Intelligent neuroprostheses mimic natural motor control
31 March 2015
Neuroscientists are taking inspiration from natural motor control to design new prosthetic devices that can better replace limb function.
In new work, researchers have tested a range of brain-controlled devices – from wheelchairs to robots to advanced limbs – that work with their users to perform tasks intelligently.
These neuroprosthetic devices decode brain signals to determine the actions their users want to take, and then use advanced robotics to do the work of the spinal cord in orchestrating the movements. The use of shared control – new to neuroprostheses empowers users to perform complex tasks, according to José del R. Millán, who presented the new work at the Cognitive Neuroscience Society (CNS) conference in San Francisco on Monday (March 30).
Millán has been working on 'brain-computer interfaces' (BCIs), designing devices that use people’s own brain activity to restore hand grasping and locomotion, or provide mobility via wheelchairs or telepresence robots, using people’s own brain activity.
“The prostheses and robots that our BCIs control are intelligent, as they can interpret many low-level details that are not necessarily coded in the mental commands,” he says. Importantly, they also work autonomously if the users do not want to change their behaviour. This function mirrors how our deep brain areas, spinal cord, and musculoskeletal system work together in many routine tasks, allowing our bodies to do simple tasks while we focus our attention elsewhere.
In his and colleagues’ latest work, they tested a variety of brain-controlled devices on people with motor disabilities, in some case quite severe. The participants successfully completed tasks ranging from writing to navigation at similar levels of performance as healthy control groups.
The individuals operated the devices by voluntarily and spontaneously modulating the electrical brain activity (EEGs) to deliver commands. EEGs have the benefit that they can be recorded non-invasively through probes on the scalp, rather than requiring surgery or sophisticated machinery. “It also provides a global picture of our brain patterns, what is necessary to decode all the variety of neural correlates we want to exploit,” Millán explains.
The participants needed a relatively short training period of no more than nine sessions before being able to operate the devices. And those using telepresence robots were able to successfully navigate through environments they had never visited. Key to their success was the concept of shared control – using robots’ sensory capabilities to interpret the users’ command in context.
The BCI processes users’ intentions and decision-making mainly from the cerebral cortex . But, Millán notes that many elements of skilled movements are handled in the brainstem and spinal cord. By designing the intelligent device to control the lower-level movements in concert with the higher-level brain activity from the BCI, the neuroprostheses come closer to natural motor control. “We aim to interact with these neuroprostheses as if they were our new body, using the very same neural signals and principles that control our muscles,” Millán says.
Two of the biggest challenges for neuroprosthetics are finding new physical interfaces in addition to EEG that can operate permanently and over long periods of time, as well as providing rich sensory feedback. “This sensory information will make users feel the neuroprosthesis and the environment, what is essential to promote user’s agency and ownership of the prosthesis,” Millán says.
“The third major challenge is the one at the core of cognitive neuroscience: We must decode and integrate in the prosthetic control loop information about perceptual cognitive processes of the user that are crucial for volitional interaction,” he says. These processes include awareness to errors made by the device, anticipation of critical decision points, and lapses of attention.
“Future neuroprostheses — robots and exoskeletons controlled via a BCI — will be tightly coupled with the user in such a way that the resulting system can replace and restore impaired limb functions because it will be controlled by the same neural signals as their natural counterparts.”