People with limb injuries severe enough to require an amputation have few options to regain meaningful function in their arms or legs. Commercial prostheses, even modern ones, don’t come close to what nature created, enabling movement that feels disjointed and artificial.
In a new study published Wednesday in Science Translational Medicine, a team of University of Michigan surgeons and computational scientists report a new procedure that captures electrical signals from nerves in the arm severed during an amputation and uses them to guide fine movements of a prosthetic hand. The work is one of a number of efforts underway to better integrate human physiology with robotics to improve the functioning of artificial limbs.
In the new technique, a flap of muscle is wrapped around the severed nerves to allow the nerves to grow and fire electrical signals. The signals are then picked up by an implant placed in patients — the implant serves as a connection between the nervous system and a computer that patients are later hooked up to. Finally, a machine learning program interprets those nerve signals to allow patients to move a prosthetic hand seamlessly.
The procedure described in the paper was initially tried on four patients, and the researchers report that the implant worked well in these volunteers for up to 300 days, which is how long the patients were observed. The patients were able to pick up small toy blocks and food cans and make a fist or pinch fingers together. The technology will need to be miniaturized and made wireless.
This is not the first such attempt to improve the state of prosthetics. Researchers at Brigham and Women’s Hospital and MIT, for instance, have developed a new amputation method — thus far tested in people requiring below- and above-knee injuries — to help preserve patients’ sense of proprioception; the team has reported that patients who have undergone the experimental surgery feel as though the advanced prosthetics they wear are a natural extension of their body.
STAT spoke with Cynthia Chestek, a biomedical engineer at the University of Michigan and an author of the new study, to learn more. This interview has been lightly edited and condensed.
Why is it difficult to integrate prosthetic control interfaces with nerves?
During amputation, a nerve is cut and while that nerve continues to carry signals about intended movements, it’s really hard to get those signals out. People for many, many years have tried to get prosthetic control down from a nerve, but the physics is just terrible. It’s very hard to record these very small [nerve] signals, and it’s hard to put anything inside the nerve because it causes a lot of scarring. The state of the art right now is to record from whatever muscles are left at the skin, and that both gives you a very small signal and usually it’s not the signal that you want. So, for example, you’re often using an elbow signal to make a hand open and close or operating a foot switch to make a hand open and close.
What is the advancement here?
Our surgeon developed a technique where he takes a very small piece of muscle and he wraps it around the end of the amputated nerve. And then the nerve goes into that piece of muscle, and then what would have been an extremely small signal becomes a large electrical signal.
What happens next?
So, then my team takes these electrical signals … and we apply a variety of machine learning algorithms to them. And this enables our study participants to then control a prosthetic hand in real time. And these are the largest ever nerve signals ever recorded from a human being, and we’re showing what that enables us to do with a prosthetic hand. If you’re trying to get a signal from a nerve, you’re actually closer to 10 to 20 microvolts, whereas we get hundreds of microvolts and sometimes more.
Why does just wrapping a piece of muscle lead to this expanded nerve signal?
The nerve is looking for a target to regrow into. And the nerve keeps growing until it finds something. Another major benefit of the surgery is as the nerve is growing into the muscle, it prevents a lot of pain [in a condition known as neuroma]. But now the nerve is happy, it’s found its muscle and the reason the signals get larger is because as muscles fire, they create this big electric field around them. So one nerve might be controlling a whole piece of muscle and that’s going to create a much bigger signal.
What do you ask the volunteers in the study to do, and how do they feel about the prosthetic?
We just ask the person to move as they normally would. We calibrate the machine learning [to these movements]. But we learn from that information and then use those signals to control the hand. So if the volunteers want to move the thumb across the hand, they just have to imagine it because the learning is in our algorithm, and not in the person. And importantly, after this surgery, they have so far all felt like they can move their fingers when they’re flexing that small muscle graft.
Does it take long for people to get used to it?
It works on the first try. We were able to just ask people to make a bunch of movements that we’re showing them on the screen and then they’re able to replicate that with the hand. We’ve increased the difficulty [of the movements] over the last couple of years.
What’s new about the algorithm?
We’re using a lot of the same sort of algorithms that are under the hood of autonomous vehicles. But what’s new that patients are able to do is they’re able to control the position of their fingers better. There’s no commercially available system in which [they] can precisely control the location of the end of [their] thumb. So I think probably the coolest thing that we were able to show is our few participants positioning the point of their thumbs in two dimensions, and that’s really useful if you’re trying to pick up things. If you can’t orient your thumb around an object, it’s very hard to grasp anything. And I’m not aware of any other way of doing that because it really requires nerve signals since so much of your thumb muscles are in the hand.
The cool thing about this surgery is it works on any type of amputation. That’s the good news. To take this [technology] home, this is still a report of the pilot study — this still needs to be replicated to work in more people moving forward.
We are trying to figure out how to get people off the computer cart. Everything that we’ve done so far has been six feet away from a computer cart [and people come in once or twice a week for the trial]. We want people to be able to do this with an implantable device so we can move away from the cart. We hope that, however many years it takes, this is something that one day enables people to control the fingers of the prosthetic hand at home and in their daily lives.