A team led by researchers at Chalmers University of Technology, Sweden, has decoded leg movements directly from the remaining nerves in people with transfemoral amputations, the first time it has been achieved, they say. Using novel implantable neurotechnology and a method based on the nervous system’s language that uses artificial intelligence (AI), the researchers could interpret detailed movements—even the will to wiggle toes. The research team expects the technology to lead to prostheses that feel and act more like a natural part of the body.
“When you tell your body to move, signals travel through the nerves to the muscles which carry out the action—even if the limb is no longer there,” said Giacomo Valle, PhD, assistant professor at Chalmers and one of the study’s senior authors. “This means you can find all the information needed within those nerves. The major challenge is extracting that information and understanding the neural code behind it—and that’s been the focus of our work.”
According to Valle, the ability to read and interpret movement signals directly from within nerves is key to developing future prostheses that are more responsive and intuitive.
“If an implant can be connected directly to the remaining nerves, instead of through residual muscles, you can use exactly the same natural signals used to move your limbs. It greatly increases the potential to create prostheses with natural control, sensory feedback, and unprecedented resolution,” he said.
However, extracting nerve signals directly from the remaining nerves of amputees is extremely challenging. Very few studies have been successful, and all have focused on the upper limbs. The research is complicated by the fact that the remaining post-amputation nerves produce weak signals that are difficult to capture reliably.
The research group succeeded in meeting this challenge with a new approach focusing on people with lower-limb amputations, in which the key role is played by a neurotechnological implant, combined with a new AI algorithm.
The same type of neural implant, which was developed at the University of Freiburg, Germany, has been used in previous prosthetics research, but only to stimulate the remaining nerves and restore touch sensation.
In this study, the researchers also succeeded in using the technology to read nerve signals in a precise and controlled manner.
In the next step, the researchers employed a new, AI-based technique to interpret the recorded nerve signals. The technique is based on spiking neural networks, which differ from conventional AI systems (such as those used in, for example, ChatGPT or image recognition) by processing time-based signals known as spikes, rather than continuous numerical values.
According to Elisa Donati, PhD, professor at the University of Zurich and ETH Zürich and the other senior author of the study, these signals therefore mimic more closely how biological neurons communicate.
“Our study shows that decoding peripheral nerve activity works best when it respects the language of the nervous system,” she said. “Peripheral nerves communicate through discrete electrical impulses—or spikes—and spiking neural networks are therefore naturally suited to processing this type of signal. By aligning our computational models more closely with biology, we can extract movement intent efficiently, using compact models and relatively limited data. This is an important step towards low-power, fully implantable systems for more natural control of prosthetic limbs.”
In the study, the researchers carried out tests with two participants with transfemoral amputations. Four ultrathin neural implants—each about the size of a human hair and both flexible and pliable—were inserted into the tibial branch of the sciatic nerve, which plays a central role in driving leg movement and sensation.
When the participants were asked to attempt different movements with their “phantom leg,” the researchers recorded the outgoing nerve signals and decoded them with unprecedented high resolution using their AI-based algorithm.
“This is the first study to demonstrate that signals recorded directly from peripheral nerves can be used to accurately interpret intended leg movements in amputees,” said Valle. “With this approach, we were able to map specific nerve signals to specific movements and predict, with high accuracy, which movements the participants were attempting.”
The method provides the opportunity to interpret very specific leg movements for the knees, ankles, and toes—even those that were previously impossible to decode.
“The study provides unique insight into how the nervous system transmits information. We’ve cracked the code of nerve communication and shown that it’s possible to interpret detailed leg movements, even in amputations where most of the leg is gone. It was amazing to see how electrodes placed high up in what remains of a leg could decode attempts to wiggle the toes,” Valle said.
According to the research group, another advantage is that the technology can be used for both motor control and restoring sensation with a single implant. Until now, several different implants have been required for movement and sensation.
“The system is bidirectional,” said Valle. “Once electrodes are implanted inside the nerve, they can be used to communicate bidirectionally with the nervous system. So, for the first time, a single neurotechnology can provide both natural neural control and sensory feedback in the same implantable device.”
The study is a proof of concept, demonstrating that the technique is feasible. The next step is to test it on real prostheses. While the findings are particularly significant for the development of prosthetic legs, Valle believes the method could be extended to other types of prostheses in the future.
“I believe these results could significantly influence the field. The next step is to integrate and test the technology into a prosthetic leg that can be controlled directly and that can return natural sensation,” he said.
Editor’s note: This story was adapted from materials provided by Chalmers University.
The open-access study, “Decoding phantom limb movements from intraneural recordings,” was published in Nature Communications.
