Neuroprosthetics

Popular Science has a great article on the recent advances in prosthetics. They hit on one of the topics that I think has been really under-researched: neural to machine interfaces. What you would really like to do with a prosthetic is have it communicate directly to and recieve information directly from the central nervous system. To whit:

Once science figures out better ways to attach artificial limbs, prosthetics themselves need to become smarter, able to act on signals sent directly from the brain. Consider the case of Jesse Sullivan, a power lineman from Dayton, Tennessee, who lost both arms at the shoulder after being electrocuted on the job in 2001. A year later, doctors transferred four nerves (which were no longer infusing muscle) that had controlled his left arm out of his shoulder area and into his pectoral muscles. Six months after that, Todd Kuiken, director of the Rehabilitation Institute of Chicago's Neural Engineering Center for Artificial Limbs, detected signals in the nerves. Kuiken's team studded the surface of Sullivan's chest with electrodes and joined them with wires to a multi-jointed prosthetic. The goal was to connect brain to artificial arm by redirecting signals from Sullivan's severed nerves. It worked. When doctors asked Sullivan to think about opening his hand, the device, almost instinctively, sprung open. "It was the greatest feeling I'd had since I'd been hurt," Sullivan says. He can now eat, mow the lawn, and do his laundry, but his arm fulfills only a small fraction of the nerves' potential power. The nerve for hand closing controls at least 20 muscles, Kuiken says, "and I'm using it for just two different signals. If we tease it out, we might get better and better control." Kuiken is now developing sensors that will allow Sullivan to feel what he is touching.

Another way to power artificial limbs is to bypass the nerves and tap directly into the brain. That's what John Donoghue, director of Brown's Brain Science program and the chief scientific officer at Cyberkinetics Neurotechnology Systems in Foxborough, Massachusetts, is working toward with the invention of BrainGate, a chip that was implanted in 2004 into paralyzed 25-year-old stabbing victim Matt Nagle. With the four-millimeter-square chip in his primary motor cortex, Nagle thinks about moving a cursor on the computer screen to the right. His neurons fire in a certain pattern, and those data are transmitted through a plug affixed to his skull to the computer, which moves the cursor. Soon, BrainGate's developers got really ambitious. They lay a prosthetic arm, tethered to the computer, on Nagle's lap and told him to open the hand. He did, just by thinking, and swore in amazement as the hand unfurled. Donoghue promises that future versions will operate wirelessly; Cyberkinetics is developing a control system that uses wireless transmitters and fully implantable power sources.

This article may be a little bit bullish about the state of the field when it comes to neural interfaces. I saw a talk by one of Donoghue's colaborators named Michael Black. Basically the take home that I got from it was that we do not yet understand the basics of motor encoding that are necessary for outgoing brain-to-machine signals. They are making strides -- if for no other reason than they are now actually recording the signals they are going to need to decode -- but they don't have a high ability to resolve those signals into subtle motions.

Likewise, I am not aware of anyone who is trying to use spinal or other implants to put touch sensory information back into the nervous system.

I think this is a great area to study, but if we want to come through with what the hype says the action in my opinion is in neural encoding.