Capstan Medical’s R&D head discusses the heart valve and robotics startup’s tech, engineering challenges and solutions, advice for others in medtech and how to join his team

As you move from the right atrium to left atrium, how can you minimize complications from the transeptal puncture?

When I started learning about this procedure, I said, “What? You punch a hole through the heart? What are you talking about?” For people who aren’t familiar with mitral procedures like this, it is the standard way to do it. You puncture that hole with a guidewire and then you use a little balloon catheter to blow up the hole a little bit so you can get our 11 mm diameter instrument through it. Minimizing damage is where having the robotic control can be a really nice advantage. We work to optimize our kinematics of the instrument so we can move to where we need to move to in the heart. Part of that is you can get around the corners, you can get to the target, but also the shape of the delivery system is the correct shape for the procedure. So now you have an instrument that’s designed to be optimal for it, but you’re also controlling how it moves under robotic control. So we have the ability to pick these points in space and don’t let them move too much. If your catheter is going through that puncture, you can move the instrument and not apply a lot of lateral load to that puncture. Because we have control over those kinematics and because we have that feedback, we’re able to hopefully make a real difference there.

What are the advantages and disadvantages of using robotics versus human or physical delivery and deployment of the valve?

On the advantages side, if you think about the motions you have to make to go around the top of the heart and land in the mitral annulus, it’s a multi-degree-of-freedom move that you have to make. You can say, “What if we do a catheter with knobs and levers marionette-style to do it?” There are attempts at that. It is possible to do it. But the amount of mental energy you have to spend to do that is pretty high, because if I want to move the tip of the instrument this way, which of the five knobs do I turn to make that happen? With a robot, we can say, “OK, I want to move the tip of the instrument this way.” We have a joystick on our controller where you push up and then we figure out what all the motors need to do to make that happen. You turn what is a complicated, multi-degree-of-freedom problem into something that’s a simple one-degree-of-freedom action for the physician. The negative side is the stuff that we work to make better. It’s a more complicated system — you have motors and sensors — so we at Capstan need to make sure the device is designed appropriately and that for the different failure modes that can happen, we have good mitigations. That’s our job. That’s what we’re here to do. We have experience doing that in previous companies, and our goal — and I’m confident we can get there —  is we make the system be safe and simple for the user by absorbing that complexity on our end.

Is all this navigation pre-planned? Is there any in-flight adjustment or compensation?

We pre-plan the path that the instrument’s going to take, and we do that based on CT data that’s taken of the patient a couple weeks before the procedure. We use that data to figure out the trajectory you want it to follow. We program that into the robot controller so once everything is aligned in the procedure, as you push forward, it’s going to try and follow that trajectory. In reality, you’re not going to go exactly where you want to go because of a couple of reasons. That pre-planned trajectory might not be perfect. Your CT image might be a little bit different than your day-of image because of gravity or whatever. Instrument hysteresis can cause it to not follow the path perfectly. We have a notion of coordinated, path-following motion, but with the ability for what we call internally “tweak modes.” So you’re trying to follow this path, and then you can use the joystick to say, “OK, move down just a little bit,” to get back to where you want to be, and then those trajectories are updated based on where you’ve moved, where you tweaked or adjusted the instrument tip to. The physician is always in control of what the instrument’s doing. The physician pushes forward or pushes go to follow the trajectory. It will continue to try to follow that trajectory until the physician stops saying, “Go.” And then they have the opportunity to adjust to get back on to exactly where they want it to be.

Are you teaching or programming the robot to follow the trajectory based on CT before the actual surgery, and if so do you do a trial run?

We have simulators. We trial it, fully digital. We also 3D print models of the patient’s anatomy, whether it’s a full heart or the septum and the mitral annulus in 3D space and put it in a tank, and we’ll steer and drive the instruments through that to make sure we like the paths that we plan and that they’re going to deliver the valve where they need to. Eventually, when we’re selling lots of these and doing lots of cases, we won’t be doing that for every patient, but right now it’s an incredibly powerful way to learn and to make sure we’re going to get the results that we want in the cases.

Are you using your own developed software to program the robot, or is it a third party software that you’ve integrated into your system?

A little of both. We use a variety of commonly available libraries that are well developed and well tested. The controllers, the path-following algorithms, the visualization, we own that, we develop it, and we develop it kind of on top of those libraries

Is there active software as part of the procedure providing the surgeon with guidance, and if so, what is the software doing?

