in the looking at the structural mechanics’ side of things.

Editor’s Note: Doug gives a fantastic presentation. At its conclusion, a lively conversation breaks out about predictive modeling versus simulation (from Arlen Ward’s presentation the day before) in bringing your medical device to market in the most efficient and cost-effective manner.

Doug Fankell: I’m going to talk about a “new” take on prototyping that could save you millions. It’s not necessarily new. And it’s going to be following up on some of Arlen’s presentation yesterday as well.

So just to give you a little bit of background about me, I’ve sort of been all over the map for school. I did my undergrad and master’s at the University of Wyoming, and UC Berkeley in structural engineering. So I have a weird roundabout path to how I ended up in biomedical device design.

I actually worked as a structural engineer for about a year and a half, decided I didn’t like it, and so I ended up at the University of Colorado in the Advanced Medical Technologies Lab working on both device design and then, more specifically, I was working on using analysis and finite element analysis to actually inform how we design these devices.

I graduated there a little over a year ago, I guess, and then started at Structural Integrity Associates where they’ve been doing finite element analysis of really hard nonlinear problems for a long time.

We have kind of 250 highly technical staff all over the country. We have some material testers, some material scientists, and they specialize in advanced nonlinear mechanics and FEA in highly regulated industries.

So they started in the nuclear power plant world which is very regulated and, since then, have expanded into different industries, such as aerospace; we’ve done a little bit with automotive, but I’m kind of their first foray into the biomed world.

They had all these tools, but they didn’t have anyone that really liked working on tissue and devices and things like that. So that’s where I came in.

So just to give you a background, I’m going to tell you a bunch of numbers that you guys probably have more understanding of and a better idea of than I do – about how much it costs to bring a device to market. And then hopefully, I’ll open your eyes on some benefits of using predictive computational modeling and then give you a couple of examples of projects that I work on that I can sort of talk about, and then just talk about where I think this is leading as well.

Background – Cost of bringing a surgical device to market

So here’s a general study that was done by Stanford in 2010 where they say it costs an average of $31 million to take a 510(k) product to market from initial concept, and an average of $94 million to take a PMA product to market.

So if we can reduce these numbers, I’m sure people would be very happy. And so where I’m really going to talk about how you can use computational modeling to help this is in the concept development and proof of concept areas, and then also how we can use it to inform our clinical tests and things like that.

Let’s make experiments better; let’s make our tests better so we actually know what we’re looking for before we spend all this money and then realize we have to reduce that.

Background – Time to bring a product to market

This talks about the time it takes to do this. So you see, on average, it takes a 510(k) product about 20 months just in that concept development and proof of concept phase.

I know that’s pretty widely varying depending on the type of product you have and what you’re doing, but just to give you a general idea. We’re going to hopefully help you reduce that time.

Background – Estimated costs of delays for a 30-person company

And then lastly, it comes down to money. Estimated costs of delays for a 30-person company, so eight-week delay. Sheldon et al. estimates that it costs a company about $1.8 million; 20 animal study, $5.5 million. I think that was for a year of work as well and a year of delay, and same with the additional hundred patients’ study.

So these are big numbers. So if you can avoid doing additional studies and reduce your amount of delays– I’m not exactly reinventing the world. That’s what everyone’s been talking about throughout this whole conference – let’s do things better to save money and make better products.

How Advanced Predictive Computational Modeling Can Help

So how can we use predictive computational modeling to help us? Well, you can reduce your time to market. You can speed up the design iteration process, so you can do more iterations. You can build a better device.

Again, I think there was a talk earlier today that was talking about, we need to get back to focusing on people and focusing on what’s our ultimate goal when we’re designing these medical devices.

We want to help the patient. We want to build better products, and a byproduct of that is hopefully we’ll also be making more money with these better products.

You can foresee potential errors that you might not have thought of as obvious before you actually are building these devices, and reduce experimental testing, and then also improve what you’re looking for in these tests, improve how you conduct these tests and the likelihood for FDA approval.

My boss asked me to summarize this in one concise statement. And it’s “I want to help produce safer, more effective products faster.” That’s the end goal. I know that’s very general but, again, I think if we don’t look at the end goal, we are going to get bogged down in the details and lose that.

Predictive Modeling

Arlen touched on this yesterday. Until recently, modeling of medical devices, I felt like it fit into one of two categories.

