Building a beautiful robotic hand is one thing. Getting it to do your bidding is another. For all the hand-shaped prostheses designed to bend each intricate joint on cue, there’s still the problem of how to send that cue from the wearer’s brain.
Now, by tapping into signals from nerves in the arm, researchers have enabled amputees to precisely control a robotic hand just by thinking about their intended finger movements. The interface, which relies on a set of tiny muscle grafts to amplify a user’s nerve signals, just passed its first test in people: It translated those signals into movements, and its accuracy stayed stable over time.
“This is really quite a promising and lovely piece of work,” says Gregory Clark, a neural engineer at the University of Utah who was not involved in the research. It “opens up new opportunities for better control.”
Most current robotic prostheses work by recording—from the surface of the skin—electrical signals from muscles left intact after an amputation. Some amputees can guide their artificial hand by contracting muscles remaining in the forearm that would have controlled their fingers. If those muscles are missing, people can learn to use less intuitive movements, such as flexing muscles in their upper arm.
These setups can be finicky, however. The electrical signal changes when a person’s arm sweats, swells, or slips around in the socket of the prosthesis. As a result, the devices must be recalibrated over and over, and many people decide that wearing a heavy robotic arm all day just isn’t worth it, says Shriya Srinivasan, a biomedical engineer at the Massachusetts Institute of Technology.
But there’s another way to tap into a person’s intended movements: by listening to the nerves that transmit the brain’s commands down the arm. Wires planted directly into these nerves can capture electrical signals to control a prosthesis. But the signals are faint, and small movements of the fine nerve fiber relative to the recording electrode can change or obscure the nerve’s subtle message.
So researchers have tried to boost nerve signals by connecting them to a muscle. Some have rerouted nerves from the arm into a chest muscle and picked up the strong electrical signal as a person contracts that muscle by thinking about moving their hand.
But surgeons must strip out some existing nerves in the chest to route in the new ones. As Clark puts it, “There’s only one car that can fit in the parking spot.” That means the procedure may compromise a muscle some amputees use to move their remaining upper arm.
For about 10 years, a team led by plastic surgeon Paul Cederna at the University of Michigan (UM), Ann Arbor, has been developing an alternative approach: Give the nerves new minimuscles of their own. The researchers isolate bundles of fibers from each of the major nerves in the arm and wrap each bundle in a chunk of muscle tissue roughly the size of a paper clip, often harvested from the thigh. The process basically creates a new set of finger muscles inside a person’s forearm or bicep.
Because wrapping nerves this way also relieves certain types of pain common after an amputation, hundreds of people have already had the procedure—but without the wire implants that could record from the muscles to control a prosthesis. In a new study out today in Science Translational Medicine, Cederna and UM neural engineer Cynthia Chestek describe the first test of that control step.
In three participants with amputations at different points along the arm who already had muscle implants, wires inserted through the skin near the muscle grafts could easily pick up their electrical signals, the researchers report. Even with an amputation up near the shoulder, a computer could interpret which tiny muscles were contracting, and by how much, to isolate different intended movements—a flex of the pointer finger versus the thumb, for example.
“The isolation that they get with these little muscle grafts is really quite remarkable,” Clark says.
Two of the participants—both with amputations at the wrist—opted for long-term electrode implants, which allowed further tests of their hand control. Using computer algorithms that “learned” to translate electrical signals into intended movements, the participants could prompt a virtual hand on a computer screen to assume any of five positions on cue. And when controlling a commercially available prosthesis called the LUKE arm, both participants could move the thumb to precise targets in space and pick up and stack a set of small wooden blocks.
Because the prosthesis relies on signals from nerves naturally involved in hand movement, participants could get it to move the way they wanted it to on the first try, Chestek says; there was nothing for them to learn. And without recalibrating the system, her team found that participants maintained the same degree of control after 300 days. “There’s no reason it would go away,” Chestek says. “The nerve is stable and happy.”
The setup isn’t ready for prime time. For now, wires tether participants to the lab equipment that reads and interprets the electrical signals. Chestek and Cederna eventually plan to develop a compact implant that doesn’t require wires that stick out through the skin. If the device can be optimized and win regulatory approval, it might offer amputees robotic appendages that are less of a handful.