Our software is actively running to define that path, and then define the motor motions that have to happen in the robot to follow that path. We also are reflecting to the physician a model of what we think the instrument looks like, as well as information about how close the instrument is to any limits, like a joint limit where you can’t bend any farther. We have kind of a lightweight UI that we’ve developed that we can show to them during the cases. What specifically we show and how we show it and what’s the most valuable thing in the heat of the moment is a really big discussion. You can always show more, but there’s a lot of value in figuring out what are the minimum number of pixels that we can light up to give information to the physician, because they have their brains wrapped around a whole lot, and we need to make sure that we’re giving them just the additional stuff they need. Another one of the fun parts of this challenge is understanding how do they think about things as they’re operating and how can we make sure that what we do lines up with how they’re thinking about the world? … It’s pretty incredible. I was in a couple of cases in Houston two weeks ago. They have these giant monitors over the patient and there’s a fluoroscopy image, two different blood pressure and heart rate monitors shown, a transesophageal echo being shown, beeping and data and noise and all this stuff going on all the time. Finding our place in that and making sure that we’re not just adding more to the overload is a really interesting challenge.

How does Capstan determine the location of the instrument during the procedure?

We don’t have any sensors in the delivery system itself. The location is determined by the fluoro image. We see what the shape of the delivery system is in fluoro, and then we tell the controller basically what is that shape that we’ve looked at from fluoro. It’s a very manual, fairly laborious process right now. You can imagine there are a lot of things we might want to do there to automate that and to better integrate with some of those imaging systems. The sensing, the imaging, the data fusion, all of that is a very active area of R&D for us right now.

Let’s talk a little bit about some of the critical components in surgical robotics and structural heart. Should we start with the robot?

On robots, the motors, sensors and brakes end up often being the critical components. There’s a lot of them. Every degree of freedom on a robot has at least a motor and likely a gearbox, two or three sensors. All those things have to work together in harmony to get your system to work. And especially as you get closer to the patient with the robot, size constraints start to really matter. You want those motors to be small so you’re not taking up a bunch of space over the patient. As the motors get smaller, they’re less able to produce the amount of torque that you need for a given amount of power input, so they get hotter. Balancing how small can you get it while still maintaining the thermal performance that you need, there are lots of trade-offs there. And you want to get pretty close. You want to get just about as hot as you can, which isn’t that hot. It’s in the low 40s (degrees Celsius), which feels warm, but if you’re going to have this thing in contact with an anesthetized patient for a long time, even those low temperatures can do damage. We take a lot of care to make sure that we’re being safe. On the sensor side … we use magnetic sensors to figure out where the motor is. Those need to be reliable, they need to be accurate, they need to not generate too much heat, they need to package into the volume that you have. Lots of trade-offs and three-dimensional Tetris gets played around that. There are lots of strategic suppliers we work with there to make sure that we’re getting the best stuff and the best supply chain for it as well.

And how about the catheter?

The catheter has some pretty interesting components in it. The distal end, the working end of the catheter, has lots of very small metal components, machined stainless steel components with features that are low hundreds of microns in size. Doing that design work, getting manufacturers that can actually make it and clean it and inspect it, lots of work goes into that. There’s some cool stuff on the catheter around the coatings and the sheaths that go on it that I’m really just starting to learn about.

And the valve?

The valve is equal parts magic and technology packed into there. Nitinol is a big one. I know you guys at MDO love nitinol. I’m getting more and more experience with nitinol, which I’m pretty excited about. The skirt that goes around the valve is a really interesting, custom-woven polyester fabric that gets made for the valve. There’s one or two suppliers in the world that can do that. And the five, eight or so components  that go into that valve have to be perfect. They have to be well designed. They have to be brutally tested. Lots of work goes into that.

What sort of materials are uniquely useful in surgical robotics and structural heart?

The sheath on the catheter delivery system needs balance. It has to be strong. You don’t want it to tear during the procedure. It has to be lubricious. As the sheath is wrapped around these metal components and as those metal components move, they have to slide relative to the sheath, so how they slide in there matters, and it matters to the performance of the device. You’re trying to find these Goldilocks zones of balancing a lot of different things on — I don’t know what the thickness of the sheath is, maybe 200 microns thick — it’s a really delicate balancing act. On the valve side, those woven skirts are super interesting. They do a lot of work. That’s the actual material that’s touching the patient’s anatomy and is left behind in there. So you really have to make sure that stuff is dialed in.

You mentioned nitinol. What’s it been like working with that special alloy so far?