  • We’ve either looked at the device itself – but then how do you model it in vitro? With all the soft tissue that’s acting on it, how do you actually model those boundary conditions? That’s really hard. And then what’s it doing to the body? You’re putting something foreign into the body;
  • or we’ve looked at the tissue itself.

A lot of people have struggled at actually combining those two – looking at how medical devices interact with the tissue and how they actually work. So that’s what my PhD was focused on, and then also what I’ve been working to do now.

So predictive modeling, now it not only models the device but also the physics within the tissue when it’s acted on by the device.

So this could be simple like:

  • Deformation. If you have a needle and you’re poking it through the skin, what’s going on in the skin? What stresses do you need to reach to push it through the skin?
  • Fluid flow through the biological tissue. My thesis was developing a thermoporomechanics model of biological tissue I won’t bore you with. That’s a lot of math. If you ever want to see it, I can send it to you. It was like 2000 equations or something in my thesis.
  • Within the biological tissue. We can look at temperature, chemical reaction, tissue remodeling within the biological tissue.

And honestly, most of the time, we have to do many of these things and explain how they affect one another.

The Design Iteration Process

So again, this is a very simplified slide but I think it really hammers home the point of – this could be a device; this could be even just a little portion of your device – “You have some need that you have to meet, right? You come up with ideas of how you’re going to solve that need, you design a prototype.” Only, most of the time, I swear I spend half of my time in SolidWorks.

But you come up with the design, you come up with ideas for this design, then you produce. You’d say, “Okay, this is the best one. We want to produce this prototype.” Or maybe you come up with a simple prototype of three different designs, then you test it, and you go through this loop over and over and over again until you get down to the design that works the best in what you want.

Well, I’m proposing that we want to speed up that design. So you still have your need; you come up with a bunch of ideas; you prototype this digitally in SolidWorks or in whatever you feel like you want to prototype it in; and then we simulate the design performance.

Now, this could be as simple as doing stuff that we already have the tools to do. Or maybe we have to do some laboratory benchtop testing to validate the computational models.

A lot of the times, even just doing simple computational models, you discover new needs or you’re like, “Oh, this might not actually do what we want it to do.” So then you go back, come up with new ideas, come up with new designs, but you’re not having to build any physical prototypes and take that time to build those prototypes.

From that, you can also say, “Well, this one looks like it is the best design.” But now, instead of coming up with 10 designs and then narrowing it down to one, maybe you can come up with 100 or you can do an optimization study and narrow down on what you think the best design, produce that, test that. And then lastly, you get additional experimental input or you can look at what you might really want to focus on with your testing. It just helps inform that testing as well.

Example I – Arterial Cutting and Fusion

So now, I’m going to take you through a couple of examples because this is all great theory, but unless you actually see how it’s applied, it doesn’t really matter. Right?

So during my Ph.D., ConMed Electrosurgery was really interested in developing a model to inform and speed up their design process. So this was taken a while and it was of their tissue fusion devices which they used to both cut and fuse arteries.

And the goal was to develop a predictive FEA model – we knew we were going to use experimental data, but they didn’t want to use a ton of experimental data. And most of the data that we used was stuff that they were collecting anyways as they went along – and then use that to inform device design.

So I came up with a model. This I ran on my desktop computer with six gigs of RAM. It was back four or five years ago.

And this is kind of what it looks like. The top, you’re looking at the Isochoric Strain Energy, which is a fancy way of comparing things to stress and strain. It was a more informative value, I guess, and then temperature at the bottom.

And from this and then from a little bit more experimental testing, we were able to actually take these terms and come up with a way of measuring the damage in the tissue. And then actually I wrote a little user subroutine that modeled how the tissue was mechanically being damaged.

And we didn’t care if the cells were staying alive, we just cared if there was actually a mechanical strength left in the tissue or if it was being cut all the way through.

And we were able to predict pretty well. This is my favorite picture. I always have to put this caveat of “Not all the pictures lined up exactly like this,” but this test actually ended up doing that. We were able to say, “We’re going to predict the outcome of this surgery.” We were able to do it almost 90% of the time. And we were also able to put bounds on “you have to press this hard with this device, and you have to get it this hot, or you’re not going to see cutting or you might only see it 50% of the time.”

