According to the latest issue of Nature Neuroscience, a new study confirms that functional ultrasound (FUS) technology developed by researchers at the California Institute of Technology can become the basis of an "online" brain-computer interface (BMI) that can read brain activity and decipher its meaning through a decoder programmed with machine learning, thereby controlling a computer with extremely short latency and accurate prediction of movement.
Anatomical recording plane and behavioral tasks. Image source: Physicist Organization Network
In 2021, Caltech researchers developed a way to read brain activity using functional ultrasound, a much less invasive technique.
Ultrasound imaging works by emitting high-frequency sound pulses and then measuring the echoes of these sound vibrations in materials, such as various tissues in the human body. Sound waves travel at different speeds in these tissue types and are reflected at the boundaries between them. This technology is commonly used to take images of a fetus in the womb and for other diagnostic imaging.
Because the skull is impermeable to sound waves, brain imaging using ultrasound requires a transparent "window" in the skull. The ultrasound technology does not require implantation in the brain itself, which greatly reduces the chance of infection and leaves the brain tissue and its protective dura mater intact.
Changes in neuronal activity can cause changes in their utilization of metabolic resources such as oxygen. These resources are replenished through the blood, which is key to functional ultrasound. In this study, researchers used ultrasound to measure changes in blood flow to specific brain regions. Just like an ambulance siren changes pitch with distance, red blood cells raise the pitch of the reflected ultrasound waves as they approach the source of the sound, and lower the pitch as they move away from the source.
By measuring this Doppler effect, researchers can record tiny changes in blood flow in the brain in a spatial region just 100 microns across, about the width of a hair. They were able to simultaneously measure the activity of tiny populations of nerve cells widely distributed throughout the brain, some as small as 60 neurons.
The researchers used functional ultrasound to measure brain activity in the posterior parietal cortex (PPC) of nonhuman primates, an area responsible for planning and helping to execute movements. The animals were taught two tasks: moving their hands to guide a cursor on the screen and moving their eyes to look at specific parts of the screen. They only have to think about performing the task, rather than actually moving their eyes or hands, because the BMI can read their brain activity.
Ultrasonic data is sent to the decoder in real time, which then generates control signals to move the cursor where desired. BMI was able to successfully do this for 8 radial targets with a small average error.