Three patients with quadriplegia used a brain-computer interface (BCI) to control a tablet device just by thinking about moving and clicking a cursor, a new study from the BrainGate Consortium reports. The researchers developed the system for use with an unmodified commercial tablet rather than with custom-built devices commonly used in BCI research. The hope is that one day, such systems could build a bridge between conventional technology and people who are unable to use the current options.
The study involved two patients with paralysis from amyotrophic lateral sclerosis (ALS) and one patient with a spinal cord injury. Each patient received one or two “baby aspirin-sized" implants in the motor cortex of their brain—each implant was an array of about 100 silicon microelectrodes measuring 4 by 4 millimeters. Signals from this implant were sent to a Bluetooth interface and configured to work like a wireless mouse.
“Each electrode penetrates about 1.5 millimeters into the brain, into just the outer layers where the brain cells that control movement on opposite side of body live,” said Jaimie Henderson, M.D., a neurosurgeon at Stanford University Medical Center and a senior author on the paper. “Those probes end up very close to those neurons and essentially eavesdrop on those brain cells and allow us to listen to the firing patterns of those neurons.”
By “eavesdropping” on neurons’ firing patterns, the researchers could figure out how changes in different neurons’ firing rates correspond to different movements. They used this knowledge to build decoders to translate these signals into cursor movements on a screen, Henderson said.
The patients used the BCI to operate an off-the-shelf Google Nexus 9 tablet. They were asked to complete a series of tasks—such as checking email and composing a message, performing a Google search, playing YouTube videos and streaming music—but also tried using apps of their choosing. Two patients chatted to each other using an instant messaging program and one of them played music on a digital piano app.
The patients used different visualizations to get the cursor to move, Henderson said. One patient envisioned rolling a pool ball on a table, forward and backward, left and right, while another patient used a thumb and forefinger visualization, he said. But as they got used to the system, they stopped using intervening visualizations and felt as though they were controlling the cursor directly.
The patients were able to make up to 22 point-and-click selections per minute and type up to 30 effective characters per minute, the study showed. While this may not seem particularly fast, it would be a game-changer for someone who is not able to communicate at all.
“This comes in at about six words per minute or so, and that’s maybe a third of the speed that a facile user could type on a phone app, for example,” Henderson said. “So, it’s still not nearly the speed that an able-bodied person could use these interfaces, but it’s also often better than what could be otherwise used and for people with communication difficulties.”
"The assistive technologies that are available today, while they're important and useful, are all inherently limited in terms of either the speed of use they enable, or the flexibility of the interface," said Krishna Shenoy, a senior author of the paper and an electrical engineer and neuroscientist at Stanford University and Howard Hughes Medical Institute, in a statement. "That's largely because of the limited input signals that are available. With the richness of the input from the BCI, we were able to just buy two tablets on Amazon, turn on Bluetooth and the participants could use them with our investigational BrainGate system right out of the box.”
The most important part of the study, the researchers said, was that it used an unmodified device with any special accessibility software switched off.
“All [BCI research], as far as I know, has been based on building very specific software that works in a specialized way with the computer that is doing the BCI decoding,” Henderson said. "What we’re trying to do here is to decouple that and say we can decode these brain signals to get cursor movements and once we’ve done that, we can use any device that uses cursor movements. It’s a much more general-purpose way to do things.”
Rather than build completely new devices, this approach would allow neuroscientists and researchers to piggyback on the interface design and software expertise of others. This frees them to work on improving BCI technology. The BrainGate team highlighted several challenges it plans to tackle, such as adding click-and-drag or multitouch capabilities—the latter meaning the use of multiple fingers on a tablet screen, for example, pinching to zoom into a picture.
Ultimately, Henderson hopes for a BCI system that could work with any computer: “What I would love to see would be a fully wireless [implanted] device that fed to a USB key that you could plug into any laptop, for example, and run that computer with your brain signals,” he said.