Operating a computer with neural signals

Imagine controlling a computer's cursor simply by focusing your thoughts. A device invented at GTRI could one day help disabled people communicate in exactly that way.

Philip R. Kennedy developed and patented the "neurotrophic electrode" in the mid-1980s, while working as a neural prosthetics researcher at Georgia Tech. His work capitalizes on the fact that the act of thinking prompts physical activity in the brain in the form of electrical impulses.

Implanted into a patient's brain, the electrode detects and captures those impulses. The electrical signals are then processed by customized microelectronics and software to move a cursor and select icons on the screen. In effect, the brain's neural signals become a computer mouse.

The system underwent its first successful human test in 1998 on a 53-year-old, bedridden man, almost completely paralyzed for six years. Six months into the test, he learned to select icons representing phrases such as "I am too cold." After a few more months, he mastered the keyboard and began forming his own words—even holding short conversations with doctors.

The key component of Kennedy's cortical control device is the hollow glass, cone-shaped neurotrophic electrode. About the size of a ballpoint pen tip, the device contains a pair of microscopic gold wires and is coated with a biocompatible substance. This coating encourages neurites—tentacle-like structures extending from neurons—to migrate into the electrode, thereby ensuring a solid electrical connection and also holding the device firmly in place.

The ability to remain stationary distinguishes Kennedy's electrode from similar efforts. It's a critical attribute if a patient is to achieve control with consistent results.

Neural Signals Operate a Computer Mouse

The low-amplitude signals traveling across the local neurons pass through the electrode. The electrode relays the signals to a tiny power induction amplifier and FM transmitter inserted just under the scalp. The signals, amplified about 1,000 times, are broadcast to a computer where signal-recognition software translates them into cursor movement.

Two electrodes are employed because basic computer operation is a two-step process. The first requires moving a cursor among a number of options; the second involves selecting one of those options.

If the electrode is implanted in the motor cortex in an area associated with, for example, finger movement, the patient can generate electrical signals to move the cursor by concentrating on moving a finger. Another electrode, implanted in an area associated with a different kind of movement, can facilitate the computer's "select" function.