I haven’t had much experience with nitinol before I got here. I’ve obviously read about it. I’m a giant nerd. I love materials. I had some nitinol wire sitting on my desk. But the biggest thing for me is maybe the most obvious, the incredible amount of strain you can put into that material before it yields. It kind of breaks my brain a bit still. You get calibrated to the materials you’re used to: aluminum, high-strength steels, plastics. And you know you can bend this thing this far and it’s not going to break, and you kind of understand that feel. You grab these nitinol valve frames and you can just crush them and they’re totally fine. It’s really remarkable to see what the material can actually handle and then recover — and handle hundreds of millions of cycles of strain when it’s in the patient.

How many cycles are you designing the valves for?

The valves that we’re designing right now, the first standard that we’re targeting is 200 million beats — 200 million cycles — which is one of the ISO standards for valves. We’re still discussing some of the longer term targets.

Are there advanced manufacturing processes or just plain interesting manufacturing methods that you’re commonly using in structural hearts and in robotics?

The valves are all hand-sewn by incredibly skilled technicians. You get a nitinol frame, you get the fabric skirt, you get bovine pericardium valve leaflets, and a person spends about 40 hours under a microscope with tiny sutures sewing all of those pieces together onto that nitinol frame. And you get this beautiful thing at the end of it. That’s the standard of how it is done in the industry, but seeing that process was pretty incredible. The amount of skill and talent that they have and the precision and quality to do that and the fact that it’s all done by hand is pretty wild. On the advanced side, a lot of the manufacturing of the robot itself is bolting motors and stuff together kind of like building a car. It’s screws and bolts, things like that. There’s a lot of technology in the calibration of the robot and the delivery systems. You use computers and use the robot to drive the delivery system around and make sure it moves the way we want it to, what sort of things can we compensate for like friction. You can use the data during the manufacturing process to make the product better.

Are you using laser welding now or have you previously?

I’ve done some parts that have been laser welded in past lives. And on our delivery system in particular, there’s a ton of laser welding. The tube itself is laser cut. They cut all sorts of profiles into it to define the flexibility and the way that it bends. But then the disc segments that kind of flex around, they get welded onto the tube, they get welded onto the more distal segments. The shaft that rotates the little capstones to unfurl the valve, there’s a few different laser welds in that component. Lots of laser welding goes into those instruments for sure. It’s a core competency of ours.

Would you share a little bit of what you’ve learned about challenges and solutions to help other people innovate in medtech? Can you talk about some design engineering challenges that are common in surgical robotics and structural heart?

Interfaces. Anytime two things have to come together, the difficulty is in making sure it works the way you want it to. An example in our system is the delivery system. The catheter is a single-use disposable that gets put into the patient before the robot is docked to it. Now you have this catheter system that’s in the patient and you have a robot. You need to bring those together and dock them, and that needs to be smooth, easy, incredibly reliable, and then you have to transmit the torque for 10 little motors that have to spin 10 little pulleys in that instrument accurately and they have to engage without moving the catheter around. You can’t have undue friction. You have to know where everything is. There’s a lot that has to happen in that interface to get the product performance out of it that you really want. Some specific things on robots, I mentioned thermal earlier. Thermal drives a lot of what we do. We can’t get too hot. In delivery systems, I mentioned the sheath rubbing against the metal parts of the instrument that are inside of it. That friction causes what’s called hysteresis. So hysteresis is: I pulled a cable to go left, it’s going left, and now I want to go right and you start pulling the cable to go right, you have to pull that cable for a while before the instrument actually starts moving. And it’s because you have to unload the force that you put in with the left cable. It’s because the cable going right is stretching as you’re pulling on it. And then all that friction in the instrument that I mentioned, you have to build up energy that you pull through with those cables before it actually starts to move. So that friction and the compliance of this instrument that’s a long, flexible, bendy snake is a really hard thing to balance. So we use software modeling, we do a lot of a lot of heroics there to make that work the way we want it to.

And Capstan Medical has got different teams working on all these different parts, all these different problems?

The teams are my favorite part about this company. We’ve got some great people working on these problems, and the atmosphere is incredibly collaborative. My organization, there’s the robot team, the delivery system team, and the implant team. Those teams all work quite closely together at the interface points. But also, thinking about the hysteresis and the delivery system as an example, what do we fix in the delivery system versus what do we try and fix with software versus can we add different sensing technologies or feedback technologies into the robot? There’s constant trade-offs you’re making there. Those conversations and those trade-offs are the fun and exciting part of doing what we do. I’ve really enjoyed getting to know the people here that are working on that and really leading those conversations.

Every medical device developer is trying to get feedback from users and patients. Do you have any advice for getting feedback from surgeons and other users at all stages of development?