And we ran a whole parametric study on different artery sizes, on different temperatures and different pressures, and all kinds of different things that really provided some insight.

You see on the right side there, there’s sort of this little ridge. They wanted to get rid of that ridge, and I ran several simulations and said, “If you get rid of that ridge, you’re not going to be able to cut the tissue.” And one of the more experienced people there was like, “Oh, yeah. 30 years ago, we tried to do this without that and they were like, ‘No, we couldn’t cut the tissue.’” So little insights like that into the job design are pretty interesting.

So, really, these FEA models were used to iterate the jaw shape and the control parameters. We could run tens or hundreds of optimization simulations in the time it took them to build one physical prototype of these jaws.

And then we were able to do parametric studies. And there were still some upfront V&V experimental tests, but they were doing these anyway. But now they have these tools and this understanding of the tissue to where they just have to go back and validate their models. That’s a big thing in the nuclear world.

We have lots of good ways of validating models because they don’t really let you test anything nuclear anymore [laughs]. Or it’s very small scale. So just the statistics behind that as well.

Example II – Tubes

And then another example. I can’t talk too much about this one. That first one was more of a big picture – how does this affect our whole device. This was just a very simple part of their design process but it was causing pretty significant delays.

They needed a composite tube casing, and they needed special holes within the tube in order to house their tooling. And this casing needed to be able to be pushed through an artery, but if it crimped at all, it just damaged their tools pretty significantly.

And it was taking them six to eight weeks to get these special custom tubes. So they’d have to get the tube, they’d go test it in the lab, be like “Oh, it’s not as stiff as we need,” or, “It’s too flexible. It’s not what we need,” as they’re trying to push it through some pulsing arteries.

Including the initial benchmarking of the material which they had already done – they’d gone through several iterations of this – we were able to come up with numerous design ideas and run simulations of them in the amount of time it took them to get one. Or a lot of the time, my colleague over there said that they would order 10 different types because it was taking six to eight weeks for them to get these tubes in the [case?? 00:13:04].

So this is just one very small part of a design that was causing pretty significant delays in them moving forward with producing their product, where it was something we could really help them out with and say, “Yes, let’s run through several iterations on a computer and come up with the best design. And then you can have that produced instead of having to wait six weeks just to get a product that wasn’t working how you wanted it to do.”

Summary

So just summarizing everything, what can we do with predictive computational modeling? Well, we can come up with faster design; you can iterate more quickly; you can iterate more so you can converge on a better solution, and you get additional insight into the performance. Then you get a better design product. You hopefully see better performance during testing, and you see less testing or what you need of less testing. And then lastly, hopefully, you get a better product; it gets to market faster; and it costs less.

That’s the ideal world. And this is just one way that we’re helping companies do that.

Future

And then lastly, looking at the future, Arlen touched on the ASME V&V 40 that’s supposedly coming out with guidelines this year.

And then, one of my favorites, he also touched on the high computing performance scene, like if you have a laptop, you can run these hundred-core simulations now. And it’s a million degrees or 10 million or 100 million degrees of freedom. And the simulation is not a problem anymore.

And then lastly, my favorite quote was from a paper that just came out from the FDA, where they say they believe computational modeling is poised to become a critical tool for accelerating regulatory decision making and adopting this will be essential for advancing their mission.

So that’s speaking to where this field is going and how important being able to use these tools effectively is. And with that, I’d love to answer any questions.

Joe Hage: I’ll start. Dr. Fankell?

Doug Fankell: Mhm?

Joe Hage: Respect, man.
[laughter]

Most of what you said went over my head, but that’s when I know it’s really good. And I never really thought about the parallels between the structural work that you used to do and the structural work that goes into making a medical device. It’s just, I don’t think everyone in TV Land, certainly I, would have made that intuition.

You referred to Dr. Ward’s work yesterday. And both of you are on different planes as it relates to simulating. He talked a lot about as it relates to clinical trials and animal testing, and you talked about it from another angle. Where do you see your work overlapping and where is it quite different?

Doug Fankell: I would say mostly, it’s different in our expertise. So we both got our Ph.D. looking at these vessel-sealing devices.

It was kind of funny, but I’ve been meeting with him. We’re both in Denver as well. But we hadn’t met until about six months ago.

Joe Hage: Oh, you knew one another.

