ATLANTA Emory University and Georgia Institute of Technology have collaborated to perform calculations with living leech neurons interfaced to a computer. The project hopes to perfect the living-neuron computation process, and eventually grow to a full wetware computer on a silicon substrate.
"What we have done so far is demonstrate the feasibility of isolating nerve cells so that we may use a computer as an intermediary in their communications with each other. We are a long way from creating living computers, but we do eventually want to grow neurons on a silicon substrate as a first step," said Ronald Calabrese, a professor at Emory University, and the director of a laboratory dedicated to studying the neural circuitry of the medical leech. Calabrese chose leech neurons because they are relatively large and well-cataloged.
His laboratory has made extensive models of the leech's neural structure and organization, especially in its timing networks for locomotion and heartbeat. Calabrese said the neural networks that manage a leech's rhythmic motor behavior result from coupled oscillations among various segments. In all, the leech has about 20 segments, called ganglia, each of which houses about 10,000 neurons. Motor activity, for instance, results from progressive phase differences among the nerve impulses.
Likewise, the leech's heartbeat can also be modeled as two coupled oscillators. Until now, Calabrese has had to be content with studying the intersegmental coupled oscillations in the leech nervous system by virtue of computer models. In particular, he modeled individual neurons as single compartments of a Hodgkin and Huxley conductance. "We have been testing the observed 15 percent phase lag between the segmental oscillators to see if it was the result of inherent differences in the periods of the segmental oscillators. We suspect that the asymmetries in the synaptic coupling between the oscillators affects their phase relationship," said Calabrese.
When Georgia Institute of Technology researcher William Ditto offered to collaborate with Calabrese in making direct recordings of the interactions among leech neurons, instead of merely simulating them on a computer, Calabrese jumped at the chance.
"Ditto has a theory that the mathematics of chaos can best describe the actions of neurons that's his motivation in our collaboration. But I just want to discover how the leech neurons perform computations regardless of what mathematics best describes them," said Calabrese.
The experimental setup devised by the pair involved surgically isolating two of the 20 ganglia from a leech, allowing a computer to supply the input to each, and also record the output. The computer, acting as an intermediary, was able to interpret the signals coming out of the neurons, in response to the stimuli going in, and effectively manage the communications between the two.
"There is a lot riding on the interpretation we make of the outputs from each ganglion, regarding the claim that they are making calculations. But we have definitely got them communicating with each other," said Calabrese. The downside of the experimental setup is that the leech's nervous system can only survive for about three or four hours after it is hooked up to the computer. Eventually, however, Calabrese hopes his work will result in a technology that places neurons on a silicon substrate that "feeds" them.
The setup would work like the team's current experiment, in that the substrate would stimulate their inputs and read their outputs, but nutrients flowing onto the chip would keep the neurons alive. "Other labs, like Carver Mead's at Cal Tech, have experimented with feeding neurons on a silicon substrate. Our motivation is to develop a better understanding of the way neurons operate," said Calabrese.