Paralyzed patient gains ability to control robotic arm with mind

Scientists from the University of California, San Francisco (UCSF) have created a platform that can restore the ability of paralyzed people to act independently using robotic arms. The manipulator can be controlled by the power of thought, imagining the desired action. Artificial intelligence plays a key role in this process.

Image source: UCSF

The platform developed by the scientists is based on electrocorticography (ECoG), a system for identifying active areas of the cerebral cortex. It is usually used before brain surgeries so that the surgeon does not affect critical areas. The ECoG system and implants are quite expensive, but in this case they have demonstrated high efficiency.

A patient with complete paralysis of all limbs was connected via ECoG to a computer with AI, which in turn transmitted commands to a robotic arm. An algorithm based on machine learning recognized the activity of the brain areas responsible for hand movements and transmitted the corresponding commands to the manipulator. The patient observed the process and mentally adjusted the hand movements in real time, which was immediately transmitted to the robotic arm.

The researchers note that this method of training the algorithm with instant error correction based on the patient’s wishes has shown high efficiency. Moreover, the AI ​​has learned to track the shift in activity of the brain areas responsible for movement to neighboring areas of the cortex.

Overall, the activity pattern did not change, but the activity zone “drifted” across the cortex, the scientists noted. Normally, this would require recalibrating the implants every few days. However, the AI-based platform developed was able to independently account for this shift and remained operational for more than seven months without significant intervention in the settings. This alone makes the project promising.

With the help of the robotic arm, the patient was able to pick up and put down objects, use dishes, and fill a glass with liquid from a dispenser. Scientists also believe that they can improve the platform, for example, by equipping the robotic arm with machine vision. This will allow the machine to help the patient perform actions more accurately, and will also improve the AI ​​learning process.

«“I’m completely confident that we now know how to create the system and we can make it work,” shared one of the participants in the study, the results of which were published in the journal Cell.