A new technology enables prosthetics to make more fluid and natural movements, without requiring a lengthy training process. To make this possible, the technology intercepts muscle signals from patients, decodes these signals and translates the data into commands to control prosthetic arms and hands.
Prosthetics are getting smarter. For instance, modern prosthetic technologies like machine learning are deploying machine learning to recognise patterns in users' movements and control the prosthesis based on these patterns. In practice, this means that users need to train their prosthesis to teach it to recognise and correctly translate specific patterns in their muscle signals into commands such as opening or closing a hand.
Long-term training process
"Control via pattern recognition requires patients to go through a lengthy process of training their prosthesis," explains He Huang, professor of joint biomedical engineering at North Carolina State University and the University of North Carolina Chapel Hill. "This process can be both monotonous and time-consuming."
Researchers at North Carolina State University and the University of North Carolina Chapel Hill have developed a solution to this problem. A new technology makes it possible to significantly speed up the learning process of prosthetics. "Every time you change your posture, the muscle signals for generating the same hand or wrist movement also change. Relying solely on machine learning means that the device has to perform the same actions multiple times; for each posture again, once for situations when you are sweaty, once for when you are not, et cetera. Our approach largely circumvents this."
For this approach, the researchers developed a muscoskeletal model based on data collected with electromyographic sensors. These sensors were placed on the forearms of six volunteers without amputations to accurately image which muscle signals are sent when they perform a variety of movements with their wrists and hands. This data was then used to develop a generic model capable of converting these muscle signals into commands to control a prosthesis. "We wanted to focus on what we already know about the human body," Huang said. "This is not only more intuitive for users, but also more reliable and practical."
Intercepting muscle signals
"If someone loses a hand, their brain is still programmed as if the hand is present," Huang said. "So if someone wants to pick up a glass of water, the brain still sends the same signals to the forearm for this purpose. We use sensors to intercept these signals and transmit this data to a computer, which feeds this data to a virtual muscoskeletal model. This model observes the position of the muscles, joints and bones and calculates the movements that would take place if the hand and wrist were still fully intact. The results of these calculations are passed on to the prosthetic wrist or hand, which performs the relevant movements in real-time in a coordinated manner. This enables prosthetics to move in a more fluid and natural way."
The model was put to the test by the researchers during a series of experiments. During these tests, both patients with and without amputations were found to be able to use the model to perform all the required hand and wrist movements without extensive training. The researchers point out that their model has not yet gone through clinical trials. They therefore expect that the model will not be commercially available for several years.
In time, the researchers expect that the model can also be used for other purposes. "This can also be used to develop computer interface devices for healthy people," Huang said. "Think of devices for gaming or for manipulating objects in CAD programmes."
Research paper
More information and all the researchers' findings can be found in the researchers' paper 'Myoelectric Control Based on A Generic Musculoskeletal Model: Towards A Multi-User Neural-Machine Interface', published in EEE Transactions on Neural Systems and Rehabilitation Engineering.
Author: Wouter Hoeffnagel
Source: North Carolina State University (pictured)
Source: University of North Carolina Chapel Hill
Source: Research paper 'Myoelectric Control Based on A Generic Musculoskeletal Model: Towards A Multi-User Neural-Machine Interface'
