The Brain Machine interface

As some of you will know, I have a particular interest in the future of human/machine interaction. Not from an ergonomics perspective, but from the idea that eventually, our thoughts, our feelings and our sense of connection with others will influence more overtly the world around us. I won’t go into the spiritual aspects of this right now, but a fascinating piece of work was published this week (Nature Physics, DOI: 10.1038/nphys1099) that heralds this kind of future.

For all its sophistication and power, our brain is built from unreliable components – one neuron can successfully provoke a signal in another only 40% of the time. This lack of efficiency frustrates neuro-engineers trying to build networks of brain cells to interface with electronics or repair damaged nervous systems. Our brains combine neurons into heavily connected groups to unite their 40% reliability into a much more reliable whole. Now human engineers working with neurons in the lab have achieved the same trick: building reliable digital logic gates that perform like those inside electronics.

Built from scratch

Elisha Moses at the Weizmann Institute of Science in Rehovot, Israel, and his students, Ofer Feinerman and Assaf Rotem, have developed a way to control the growth pattern of neurons to build reliable circuits that use neurons rather than wires.
The starting point is a glass plate coated with cell-repellent material. The desired circuit pattern is scratched into this coating and then coated with a cell-friendly adhesive. Unable to gain purchase on most of the plate, the cells are forced to grow in the scratched areas. The scratched paths are thin enough to force the neurons to grow along them in one direction only, forming straight wire-like connections around the circuit. Using this method the researchers built a device that acts like an AND logic gate, producing an output only when it receives two inputs.

Better together

The gate is made from a network of neurons in a square shape approximately 900 micrometres on a side. Three of the sides form a “horseshoe” 150-micrometres wide, and packed with neurons. On the fourth side an isolated neuron island is linked to the other sides by two thinner bridges (see image, top right). Neurons send their wire-like extensions that carry signals – axons – across those narrow bridges to the neuron island. When stimulated with a small dose of a drug, the neurons send signals around the circuit. An ion blocker is used in the centre of the horseshoe to electrically isolate one side from the other.

By changing the width of the bridges, the researchers are able to control how many axons link to the neuron island, and tune their device to behave like an AND gate. The neurons on the island only produce an output after receiving signals through BOTH of the thin bridges. Like a natural system, the device transcends the performance of individual neurons – achieving 95% reliability from a collection of 40% reliable components.

Brain interface

Rotem thinks that this provides a useful model for real brain function. “The existence of a threshold level for activation plays a central role in neuronal computation,” he says. In his logic gates and real brains alike, many neurons contribute to generate a signal strong enough to excite another group of neurons, he says.

Charles Stevens at the Salk Institute in La Jolla, California, is not so sure, pointing out that real brain “circuits” do not resemble logic gates. However, achieving reliable performance from lab-grown neurons is still impressive, he adds; “There is a sort of fascination with neural networks grown in culture, and this paper improves on the usual random networks.”

Rotem says that brain-cell logic circuits could serve as intermediaries between computers and the nervous system; “It’s difficult to physically interface [neural prosthetics] with live neurons.” Brain implants can allow the paralysed to control robot arms or learn to talk again, but suffer a drop-off in performance when scar tissue coats their electrodes. “An intermediate layer of in vitro neurons interfacing between man and machine could be advantageous.”

Best wishes