A major step has been taken toward creating truly “living” prosthetics. In a study involving two patients after above-knee amputation, researchers were able for the first time to directly record and decode the signals the brain uses to try to move an amputated leg.
Multichannel electrodes were implanted in the sciatic nerve of the participants to capture electrical activity. When the patients tried to “move” a nonexistent leg — flexing the knee, foot, or toes — the system recorded the corresponding signals. It turned out that the brain continues to send detailed motor commands along the same nerve pathways, even when the muscles are already gone.
To decode the signals, the researchers used a spiking neural network, a type of artificial intelligence that works on the principle of neural spikes. Recognition accuracy reached 72%, which is higher than the performance of traditional machine-learning methods. In addition, the same electrodes may potentially make it possible to restore sensations such as touch or pressure.
Today, most leg prostheses remain passive — they do not receive signals directly from the nervous system and do not provide feedback. The new work shows that the nervous system retains something like a “movement instruction” even after amputation — and that this instruction can be read. This opens the way to prostheses that will be controlled by thought while also returning a sense of contact with the surface, making movements more natural.
