Experimental neurochips fabricated in a lab.

22 August 2010

Fans of the manga Ghost In the Shell no doubt remember one of the more visually stunning pages at the beginning of the saga, CG art depicting a neurochip, which in the series was the technology underlying artificial intelligence and the prosthetic brains which made full body cyborgs possible. Not a few of us have dreamed of the day in which it would be possible to directly interface doped silicon processors with our wetware and move information out of one and into the other with little more than a thought. However, our science fiction-fueled dreams are just that, dreams, and probably will remain so for a number of years to come. However, a news article in the Globe and Mail late last week gave no small number of us pause: researchers at the Hotchkiss Brain Institute of the University of Calgary have made significant advances fusing organic neurons to silicon chips in the lab. Six years ago they succeeded with snail neurons and they say that they are now ready to start experimenting with human brain cells.

Details are a bit sketchy right now but the gist of it is that they've made the mechanisms which detect the minute and nuanced puffs of neurotransmitters and electrical impulses (properly called action potentials) much more sensitive. The experiments they have planned for the immediate future involve testing psychotropic compounds on samples of human brain tissue interfaced with their labtype neurochips to determine which are likely to be most efficacious in treating certain brain disorders. The standard protocol right now is to diagnose someone and look up what drugs (in decreasing order of likelihood) can be used to treat a disorder; the thing is sometimes the first choice doesn't work and something else has to be tried. Natural differentiation makes a species resilient, after all, but that differentiation also means that there are no silver bullets for treating a lot of things. By testing the compounds on the interfaced samples they can observe how the neurons react and better determine what dosages, compounds, or combinations of compounds might work. The brain tissue in question will be donated by an epilepsy patient undergoing surgery to ease their symptoms.

The desired eventual conclusion of this research is hoped to be a fully functional bidirectional interface between silicon chips and living brains or nerves, which is still a relatively primitive technology right now. Late generation prosthetics, while impressive, are not interfaced with individual nerves yet; they are instead connected to bundles of nerves which introduces problems of resolution. The pickups are often larger than their intended targets so interfacing with multiple signal sources cannot be avoided. To put it another way, plugging into a group of 50 nerves results in a signal which is the composite of the signal traveling through every nerve in that bundle plus some amount of noise picked up by the electrodes, plus some amount of noise introduced by the act of amplification. It's hard to avoid touching multiple nerves because fabricating pickups small enough to avoid doing so is nontrivial. Lilly electrodes can interface with individual nerve cells but they require microsurgical procedures to install and there are nonobvious difficulties in using them for long periods of time, such as the continued growth of nerve cells causing contact with the electrodes to be broken through migration or formation of fiberous tissue.

We're not there yet, but we can see the horizon.