Doug Fankell: Yes, we do.

Joe Hage: Okay. Go over one another’s houses at all or…?

Doug Fankell: [laughs] We have coffee every now and then.

But he’s much more in the energy tissue interaction world. So hitting it with lasers, hitting it with ultrasound, things like that. I’m much more in the looking at the structural mechanics’ side of things.

And honestly, they do overlap some. Very rarely are we looking at one physics anymore. My Ph.D. thesis was looking at the fluid flow, and the water becoming vapor, and how it was expanding the tissue, and what was going on as far as the deformations of those tissues. So there is some overlap, but I would say I’m much more of the structural mechanics’ side of thing, looking at structural damage, not necessarily thermal damage, Arlen Ward: I think the last time that we met for coffee, this was one of the things that we talked about quite a bit because it’s easy to say, “Oh, we both work in simulation so we’re natural competitors, right? People that would work with me would work with him. So one of us will be losing out between that.”

But really, in the conversation, what we realized is, there’s a lot of different areas of expertise that people outside of the simulation world may not really have an appreciation for.

And looking at it more as “we have a network of people that each are focused in different areas,” it helps when a device company comes to me and says, “We have a problem.” A good example of that is insertion forces for needles. My response to that is, “I know someone you should be talking to.” Could we do it? Yes, probably, but I know somebody that can do it a lot better than we could, and point them in that direction.

And being that connector like you and I talked about before, I think, is going to be necessary in the simulation world as it relates to medical devices because we can’t expect medical device companies to have that knowledge of the entire area.

Joe Hage: One more thing and I’ll come right to you, Tor and Hilary.

Because I don’t understand the discipline like you doctors can, I think about design and development firms as “I’m going to delegate that. They’re going to design and develop it for me.”

And this term about structural integrity and all of these things, they never even occurred to me. I don’t know, – and Torey, you are a great person to hand the mic to – should I think most design and development companies have this expertise or do they typically need to find someone like Doug or Arlen?

Tor Alden: Sure. We talked about this a little earlier. So as product development firms, we have simulators or some of the lower-end ones. We get in trouble when we start working with non-Newtonian fluids and other things.

Even though we have these capabilities and we rush to make prototypes and prototypes, – you mentioned the needles puncturing through tissue. That was one of the issues that we had – the challenge for us was really to convince the client of the importance of getting the simulation versus going into actual [?? 00:19:00] testing.

And so we ran into trouble when we actually tested on animals because after one or two needle punctures, the needle dulled slightly and the whole instrument blew up in our hand. So we had to go back to the drawing board, and the needle became the issue. But if we had earlier involvement with companies like yourselves, we probably could have avoided that.

But again, the challenge is, how do we sell your services to our clients?

Joe Hage: So I’m curious, because I know time is a premium, you want to get the market. And I guess in most cases, time is more valuable than money.

But I wonder, if I were in this business and I had to make these decisions, I have a certain cost associated with you, but I might be able to avoid it if the testing goes fine and I didn’t have the dulled needle, so I saved. But when it didn’t work out that way, how much money did it cost to do it again, not to mention the time? And would that be exactly how much he charges or a fraction of it? Do you have a perspective on that? Or you, Tor?

Doug Fankell: I would say that’s exactly the argument that I’m making. You can avoid all these delays.

I’m trying to remember who was talking yesterday where they were talking about the cost of finding out these errors later on in the process. I mean, it’s 500-fold or a thousand-fold or 100-fold. It’s much more expensive.

And then also, just coming up with these computer models, while it’s expensive, animal testing can get really expensive really fast if you’re not careful. And so I would say you can both look at it from a time aspect. I mean, I can set up 15 simulations and run them overnight. I can press go on Friday and have an answer Monday morning.

Joe Hage: Do you think it would be a fair parallel to say that hiring someone like your firm for simulations is like an insurance policy, like I’m paying a certain premium to mitigate risk?

Doug Fankell: I would say, “Yes, that’s one way of looking at it.” Maybe not necessarily the whole way of looking at it. We also can help drive innovation by being able to test your crazier design ideas and things like that.

Tor Alden: I think we actually ended up using more finite element analysis to our advantage when we started looking at the different needle tips, whether or not it was a single probe or a trilobe. As we worked through it, we let the simulation do the work. And then we were able to optimize prototype, then go back and take that prototype and optimize it again.

