One of the biggest challenges for amputees is that they need a way to control their prosthesis, says Giacomo Valle, assistant professor at Chalmers.
Prosthetics for people who have had an arm or hand amputated are often controlled by muscles, but this requires that the relevant muscles remain intact. For leg amputees, it is more common to have prostheses without any active control from the user.
Giacomo Valle and his colleagues hope to change that.
All the information needed to control our body parts remains in the nerves, even if the body part is no longer there. But those signals need to be decoded, says Giacomo Valle.
AI inspired by biology
In the study, the researchers placed four electronic neural implants in the test subjects' sciatic nerves. The implants are as thin as a hair and flexible.
A new AI method was then used to interpret the nerve signals recorded by the implant. The method is based on so-called spiking neural networks, or SNNs, which means that information is transmitted through brief spikes rather than continuous signals.
The technology is different from the usual AI systems we are used to, such as ChatGPT or image-recognition systems, says Giacomo Valle.
These spikes are similar to the way our nervous system communicates. It is inspired by our biology, which makes it possible to understand how the brain communicates.
“Cracked the code”
Using the method, the researchers were able to interpret very detailed intended movements in the test subjects - even movements as small as the desire to wiggle their toes.
We have cracked the code, says Giacomo Valle.
But the technology cannot yet be used in real life. The study is a so-called “proof of concept,” a study that tests a technology and shows that it is feasible, and it was carried out on only two test subjects.
The next step is to integrate the technology with prosthetics, use it in a more realistic everyday scenario and ensure that it is truly functional and something that patients can benefit from, says Giacomo Valle.
The study has been published in Nature Communications.
