Science to restore speech is moving fast.
A Science News podcast episode by reporter Laura Sanders surveys brain-implant research aimed at restoring speech, emotional tone and gestures to people with paralysis, including a new wireless device implanted this summer.
The Deep End: Talk to Me, a Science News podcast hosted by Laura Sanders, surveyed current efforts to give words, emotions and gestures back to people who cannot move or talk. The episode says AI-based methods for decoding speech signals and predicting words, along with advances in technology, are driving the progress.
Casey Harrell, who has amyotrophic lateral sclerosis (ALS) and can no longer speak, volunteered for the BrainGate clinical trial. His implant detects neural patterns when he tries to say a word. Leigh Hochberg, director of the BrainGate consortium, estimated that fewer than 10 people, possibly five or six, have implants that decode attempted speech, and he expected that number to grow quickly. He said a dozen or more companies are now pursuing implantable brain-computer interfaces.
One of them, Paradromics, makes a dime-sized device with 421 sensors that sits on the motor cortex. Unlike Harrell's system, which uses ports on his head connected to external wires, it is fully implantable and wireless. Neurosurgeon Matthew Willsey of the University of Michigan Health implanted it this summer in a 68-year-old Michigan woman with primary lateral sclerosis. The system is running, and she has used it to talk to her grandkids on the phone. Sanders describes the project as only at its very beginning.
Speed is the next target. Harrell's system puts decoded words on a screen, and he then chooses to have a digital voice say them. Researchers want to skip that step and go straight from brain signals to speech. In the latest tests, brain-to-voice output reached about 95 percent accuracy for listeners, according to a preprint on bioRxiv. Harrell called the goal "the holy grail of what this will become in terms of speech."
Teams are also working on expression. Vikash Gilja of Paradromics described decoding whether a sentence is a statement or a question, along with speed and loudness. Sergey Stavisky of UC Davis said his group is exploring decoding emotional tone and has tried humming, noting that singing remains difficult. Samantha Brosler of UC San Francisco reported that a single implant decoded both words and gestures, such as thumbs up or a shrug, to drive a digital avatar, though only a few phrases and gestures were tested.
David Brandman, the UC Davis neurosurgeon who implanted Harrell's device, said a conversation offering patients a fully wireless implant is about five to 10 years away.
At a Glance
- Hochberg estimated fewer than 10 people, perhaps five or six, have implants decoding attempted speech.
- Paradromics' wireless device has 421 sensors and was implanted this summer in a 68-year-old Michigan woman with primary lateral sclerosis.
- Latest brain-to-voice tests with Casey Harrell reached about 95 percent accuracy for listeners, per a bioRxiv preprint.
- Hochberg said likely a dozen or more companies are pursuing implantable brain-computer interfaces.
- Surgeon David Brandman estimated a fully wireless implant for ALS patients is five to 10 years away.
Sources
This brief was written by Brightside Report from the reporting linked above. Spot an error? Request a correction.