The latest in prosthetic research, direct connection to the nervous system promises a more natural leg prosthesis

A technological breakthrough that changes the lives of amputees

For patients with missing limbs, having a prosthetic limb that can "move at will" is an important hope for restoring independence and dignity in life. Recently, an international research team led by Chalmers University of Technology in Sweden has made significant progress - for the first time, they directly decoded the motion signals of the sciatic nerve of above-knee amputee patients through artificial intelligence pulse neural networks, and accurately captured subtle movement intentions including wiggling the toes. This achievement not only takes a crucial step towards achieving "natural control" of prosthetic limbs, it is also expected to simultaneously restore motor function and tactile perception through a single implanted device, bringing a new hope to life for amputees.

The "control dilemma" of existing prostheses

For a long time, helping amputees restore functional activities and independent lives has been a core goal in the biomedical field, but existing technologies have always had obvious limitations. For arm and hand prostheses, the current mainstream control method relies on residual muscles: nerve signals from the brain first activate the remaining muscles, and then sensors capture muscle movements to drive the movements of the prostheses. However, the premise of this method is that the remaining muscles must be preserved intact. Once severe amputation occurs and a large amount of muscle tissue is lost, this control mode will completely fail.

The situation of leg prostheses is even more special - due to the complexity of walking movements, most existing products adopt "passive adaptation" designs: relying on built-in mechanical structures and sensors to sense changes in the road surface and automatically adjust gait without the need for active control by the user. But this also means that users cannot flexibly adjust their movements according to their own wishes. Fine movements such as turning quickly, tiptoeing, and going up and down steps are difficult to achieve. The "practicability" and "natural feeling" of the prosthetic limb are greatly reduced.

“The core problem is that we have not been able to directly connect to the ‘command source’ of the brain.” Giacomo Valle, senior author of the study and assistant professor at Chalmers University of Technology, explained, “When a person wants to move a limb, the nerve signal is transmitted directly to the target muscle—— Even if the limb is missing, these nerves will still remain active. Theoretically, these nerves contain all the key information for movement control. Our challenge is how to ‘read’ these signals.”

The “two major difficulties” in deciphering nerve signals

It is a recognized technical problem to directly extract movement signals from the remaining nerves. There have been very few relevant studies before, and all of them focused on the upper limbs. Research on the legs is almost blank - which is in sharp contrast to clinical needs: most amputees in the world lose their legs.

The difficulties mainly focus on two aspects:First, the signal is “weak and difficult to capture”. After amputation, the electrical signal intensity of the remaining nerves is extremely low and is easily interfered by surrounding tissues. Conventional sensors cannot collect stably at all; The second is that the encoding is "complex and difficult to interpret." Nerve signals are not continuous values, but are transmitted in the form of discrete electrical pulses (also known as "spike signals"). This "pulse language" is completely different from the signal pattern of human daily cognition, and it is difficult for traditional algorithms to crack it.

To solve these problems, the research team adopted a combined solution of “hardware innovation + algorithm breakthrough”. In terms of hardware, they used an ultra-thin neural implant developed by the University of Freiburg. This implant is only as thick as a hair and has good flexibility. It can be accurately implanted into the tibial nerve branch of the sciatic nerve (this branch is directly responsible for leg movement and sensory transmission). It will not damage nerve tissue and can capture weak signals at close range. It is worth mentioning that this type of implant was previously only used to stimulate nerves and restore touch. This research has achieved "two-way communication" for the first time: it can both read nerve signals and transmit sensory feedback in reverse.

In terms of algorithms, the team abandoned the traditional artificial intelligence model and instead adopted a "Spiking Neural Network (SNN)" that is closer to the biological neural communication model. Unlike traditional AI such as ChatGPT and image recognition that rely on continuous data, the spiking neural network specializes in processing spike signals based on time series, and its working principle is highly consistent with the communication method between biological neurons. "The 'language' of the nervous system itself is impulse, and using impulse neural networks to interpret it is equivalent to talking to the nerves in the 'native language'." Elisa Donati, another senior author and professor at the University of Zurich, explained that this biologically-friendly design allows the model to efficiently extract movement intentions with less data and lower power consumption, laying the foundation for the subsequent development of a "fully implantable, low-power" prosthetic control system.

Experimental verification: accurate decoding of "phantom limb movements"

In clinical testing, the research team conducted experiments on two above-knee amputee patients. They implanted four ultra-thin implants into the tibial nerve branch of the patient's sciatic nerve. When the patient was asked to try different actions with the "phantom limb" (that is, the missing limb that still exists in subjective perception), the implants recorded nerve impulse signals in real time and then decoded them through the impulse neural network.

The test results are exciting: the system not only successfully recognized regular movements such as knee bending and ankle rotation, but also accurately captured previously uninterpretable subtle movements such as wiggling toes and tiptoeing, with decoding accuracy reaching an unprecedented level. “The most surprising thing is that even if the electrode is implanted at a higher position on the residual limb, it can still clearly decode the tiny movements of the toes.” Valle emphasized that this proves that the movement information contained in the neural signals is far richer than imagined. As long as the decoding method is appropriate, ultra-high-resolution movement control can be achieved.

More importantly, the system achieves “two-way functional integration”. Previously, if you want a prosthetic limb to have both motion control and tactile feedback functions, you need to implant multiple different chips, which not only increases the complexity of the surgery, but may also increase the risk of complications. The single implanted device used in this study can not only read movement signals, but also transmit tactile feedback through neural stimulation - for example, allowing users to sense the pressure and temperature of objects touched by , truly achieving the dual recovery of "movement + feeling".

Future prospects: from “proof of concept” to “practical prosthetics”

Currently, this research is still in the “proof of concept” stage, proving the feasibility of the technology. In the next step, the research team will focus on promoting the practical application of the technology: integrating neural implants with real prostheses, conducting clinical tests, and optimizing the stability and practicality of the system.

“Our goal is not to create a ‘movable’ prosthesis, but to create a prosthesis ‘like my own limbs’.” Valle said that in the future, this technology can not only be used for leg prostheses, but can also be extended to various types of prosthetic products such as upper limbs and hands. Imagine that after an amputee wears a prosthetic limb, he can flexibly control the movements of each finger through his mind and clearly feel the touch of the object he is holding. When walking, he can adjust his gait according to the road conditions, and he can go up and down stairs, run, and turn just like ordinary people. All of this will be gradually realized with the maturity of neural decoding technology.

In addition to the field of prosthetics, this technology may also bring about a wider range of application scenarios: such as developing neural interface devices for spinal cord injury patients to help them regain control of their limbs; or being used in rehabilitation robots to provide more precise rehabilitation training for stroke patients. Its core "pulse neural network + neural implant" technology also provides new ideas for the integration of biomedicine and artificial intelligence - when we learn to use "biological language" to talk to the human body system, more "medical problems" that have been in the past may be solved.