AI and ML are enabling prosthetic hands that better match each user’s muscle control patterns for an unlimited number of grips and gestures...
More than 1,500 Americans have lost a leg or arm in combat in Iraq or Afghanistan, and hundreds have suffered the amputation of multiple limbs. The physical damage is often compounded by mental stress, depression, anxiety and post-traumatic stress disease (PTSD). The value of any scientific development that can give back physical capability and independence to those who have lost it is beyond words. This brings us to AI-powered prosthetic hands, which “learn” from repetitive actions and enable hand amputees to make an unlimited number of grips and gestures.
Artificial intelligence (AI) takes these prosthetic hands to new levels when compared with the traditional prosthetics that rely on electromyography (EMG) alone. EMG measures of muscular electrical activity in response to a nerve’s stimulation in order to move the hand.
The new prosthetics also rely on EMG, but the addition of AI leads to significant improvements in how they perform.
“The biggest difference between the AI prosthetic hand and the standard prosthetic hand is the AI algorithms that facilitate users in their daily usage,” said Qing Zhu, product manager and regulatory analyst at BrainRobotics, which innovated the AI prosthetic hand. “Users are able to have a shorter training time while the captured EMG signals are analyzed into a more precise control of the prosthetic hand,” he added.
Training time is required as the user learns to use the hand and understands the interactions and relationships between what he wills the prosthetic to do, the electronic impulses from muscles that result, and the ability of these impulses to operate a prosthetic hand.
Here’s how the technology works:
Artificial intelligence is embedded in the prosthetic hand and is also running on cloud servers that we operate,” explained Zhu. “The user’s EMG signals are processed in the hand and transferred to our cloud servers. Initially, there is a hand “signal training” database in the cloud servers. It creates a hand control model that fits the usage patterns of the user and his muscle control patterns. The prosthetic hand is then embedded with this control model once the initial signal training is complete.
Zhu also noted that “if an update is needed, the signal training session can be redone, and the new model can be replaced in the hand remotely.”
Over time, machine learning (ML) in the hand “learns” from the repetitive use patterns of the user and refines the intelligence that is built into the hand, so hand control more closely conforms to the user’s usage habits and needs.
“We use AI algorithms and machine learning to capture the nuances and the movement similarities to allow for the best control of the hand,” said Zhu.
On the physical side of the prosthetic hand design, a HIFI socket (high-fidelity interface with osseostabilizing technology), developed by Biodesigns, Inc., is used to facilitate tighter alignments between the prosthetic hand and bone joints.
“This is a complete departure from standard-of-care sockets,” said Zhu. “The HiFi Interface focuses on manipulating soft tissue via an alternating, circumferential array of targeted compression to capture the wearer’s underlying bone. This enables better control and synchronization between the prosthetic hand and the body’s natural skeletal motion.”
Randall Alley, CEO, founder, and chief prosthetist and head of user-interface technologies for Biodesigns which designs and deploys human interface technologies, said “the result is improved hand function, range of motion, and stability.”
All of this matters to U.S. Army Special Forces Captain Carey Duval, an amputee who got the AI-designed prosthetic hand.
“Carey’s initial challenge was that the speed of the hand didn’t match his daily usage needs,” said Zhu. “We were able to come out with and remotely deploy a firmware update that allowed Carey to speed up the opening and closing of the hand. He was able to directly update the firmware on his mobile prosthetic hand app, which is not possible with other prosthetic hands on the market.
“For our development team, this was an excellent reminder that each user will have their own unique needs,” Zhu said. “We will be adding more features to the app so individual users can customize the hand based on their own needs and habits.”
— Mary E. Shacklett, president, Transworld Data