Tech

Software that mimics Darwinian natural selection could help boost the energy efficiency of brain implants and reduce the need for surgery to replace their batteries.

Conditions like epilepsy and Parkinson's can, in some cases, be controlled using implants that electrically stimulate nerve clusters deep within the brain.

But until now, the precise shape, or waveform, of the electrical pulse applied has been a matter of guesswork. Some implants fire out "exponential" pulses that rise and die away just as quickly, while others might use rectangular or ramp-shaped pulses. As long as the pulse activates enough nerves it will have performed its role, but as each waveform uses a different amount of electrical energy, it's possible that implants could be putting out unnecessarily strong pulses that waste precious battery power.

"When the battery is depleted, the entire stimulator must be replaced through an expensive and invasive surgical procedure," says Warren Grill, a biomedical engineer at Duke University in Durham, North Carolina.

Evolutionary design

Grill and his colleague Amorn Wongsarnpigoon decided to "evolve" the perfect low-energy pulse waveform. To do this, the pair used a genetic algorithm (GA).

Working like natural selection, the GA takes a population of random waveforms, mutates the "fittest" of them – in this case, those with lowest energy use – and then "interbreeds" the mutated forms to make new "offspring" waveforms. The process is then repeated through several "generations" until the optimal waveform is found.

The researchers began with five groups of differently shaped pulses and ran them through a genetic algorithm with 10,000 generations. Each pulse had to activate at least 50 per cent of a group of virtual mammalian nerve cells in order to make it to the next generation, and those that did so with minimum energy requirements were favoured.

Saved by the bell

In all cases the optimal waveform was a truncated Gaussian pulse – a bell curve with a vertical line cutting short the lower parts of each.

To verify that the GA had hit on the best waveform, the pair tested an implant attached to the sciatic nerve in three anaesthetised cats. Compared to all previous implant waveforms, they found the new ones were indeed superior in terms of energy efficiency while no less able to perform their nerve stimulation role.

The Duke team estimates that with the new waveforms, implant batteries will only need changing between 5 to 6 times in 30 years, against 8 to 10 times if the older waveforms are used.

More room for improvement

The energy use of implants could be cut further, Grill says. "We also want to use GAs to optimise electrode geometries to improve the efficiency of stimulation. That would further increase the lifetime of the batteries."

Computer scientist Peter Bentley, a specialist in adaptive and evolved systems at University College London, UK, and a former organiser of GECCO, a major annual conference on GAs, is impressed with the work.

"This is an excellent example of the use of evolutionary optimisation to provide tangible benefits to patients. Any technique that can minimise the number of surgeries is clearly a significant step forward," he says.

Steven Manos, a computer scientist at the University of Melbourne, Australia, and who won a GECCO award for a novel broadband optical fibre designed with GAs, agrees. "It's a very nifty and relevant application of a GA," he says.

Journal reference: Journal of Neural Engineering, DOI: 10.1088/1741-2560/7/4/046009

If you would like to reuse any content from New Scientist, either in print or online, please contact the syndication department first for permission. New Scientist does not own rights to photos, but there are a variety of licensing options available for use of articles and graphics we own the copyright to.