Can prosthetic innovator Hugh Herr build a better socket-and-liner system?
This year marks the 20th anniversary of the patent for the Rheo Knee, one of the first smart prosthetic joints ever invented. At a time when Mark Zuckerberg barely knew his algorithms from his Adam’s apple, the Rheo boasted an artificial brain powerful enough to crunch a torrent of data about its wearer’s biomechanics, then adjust in real time to integrate seamlessly with each stride.
The device heralded a new era of responsive, interactive artificial limbs. It also helped make the patent-holder, Hugh Herr, the most celebrated prosthetic pioneer of the 21st century—maybe ever. In the two decades since, Herr has continued to build a more perfect union between human and bionic anatomy. Just as important, he’s helped make this technology comprehensible and compellingly cool to the masses. Millions of people have been introduced to next-gen prosthetics via popular media profiles of Herr. His 2014 Ted Talk, “The New Bionics That Let Us Run, Climb, and Dance,” ranks among the 25 most-viewed segments in that series’ history (18 million views and counting). Time wasn’t exaggerating when it dubbed Herr “The Leader of the Bionic Age” in 2011.
That title still applies, even in the vastly more crowded, faster-moving research and development ecosystem that Herr helped create. According to Justia, he has filed or been awarded more than 20 new patents in this decade so far. He’s breaking new ground in neuroprosthetics that link minds directly to machines, and he’s collaborating on new surgical techniques that improve post-amputation function and reduce pain. And now, at age 59, he’s taking on the most fundamental, universal interface between prostheses and people: sockets and liners.
To be more precise, Herr wants to transform liner and socket fabrication from a largely artisanal process into a science-based one. Like the Rheo Knee, this innovation pairs biomechanical principles with reams of user-specific data to produce a device that “knows” its wearer’s body and works with it, not against it. And, like the Rheo, Herr’s new interface promises an unprecedented level of function and comfort.
Too good to be true? You can see for yourself at Bionic Skins, where Herr is piloting the new system. A prosthetic clinic and research facility in suburban Boston, Bionic Skins opened last summer and is actively seeking new patients. The socket-liner innovation is only available to below-knee amputees for now, but patients of all amputation levels are served at the clinic. Learn more about the company at bionicskins.com.
Herr spoke with Amplitude early this year to describe Bionic Skins’ technology and its implications for the future. The conversation is edited for clarity and length.
A lot of resources are being poured into improving socket comfort. Where does Bionic Skins fit in the big picture?
In my opinion, the dominant problem of today’s artisanal approach to liner and socket design and fabrication is that the pipeline has no memory. I’ve been an amputee since 1982, and I’ve probably received about 50 limbs between 1982 and today. If I were to go into a new clinical facility, we would be starting all over, as if I haven’t been an amputee for decades and decades.
[At Bionic Skins,] we embed artificial intelligence, physics, and clinical knowledge into an algorithm that’s personalized to the individual, and the algorithm predicts the optimal liner and socket. And once we build that algorithm, we’re done for the rest of that person’s life. We know the exact anatomy and physics and the comfort requirements for that person. All we have to do is update their image data and filter it through their algorithm. So you won’t spend an entire month getting three check sockets and end up with something that still doesn’t fit.
That’s the dominant value proposition. It’s not artisanal; it’s based on physics. We can precisely apply load and anatomical regions in a repeatable manner.
The terms artificial intelligence and algorithm are ubiquitous these days, but I think many people don’t really understand what they mean. Is there a layman’s way to describe how the algorithm translates data about your anatomy, and the physics of your anatomy, into a comfortable socket?
It’s tricky, but we’ll take a shot at it. The dominance of our algorithm is physics. We do use AI tools, but those largely allow us to move faster computationally; they offer an efficiency. But what’s predicting fit is actual physics.
The imaging data tells us where the bones are, where tendons and ligaments are. And, of course, it tells us the external shape of your limb. Those images are segmented and sliced using AI tools, and the computer builds an entire 3D model of the residuum.
We then use state-of-the-art biophysical models that tell us how those tissues react in a socket and liner. In the computer, we don a liner and socket. We add a digital representation of a liner, with all the physics of that liner, and we have a digital representation of a socket. We then can simulate loading conditions, simulate a person standing in the prosthesis. The physics tell us what the shear and pressures are at every anatomical point. We can then start varying the properties of the liner and the socket to reduce pressures where they are too high and will ultimately lead to tissue damage and sores.
Many clinicians are using 3D scanning and printing to achieve similar goals—faster turnaround time, more precise fit, and so forth. What sets your process apart?
