For the past decade, bridging the gap between our thoughts and movements has been the challenge that Paul Cederna, Plastic Surgeon and Biomechanics Professor; Cynthia Chestek, Neural Engineer and their research team at the University of Michigan, Ann Arbor, are trying to overcome as they develop mind-controlled prosthetics.

The road to mind-controlled prosthetics

Theoretically, lives of many amputees will improve tremendously if they are able to manipulate their artificial limbs like the way they do for their natural ones and regain their capabilities to carry out many of the daily tasks. However, this is not entirely achievable at the moment because one’s brain signals are not robust enough to sustain these additional bionic structures that are attached upon the body.

Robotics prostheses are available and what they do is to retrieve signals from the surface of the amputee’s skin so that one can have some control over their fingers by contracting muscles remained around the area of an amputation. Yet, if those muscles are to be absent, amputees may have to depend on other less intuitive movements, such as replying on their upper arms to get their fingers moving.

Others may choose to have wires being planted directly into the muscles around the affected areas for more robust electrical signal capture. Nevertheless, robotic limbs tend to be heavy and the signals may vary especially when users perspire or when the socket of the prosthesis is out of place as a result of constant movement. This means that the devices need to be recalibrated many times throughout the day and most users find it too much of a consequence to wear.

To overcome this signaling challenge, some researchers had tried rerouting amputee’s nerves from their limbs to a chest muscle so that a strong electrical signals can be recorded but that will require surgeons to extract out some healthy and existing nerves in the chest area to accommodate the rerouting. This may compromise the amputees other movements.

A machine learning driven implant

What Cederna, Chestek and their research team did was to separate tiny muscles from their respective major nerves around the amputees’ thigh area and bundle them up together into a graft that takes the size of a paper clip. This will permit the creation of a new set of nerves and blood vessels to be placed in an amputee’s forearm or bicep. After which, electrodes are placed into these sites so that nerve signals can be recorded and transmitted into the prosthetic arm in real time.

Machine learning algorithms will then be used to interpret when and which of these tiny muscles are contracting, how much contractions are there, and associate each of them to various intended movements, so that all the captured signals can be converted seamlessly into the intended movements. As of now, this muscle graft had been tested on three human participants with amputations at different points along their arms. Two of the amputees have opted for long-term implant so that the machine learning algorithms can further learn their nerve signals and muscle movements.

Researchers noted that since amputees rely on natural nerve signal, they do not need to put aside additional time to learn how to make movements, they can do so right away. Thus far, all participants were able to perform small precision tasks such as picking up small blocks and stacking them up and they maintained a certain level of control for up to 300 days. In spite so, researchers noted that the set up remains a lab design and the device needs to be optimized for regulatory approval and commercial testing. All related findings were recently published in Science Translational Medicine.


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.