A study published in the journal Advanced Intelligent Systems has garnered attention in the fields of high-end device manufacturing and neuroscience: the brain can adapt to an artificial third arm and use it to perform simple tasks. This brings new hope to the dream of precision mechanics, surgeons, and others who envision people proficiently using a "third arm" in the future.
In this study, approximately 20 participants learned to use a prosthetic limb in a laboratory setting. This unfinished prosthetic, equipped with a gripper at its end, was fixed to a table beside the participant. While seated, participants controlled the prosthetic via a belt placed on their diaphragm, with the limb moving forward during exhalation and backward during inhalation. Participants practiced performing a series of tasks, such as grasping blocks, pressing buttons, and moving a cursor, with scientists aiming to determine the extent to which the brain can learn to use a mechanical limb to the proficiency of a natural arm.
Previously, the EPFL team had demonstrated that participants could control a virtual arm and use a simple mechanical arm to point at objects. This study went further, testing gripping capabilities. The team, led by postdoctoral researcher David Leal, focused on measuring the complex skill of task generalization. Lead author Silvestro Micera explained that natural limbs can automatically perform generalized tasks, as the brain internalizes the principles of movement sequences and applies them to other objects.
The study results showed that generalization does occur. Participants first practiced rapidly moving blocks using both their natural arm and the mechanical arm simultaneously. In the second phase, compared to those without practice, they could manipulate various other objects faster and more accurately using both their natural arm and the mechanical arm, indicating that effective protocols for inducing generalization in natural limbs are equally applicable to robotic limbs.
However, when the operations in the testing phase differed significantly from the training phase, generalization ability declined, particularly in multi-task environments. For example, participants struggled to generalize the ability to grasp objects with the prosthetic while simultaneously typing on a keyboard. Micera believes this suggests that generalization with prosthetics may be more challenging, possibly limited to performing similar tasks, and that the training methods may not yet be optimal.
Currently, few scientists are researching the use of mechanical limbs to enhance human function, with only a handful of teams in the U.S. and Europe, including those studying integrated artificial fingers, engaged in this field. However, the approach holds immense potential. Micera noted that many professions, such as first responders, precision mechanics, and surgeons, could benefit from an additional limb, though such applications remain far from reality.
The primary obstacle lies in insufficient control precision. Even with improvements, diaphragm-controlled artificial arms remain in their early stages, far less precise than natural limbs. To overcome this, invasive interfaces like cortical electrodes may become a long-term solution for translating brain signals into executable commands for the arm, though this is currently difficult to achieve. Therefore, Micera's team is currently using only non-invasive devices, controlling the prosthetic through breathing for now, with plans to soon use electrodes placed on the scalp.
For Micera, the appeal of this work lies not in futuristic scenarios of human enhancement but in better understanding the brain and how it interacts with the body to form new connections. He stated that by improving and accelerating prosthetic training, insights could be gained that benefit rehabilitation therapies, such as for stroke patients with paralysis.
