Sunday, June 3, 2007

70. Experimental models


Mark said...

A brain on a chip

WHAT IT IS: A model of the brain’s pathways on a computer chip

LEAD INVENTOR: Bioengineer Kwabena Boahen, Stanford

SNAPSHOT: To make computers as powerful and efficient as the human brain, Boahen is developing ways to base the design of electronics on our mental circuitry. His circuits combine hardware and software such that the wiring can self-adjust to “learn” new functions. He has already recreated the networks of the retina.

FANTASY APP: Because Boahen’s electronics work like the brain, they may one day form the basis of a prosthetic cortex—yes, an artificial brain.

THE STORY: When Boahen, 41, got his first computer as a teenager in Ghana back in the ’80s, he was appalled to learn how it worked. “To do anything interesting was so complicated,” he says. “I figured there has to be a better way.” He doesn’t have a much higher opinion of today’s computer technology. While exponentially faster and more powerful than his first machine, even the latest supercomputers are woefully inefficient when compared to the one mechanism that really impresses him: the brain. Running on a mere 10 watts of power, the brain performs calculations that would require millions of Pentium processors and consume a gigawatt of power.

Rather than compete with biology, Boahen, a professor of bioengineering, seeks to emulate it on a chip. The physical realities are daunting: the brain is composed of a trillion neurons connected in a web so fine that a single cubic millimeter of cortex contains four kilometers of wiring. Moreover, the pattern in which the neurons are connected is constantly changing as we learn.

So Boahen has begun his research on a network similar in design and function to those of the brain proper, yet more self-contained and much simpler: the retina. Using transistors, he has built a circuit on a computer chip that measures light in the way that a real retina does. Instead of measuring lighting conditions at every point in its field of view, the retina compares light levels at many points and registers the contrasts. Greater efficiency means lower power consumption, resulting in a chip that can be used as a retinal prosthesis without overheating the eye.

Boahen is seeking to build a partial cortex on a circuit board within the next few years. He can imagine neurologists, soldering irons in hand, experimenting on this silicon brain with a directness that could never be attempted on a living organ. Insights from these experiments may allow surgeons decades from now to replace damaged areas of a real brain with silicon. In the meantime, Boahen’s brain-inspired electronics are demonstrating ways in which computer design can be smarter, one day allowing handheld devices to become smaller and to run for longer on less power.

Mark said...

A virtual heart

WHAT IT IS: Three-dimensional computer simulation of body parts—for starters, the human heart

LEAD INVENTOR: Computer scientist Katherine Yelick, UC Berkeley

SNAPSHOT: Yelick and her colleagues have developed a computer language that will eventually allow hospitals to create virtual copies of your organs, allowing doctors to test their treatments online, without jeopardizing the patient.

FANTASY APP: Prostheses and drugs custom designed to meet a patient’s specific needs

THE STORY: A technique that saves 99.9 percent of patients may kill or damage the remaining 0.1 percent. The only way to be certain that a drug or medical procedure will cause an individual less harm than good would be to test it first on a virtual doppelgänger.

Yelick, 46, a professor of computer science, is working one organ at a time to invent such an avatar. Using Titanium, a programming language of her own design, she has simulated the three-dimensional dynamics of a beating heart. Her programming allows her to parcel out the organ’s intricacies over many computers running in parallel, handling muscle contraction, for in­stance, on one set of machines and valve closure on another. (Others in her research group have applied the simulation technique to other body parts, most successfully the inner ear.)

The challenge of making a heart beat on a computer screen suggests how difficult creating a fully functioning avatar will be. While fluids move through a rigid machine in a relatively straightforward way, the muscle fiber of the heart is elastic, constantly changing in shape as the blood moves through it. Yelick began working on her computerized heart in 1993; eventually, using an IBM supercomputer, she was able to simulate one heartbeat every 24 hours.

Fortunately, there’s no need for a naturalistic 60-beat-per-minute pulse to test, for example, how an artificial valve will function in a damaged heart. Yelick believes that once Titanium can precisely simulate a deformity, doctors will be able to build custom prostheses, such as heart valves, rather than rely on one-size-fits-all components. Simulating an individual body’s chemistry will likewise be helpful, eventually allowing doctors to experiment with potent drug cocktails or even to custom-design pharmaceuticals for the rarest of ailments. Having an artificial patient online, biologists will be able to move beyond the observations of medical imaging and try experiments that would never be permissible on a living being.

Mark said...

Sacred Cows and Golden Geese: The Human Cost of Experiments on Animals, by C. Ray Greek, Jean Swingle Greek, Jane Goodall (Foreword)

Specious Science: How Genetics and Evolution Reveal Why Medical Research on Animals Harms Humans, by C. Ray Greek, MD, and Jean Swingle Greek, DVM