A New Computer Chip Pushes Machine-Human Interface
In a fifth-floor lab of the University of Toronto's $108-million Bahen Centre for Information Technology, assistant professor Roman Genov suddenly becomes nervous. At Genov's request, a well-meaning grad student shows a guest one of the electrical engineer's groundbreaking microchips, designed and built to record brain activity in unprecedented detail. After the student cracks the lid on the protective plastic case, she reaches to remove the pill-sized, $16,000 processor from its cushy foam bed. Genov can barely contain himself. "Don't touch this," he stammers, his hand reaching to intercept hers. She balks; Genov relaxes. "Just keep it in there."
There are only two such chips, and dropping one on the floor would be a catastrophic setback to Genov's research into the arcane world of how the brain functions. It represents no less than a new threshold in the scientific evolution of the machine-human interface. The goal is to create what is essentially a mini-computer and implant it in the brain to monitor electrical activity in someone who, for example, suffers from epilepsy. The implant would then detect the onset of a seizure and trigger its own signal to stop the event.
That day is far away, and it may never come. But collaborators say Genov's chips will certainly further our understanding of the brain, and there are few if any chips that are better. Wherever it is that we eventually find ourselves, it will be because of a collaborative alchemy, the result of a multidisciplinary effort that, in addition to Genov, includes Berj Bardakjian, a mathematics expert at the Institute of Biomaterials and Biomedical Engineering, and U of T electrophysiologist Dr. Peter Carlen. Together, the trio are melding hardware, software and biology. "The sexy application is seizure prevention and control, which may or may not work," Carlen says. "But what will work almost for sure is the ability to have wireless, low-power chips with multiple sensors, packaged with some computing power to analyze brain activity."
The vast majority of today's electronics are digital, but the language of nature is analog. The brain isn't either on or off. There's a spectrum of electrical activity that can be measured. Genov's chip consists of a sliver of silicon embedded with transistors and capacitors, on top of which sit 256 electrodes that resemble a bed of nails. "We are among the top groups in the world with the largest number of recording channels," Genov says.
Current testing monitors dissected rat brains. One key design challenge is keeping heat buildup in the chip to a bare minimum. To avoid burning body tissue, the microchip cannot rise more than one degree Celsius. Genov has accomplished this, astonishing when one considers that in performing certain functions, his chip exceeds the computing power of a Pentium processor - a chip that generates so much heat it will burn skin if touched.
In Carlen's lab at Toronto Western Hospital, graduate students Marija Cotic and Eunji Kang are in the scientific trenches - Cotic developing new ways to analyze complex data gathered from several electrodes at multiple sites on the brain, Kang working on ways to analyze brain states and control seizures. Dr. Demitre Serletis, a resident neurosurgeon working on his Ph.D., says the microchip is "amazing - this is the kind of thing you'd see in science fiction movies 10 years ago, and here we are, starting to do this research - and getting results." Carlen repeatedly refers to the chip as beautiful. Why? "That box there is an amplifier for two voltage sensing channels [to read the brain's electrical activity]," Carlen says, pointing to a device the size of a VCR. "And Roman has chips that have 256 voltage sensors you can barely see with your eye. That's beautiful."
See also NEUROSCIENCE.
Maclean's March 19, 2007