No one’s taking the data set we’re taking. Most people are scanning the external shape of the residual limb—that’s a data set you can get with your iPhone. But it tells you nothing about tissue properties. You don’t know where the bones are, and you have to know where the bones are to predict fit. You don’t know where the soft tissues are, and how all these anatomical points vary spatially. We can look right at the fibula head, the patellar ligament, the distal aspect of the fibula, the tibia crest, [and] the posterior wall and ask, “What is the exact load contribution of that anatomical point?” We actually answer that question for each individual.
So you’re creating a detailed subsurface architecture of the limb?
That’s exactly right. No one’s using CBCT [cone-beam computed tomography] scanning or any similar imaging methodology to get the comprehensive data that our approach requires. And then the algorithm does computations at incredible speeds, so we can iterate shapes and stiffnesses and do hundreds of experiments before we fabricate. You could never do that if you were actually building liners and sockets and testing them. We can find the optimal interface digitally; then we fabricate.
We also inject clinical knowledge into the algorithm. The algorithm predicts an optimal liner and a few socket variations, which have subtle differences in how much load is applied at every anatomic point. We empirically test those socket variations on the patient. So our pipeline enables an incredible clinician to actually hard code their knowledge and intuition into the algorithm. If a clinician has experience that’s not embedded in the physics or the AI—“There’s a neuroma here, and my 20 years of clinical experience tell me the load should be reduced”—we’re giving the prosthetist these powerful tools to treat their patients much, much better.
In other words, humans can override the AI.
That point is really important. And it’s not just the clinician. We also embed the patient’s feedback into the algorithm. We ask, “How does this feel?” and we listen, and we embed that into the mathematics, too, so five years later we don’t have to repeat the same thing. Over time, the algorithm becomes personalized. It’s Tom’s algorithm that Tom helped build, that all of Tom’s clinicians helped build, that AI was a partner in, and that Newton’s physics was a partner in.
Suppose I gain 40 pounds. How does that affect my algorithm?
The fundamental question we’re answering is, “Where should I distribute your body weight?” We’re dealing with fundamentals of fit that are invariant, even if you’ve atrophied or hypertrophied. You can atrophy over five years, and it’s likely that the same fraction of your body weight should be applied to the fibula head. It won’t be the same pressure force, because you’ve changed weight and you’ve changed size. But it’ll divvy up on the [same] fundamentals that led to you being comfortable before.
Is it possible that we have to deviate from those fundamentals across your lifetime?
Maybe. If you developed a bone growth or a neuroma that didn’t exist ten or 20 years ago, we would have to adapt the algorithm. If you give the algorithm new data, it’ll put the load where it needs to go. But if there’s no fundamental change, then we already know how to give you a really well-fitting socket and liner. That’s the memory piece.
How portable is this system? If someone comes to Bionic Skins’ clinic and has a definitive socket made using Bionic Skins’ algorithm and scanning technology, is the data in a form that their clinician back home would be able to use?
At that first fitting, we’re building you an entire limb, not just the socket. So you would go home with an entire limb. Years later, when you need a new limb, you can get reimaged wherever you are and send us those data. Then we can update the interface and ship it to you.
Essentially, they’d just need a new CT scan, which can be done anywhere.
That’s correct. There are even some revisions we can make after that first fitting that only require an external scan of your limb, which you can make with your iPhone. So in some cases, you can image your residual limb from the privacy of your own home, send us the data, and we can FedEx you another interface.
How do you envision this scaling up? Is the goal for Bionic Skins to build out a network of clinics like the one you have in Boston? Do you envision licensing the technology to existing practitioners so they can incorporate it into their own practices?
The first step is to get our current clinic well established. Then we’ll start duplicating that model in other regions at regional centers of excellence. They could be Bionic Skins clinics, or they could be strategic partnerships with established clinics. We don’t think it’s going to require that many clinics to cover the United States. We’re talking on the order of ten.
One last thing to think about is the arc of optometry. It used to be that you came in and you were given three lenses, and they’d ask you, “Which one is better?” Now when you go in, there are really powerful imaging tools where one’s eyes are digitally quantified with really comprehensive data, and lenses are manufactured based on the physics of your eye anatomy and the optics of vision.
The arc of that industry is the arc that O&P’s going to take. We’re just way behind. What Bionic Skins is doing puts us into the modern age.
[Correction: The Bionic Skins prosthetic clinic serves patients at all amputation levels. It only offers the science-based socket-and-liner system to below-knee amputees at this time. The print version of this article incorrectly stated that the Bionic Skins clinic only serves below-knee amputees. Amplitude regrets the error.]