The big thing for me is to show. Show prototypes, show ideas, get feedback that way. You can really start a conversation when you show a prototype of something — let’s talk about it. It’s really important to go observe cases and observe the way the physicians are using instruments and using technology in their environment, because you can ask questions about that. You can ask them to imagine they’re using the catheter and turn these knobs, what do you expect to happen? They’re going to get about 50% of that right. It’s like asking you when you’re driving a car what are all of the 20 things that you’re doing? You’re going to know a few of them, but you’re not going to be able to be super precise about that. But by going in and very carefully watching and understanding what they’re doing, understanding how they’re thinking about the problems, understanding their mental model of what they’re doing, then you can ask better questions. Then you can better understand who are these people that you’re designing for, what do they really need out of the product?

It’s so much more than just the physicians, right?

The physician’s the center of the show in there, but you can think about the operating room staff all around the physician, the hospital staff, the people that have to clean the instruments. There are a lot of hands and minds and skill that has to go into using the product as a system. Understanding their point of view and working with them helps you make the make the best possible product.

And you’re hiring for some of these roles?

We’re hiring for some engineering roles. We’re also hiring clinical development engineers, which are people who really specialize in understanding the clinical needs, the clinician’s needs, the patient’s needs, and help translate that into engineering requirements. If you know a great one or you are a great one, call me up.Related: Capstan Medical is hiring — here’s what CEO Maggie Nixon’s looking for

Do you have any advice for people applying to join the team?

Apply. We look for talent. It doesn’t necessarily have to be in surgical robotics. We’re talking to some software people that are pretty far outside of the specific discipline. But bringing in different perspectives and different skillsets, there’s real value there. Don’t be bashful. If you’re passionate about this and you have some skills and experience that you think might line up, don’t hesitate to reach out. And when we’re talking to people, I mentioned our culture earlier and how much I love our culture, how much our team loves our culture. We want to make sure that people that we’re talking to are good communicators that fit well with the team, that clearly want to collaborate and work in a highly collaborative environment like we have.

What’s it been like going to a startup after working inside a larger company?

My first startup experience was I left Intuitive and went to a 3D printing startup called Carbon. I think Intuitive was 3,500 or 4,000 people when I left, and Carbon was 80. I left because, I wanted to go to a smaller company, I wanted to be freer to move. The medical device industry has a lot of process. I got there and I said, “Oh, man, it’d be nice if we had some process.” It was a little interesting for me as somebody who doesn’t really have a lot of process instinct, but to realize the value of it. It hammered home the value of making sure that you do have good design processes and communication strategies and things like that. When I left Carbon, they were about 400 people, and now I’m back to a smaller company. I love how fast we’re able to move, how simple the communication is. The R&D organization is 25, 30 people. We all know each other. We sit next to each other. You can really make things happen quickly. It’s really exciting. A lot of energy.

At this point, does Capstan have to hand off engineering to manufacturing, or is it all one group?

It is two groups: the R&D team and an operations team of manufacturing engineers and quality engineers, the technicians who build the things. So there is that handoff there. I try to not think of it too much as a handoff. We’re not trying to put something in a box and hand it over to them. The interesting part of the design transfer is the continuum of transfer and trying to pull the operations teams in early to make sure we designed this thing so those guys can actually assemble and test it, those type of questions. And then making sure the R&D team stays with the products through that transfer and are there to support on the other side. The collaborative culture we’ve got at Capstan, the collaboration’s not just within the R&D team. The collaboration across from the R&D team into the operations teams is critically important.

Do you complete all of your engineering work in-house, or do you outsource it to other development companies?

For the hardware, the majority of it’s in-house. For software, we have a blend. We have an in-house team and then we have a couple of consulting teams we work with as well, so we’ve got a mix.

Do you have any advice for device designers and engineers generally?

A few things. Prototype a lot. Working in CAD, working in code, you can get a pretty good idea of how something is going to work, but you have to get it into the physical world and really understand if it’s going to work. I was having a conversation yesterday about design review culture and communication culture within projects, and often there can be a reluctance to talk about designs, talk about your ideas, talk about concepts early. I would encourage people not to be afraid to talk about stuff when it’s an early idea and collaborate with colleagues and peers about, “What do you think about this idea? How can we make this better? Where can we push? How does this fit with how you’re thinking about stuff?” Leaning on that communication and that collaboration is really important. The last thing is something I’ve said to lots of people: Curiosity is valuable. Be curious about what you’re working on. If you’re working on a nitinol stent, be curious about nitinol but also be curious about how are they making this stuff? What are all of the steps involved in making the material? How does everything around the thing that you’re designing work, and how does it all come together? Because if you understand those interfaces and understand the stuff around the thing you’re designing, you’re going to design a better thing.