[ad_1]
In an article recently published in the Journal of Neural Engineering, researchers from the University of Twente (NL), in collaboration with Imperial College London (UK) and the University of New York. University of Aalborg (DK), presented a totally new approach. This new approach ensures that the control of a prosthesis is more natural for users. This technology is based on musculoskeletal models and offers an alternative to the usual machine learning.
Prosthetic Arm
There are millions of people in the world who can not use part of their body as a result of amputation. Advanced robotic solutions allow immediate recovery of motor skills in these situations. Despite advances in many areas, current prosthetics offer limited functionality to their users.
Human movements result from the fact that electrical impulses are sent from the brain to the muscles. A substitute for these cerebral signals can be recorded, in the form of electromyograms (EMGs), using electrodes placed on the skin.
Available Solutions
In people with an amputated arm, EMG signals can be measured from the remaining muscles after amputation. These signals make it possible to control the prosthesis in real time. The established methodologies depend on machine learning, algorithms detecting patterns of EMG signals characteristic of a particular movement. These algorithms teach the connection between certain EMG signals and a specific movement. Once an EMG pattern is detected, it can be associated with a prosthetic movement.
Despite widespread use, machine learning algorithms are sensitive to changes in EMG properties due to noise, electrode placement, and muscle fatigue. To control the use of the prosthesis, the user must learn to systematically produce very different EMG models. which is not always possible.
This type of technology therefore often offers limited reliability and is therefore less used by amputees with one arm.
Alternative Solution
In a recent paper (M. Sartori and University of Twente) presented to an international team of scientists an alternative solution for machine learning
. The idea was to create a detailed digital model of the amputated arm (ghost arm), including the organic tissues. Such a digital model included the accurate description of lost muscles, tendons and joints. The researchers recorded EMG signals from the forearm remaining after amputation. These signals were then used to determine how the virtual muscles of the model would be activated and what movement would be produced by the virtual ghost. This predicted motion was then transmitted to the robot's prosthesis in real time. This ensured that the prosthetic arm could be used as a natural extension of one's own body. the use of machine learning technology. Instead, the person simply had to imagine moving his own ghost arm after which such a movement was accurately recorded and executed by the prosthesis.
This performance may have significant clinical benefits because, in the user's daily life, it helps to reinforce the sense of ownership and acceptance of advanced robot prostheses as a replacement part of the body . performed on three test subjects (intact) and one subject transradically amputated, which showed the ability to perform a repertoire of large movements that poses a challenge for the most modern methods. After these first experiments, the authors plan to conduct a clinical trial on a large number of patients.
The complete study is published in Journal of Neural Engineering:
- Massimo Sartori, William Durandau, Strahinja Dosen, Dario Farina. "Robust simultaneous myoelectric control or multiple degrees of freedom in hand-wrist prostheses by real-time neuromusculoskeletal modeling". Journal of Neural Engineering . 2018.
- Web: http://iopscience.iop.org/article/10.1088/1741-2552/aae26b/meta
Source: University of Twente
Source link
