Not for the first time and not a new method
A new method of training a neurocomputer prosthesis has allowed to get rid of intracranial electrodes
Now, to transmit signals using the brain-computer interface, only those electrodes that are placed on the surface of the subject's head are enough
Evgenia Shcherbina, "The Attic"
American engineers have improved the methodology of training algorithms working with the brain-computer interface. Now, in order to use it, you do not need to implant electrodes directly into the brain. The new development will allow us to use such interfaces on a much larger scale than now.
With the help of "brain-computer" interfaces, a person who has lost an arm or leg can learn to control a prosthesis with the "power of thought" – using signals from the motor cortex of the brain. Simply put, the interface simulates the mechanisms by which our brain controls the limbs. As a rule, it consists of electrodes that read brain signals, a computer to which these signals are sent for processing, and a device to which the processed signals are sent, that is, a prosthesis. Until recently, for the operation of neurocomputer interfaces, it was necessary to insert electrodes directly into the brain, and this, obviously, can be dangerous. Therefore, such mechanisms were used in very rare cases.
The main reason why the electrical signal cannot be read from the surface of the skull is that in this case there is too much interference in the signal, so it is difficult to convert it into the correct movements of the prosthesis. To take this disadvantage into account, the authors of the new work (Edelman et al., Noninvasive neuroimaging enhancements continuous neural tracking for robotic device control) were looking for such methods that would increase the efficiency of the interface even with a "dirty" signal.
To do this, they modified the technique by which researchers usually train algorithms that process the signal coming from the brain. Such techniques, as a rule, are trained on signals from the brain of people who make short jerky movements. To do this, the participants of the training experiment look at a special point in the center of the monitor and mentally take the missing limb from this point to some other place on the screen, and then return it back. Experts call this technique "from center to center".
The resulting short jerky movements are more convenient for the algorithm to process, but in this case there is a minus, because in real life people, as a rule, do not make such precise and sharp movements. There are more unpredictable deviations in the signals of the brain, which controls smoother gestures. The result is a big difference between reality and the result of learning algorithms.
The authors of the article decided to train the program differently. They attracted 33 healthy people, each of whom participated in 10 tests. A person had to "use his mind", that is, through a neurointerface, control a robotic arm, trying to make it continuously follow the random movements of the target on the monitor (this is called a pursuit task). Compared to the previous one, the new technique turned out to be more effective. The algorithm trained in this way coped with traditional tasks "from center to center" by 60% better. When such algorithms were offered other options for pursuit tasks, their effectiveness increased by 500%.
Various tests with groups of volunteers (there were 68 in total, including those who participated in the initial training) showed that this technique increases both the accuracy during control of the prosthesis and the efficiency of the algorithm in signal processing. At the same time, it was not necessary to insert electrodes into the brain – it was enough that electrical signals were read from the surface of the head.
Scientists hope that further development of training of such non-invasive neural interfaces will help their wider introduction into medical practice and make it more accessible to patients.
* * *
From the editorial office:
In fact, non-invasive brain–computer neural interfaces are news from at least a decade ago. And the test results of the encephalograph cap described above were published by a group led by the same Bin He a year and a half ago, only the design of the experiment was completely different, as shown in the video.
But what kind of press release is complete without the word "for the first time"? The message about the experiment described above is called: First-ever successful mind-controlled robotic arm without brain implants :(
Portal "Eternal youth" http://vechnayamolodost.ru