A trio of quadriplegic patients learned to control their wheelchairs with their thoughts, which allowed them to roll a relatively complex route, avoiding obstacles, in the corridors of the German hospital where they were being followed.
Research describing the feat has just been published by American and European specialists in the specialized journal iScience. This is an important milestone in studies on the so-called brain-machine interface, whose great promise is to return part of the mobility to paralyzed people through robotics and neuroscience. The work, however, also shows how complex this path is: not all participants were able to make their brain “talk” to the computer with the same efficiency.
“We showed that mutual learning, involving both the user and the algorithm of the brain-machine interface, are important to successfully operate these wheelchairs”, summarizes the study coordinator, José Millán, from the University of Texas at Austin (USA). ), in an official statement. Scientists from the University of the Ruhr in Bochum, Germany, the University of Padua, Italy, and other European institutions also sign the research.
Since the beginning of the 21st century, works that try to connect the brain with robotic systems have followed different paths. Some have bet on surgical implants, directly connected to neurons, because that way it would be possible to obtain more reliable signals of brain activity.
This more invasive approach, however, involves both the risks of the operation and the long-term impacts of electronic devices on patients’ brains. In addition, there is the question of determining how many neurons, among the billions in the organ, are needed to decode information about movements.
In the case of the research in the journal iScience, Millán and his colleagues took a non-invasive approach. The quadriplegic patients wore an external EEG (electroencephalography) cap, which is capable of picking up the electrical activity of the brain from the outside. One of the participants was completely paralyzed from the neck down, while the other two were able to move their arms and fingers a little.
At first glance, the volunteers’ task was simple: imagine that they were moving both hands at the same time —which the interface would “read” as the command to turn left— or both feet, the equivalent of turning right.
The system, which ran on a laptop behind the wheelchair, was equipped with the ability to optimize its “mind reading” over time, so that the task of controlling the wheelchair’s turns became easier. . If neither of the two commands to turn were given mentally, the vehicle “understood” that it should go in a straight line.
Although the system seems simple, it took a lot of training —three sessions a week, over a period of between 2 months and 5 months— for the patients to acquire good control skills. Two of them —designated as P1 and P3 in the study— achieved a degree of accuracy in the use of commands equal to or greater than 95%, progressing more and more throughout the sessions. The other, which the scientists call P2, however, peaked at 68% accuracy, with no clear progress after that.
In the two patients who made the most progress, the team realized that two factors worked together to make this happen. On the one hand, the brain-machine interface software, in fact, became increasingly adept in its ability to “understand” what the patients’ brain waves meant. On the other hand, however, everything indicates that the volunteers themselves learned, unconsciously, to send clearer commands via thought.
“Based on the EEG results, we see that the patient has consolidated the ability to modulate different parts of their brain to generate different ‘go left’ and ‘go right’ patterns. We believe that a reorganization of their cerebral cortex has taken place. , as a result of the learning process”, explains Millán.
In the final task, which involved traveling along hospital corridors and rooms, dodging cones, dividers and beds, again two patients successfully completed the journey, while the other did not go all the way. The team’s intention is to further investigate what factors hinder the ability to control the interface with the brain in these cases.