But it wasn’t a clear yes or no answer, where I think in our world as prototype or product developments, if you CNC something out of ABS, you’re 98% sure it’s going to work. Where in a finite element analysis, it’s a little bit more fuzzy, I think, to people that are untrained in it.

Doug Fankell: Yes. It can be–

Joe Hage: I want to follow up with Tor and ask, “You have a better sense of your competitive set. As I said immaturely a few moments ago, design and development firm, you know, they’re going to design and develop. How many design and development firms typically have functional expertise like Doug’s in-house? Would you say quarter, half or third?

Tor Alden: So I think everybody will say they have it, but I think whether or not you’re using SolidWorks simulator versus ANSYS or whatever you’re using, there’s a different level of functionality and capability.

And then it’s also who’s developing the parameters of the test to make sure the test is actually working. So yes, we can say we have it, but are we doing it efficiently and correctly?

Doug Fankell: I have a whole soapbox about SolidWorks simulation because it’s making things easy [laughter]. But as Arlen was saying yesterday, unless you’re doing a good job validating and verifying A. that it’s doing what it should be doing, and B. that it’s actually simulating the physics that’s going on in the world–

Joe Hage: The reason I’m going a little beyond time on this is I think this issue is probably something that the folks in TV Land face a lot. Because I know, as I said, I would be clueless about that.

If my design and development firm says they can do that, they can do that. So what kind of questions would I ask to even know that “well, could ya?”

Because I’m thinking – and correct me if I’m wrong – most of your clients are companies like Tor’s that do design and development instead of the manufacturer themselves because they don’t even know that they should really be asking you. Is that true?

Doug Fankell: Yes. That’s definitely been the people who have been most interested in working with us.

Joe Hage: Is that the same for you, Arlen?

Arlen Ward: Yes.

Doug Fankell: I would say asking a simple question like “how are you validating this model? How are you proving to me that the model is doing what you say it’s doing?”

That was what I spent a big portion of my thesis on. Great, I developed this crazy mathematical framework. Now, actually, is it doing what I hope it’s doing?

Joe Hage: At the end of the day, I’m going to talk a bit about medicaldevicesgroup.net – this site that I lead and I want it to be extremely useful.

Just talking with you, I’m thinking it would be great. I think everyone in this room can contribute to it. What is the compendium of questions that you didn’t even know to ask, you didn’t even know it was going to be on the test? Because I’ll meet physician entrepreneurs who have a great idea, and they’re like, you know, does hmm. And they won’t know to ask these questions in the first place.

If we could put together a resource of– It’ll be exhausting and it’ll be intimidating, but if you show it to the right people, did you ask these questions? If you have electricity running through it, did you ask these questions? Does it have IoT? Put it. How will it interact? All these things.

So, a group effort is really the spirit of the community that I attempted to build. And Hillary, you have a question?

Hillary Sweet: Yes.

So I’m just wondering if you have anyone that comes to you, not on the large manufacturing side, but with an idea of “we have this problem and we don’t know where to start?” Do you have anyone who comes to you and says, “What material or what thickness?”

I’m thinking of a particular application with someone who’s working on a urethral stent that continues to clog. I’m just thinking this could be really helpful for something like that.

Doug Fankell: Oh, definitely. We definitely have had some of that. You get people that are again, like you said, almost sort of clueless – I have this idea, but I have no idea what to do.

I used to deal with it. I’m an adjunct advisor of a grad design team, and they came up to me and they were like, “Oh, we have this great idea and this great design.” And I looked at it. I was like, “Well, you have to make that three times thicker, otherwise it’s just going to break every single time you even–“

Just little things like that.

Joe Hage: Were they like, “What do you mean?”

Doug Fankell: Yes, exactly. And I asked them, “You are engineers. Have you done any mechanical analysis on this?” And they were like, “Oh, no.” I was like, “Don’t be scared of equations. It’s okay.”

Joe Hage: Well, you have 2000 in your paper. So where can we read 2000 equations in case we—

Doug Fankell: Uhm… I can—

Joe Hage: I’m joking. I have no interest whatsoever in reading that document.
[laughter]

Ladies and gentlemen, Doug Fankell. Thank you very, very much.

Doug Fankell: Thank you.
[applause]