Optical neural stimulation to improve prosthetics

Brain-machine interfaces (BMIs; also known as brain-computer interfaces [BCIs]) have demonstrated the remarkable capability to connect neural signals to electromechanical devices. Now, a team of French researchers at Centre National de la Recherche Scientifique (CNRS) has demonstrated a significant enhancement of BMI performance using an optical neural interface to photostimulate neurons.

The ability of BMIs to collect and process neural signals to drive prosthetic actuators is remarkable, but it's missing an important element: somatosensory feedback. With current systems, for example, a person may direct a prosthetic arm to grasp and lift a cup, but guiding the prosthetic requires following it with the eyes. In contrast, original limbs provide sensory signals that integrate touch with other inputs and provide proprioception—that is, the self-perception that allows us to construct a mental image of the position of our bodies.  

Why not use an electrode array to provide proprioceptive information? Two reasons. First, crosstalk between motor and sensory signals means they can't be sent within the 30 ms or so of the intrinsic neural control loop. Second, electrode arrays are not precise enough to stimulate only neurons connected with a single nervous system impulse.

Optics provided a solution to both these problems.

Luc Estebanez and his colleagues at the Paris-Saclay Institute of Neuroscience (Gif-sur-Yvette, France) optically stimulated brains of mice genetically modified to contain channelrhodopsin in their excitatory neurons.

They removed a 4-mm-diameter section of skull and adhered a window to the cranial opening. Then, the team identified the "barrels" in the mouse cortex that respond to nerve impulses from a given whisker. (They did that optically as well, using red LEDs to identify the brain region stimulated by each whisker.) They directed 462 nm light through a Vialux digital light processor projector to activate neurons in one of the previously identified barrels.

After a few validation tests, Estebanez's team tested complete closed-loop BMI control. An implanted electrode sensed motor signals. In response to a sequence of motor signals, an external computer generated the appropriate set of complementary responses, which were translated into optical stimulation of the associated barrel region. To get a reward, mice were required to generate the motor signals necessary to direct feedback into a specific somatosensory region. Put another way, the mice were required to send motor signals as if they were moving a (virtual) object into a specific location.

Estebanez notes that this work enabled the first BMI able to complete full, closed-loop feedback using optogenetics. Although "applying this advance to human therapy remains a remote goal because of the need to express channelrhodopsin in the neurons," he said, the current work "sets up a platform that can be extremely useful to study the integration of sensory inputs into the motor cortex."

The team is now using the closed-loop setup to train mice to control a robotic prosthesis, using the structure of the photostimulation to characterize the somatosensory cortex, and transitioning to an all-optical system where both readout and feedback are optical. According to Estebanez, clinical investigations may begin as soon as five years from now.