| |
When it comes to recognizing objects, humans easily beat computers—but how our brains do it isn’t known in great detail. Human studies have relied on noninvasive methods, gathering data from fairly general brain locations and lagging considerably behind the brain’s actual processing speeds.
Wanting more precise information, researcher Gabriel Kreiman, PhD, of Children’s Hospital Boston’s Department of Ophthalmology, teamed up with Joseph Madsen, MD, of Neurosurgery. Madsen was working with neurologists and technologists in the Epilepsy Service to perform brain mapping in 11 patients as part of surgery for epilepsy, to locate the "hot" seizure zone and ensure that essential brain functions would not be harmed.
This mapping places electrodes directly on the brain, allowing Kreiman to study signals from very discrete locations. When patients were shown images—variations on animals, chairs, human faces, fruits and vehicles, displayed in different sizes and degrees of rotation—Kreiman could capture signals as fast as 100 milliseconds (thousandths of a second) later.
The results, published in Neuron on April 30, show that the brain, at a very early processing stage, can recognize objects rapidly, under a variety of conditions. "Some people think that if you don’t have feedback from higher brain areas, you don’t have vision," explains Kreiman. "But we’ve shown that an initial wave of activity gives a quick initial impression that’s very powerful." Although feedback likely occurs later, this fast visual processing may have given our species an edge in situations like encountering a predator, he says.
Kreiman also showed that specific areas of the visual cortex selectively recognize specific kinds of objects, like faces. He and Madsen are now studying patients while showing them movies—more closely simulating how images look in real life—and analyzing their brain activity frame by frame.
The goal? As the brain’s vision algorithms become clearer, Kreiman envisions teaching them to computers, which could then provide additional sets of eyes, helping to spot terrorists in airports, helping drivers avoid collisions or scanning tumor samples for malignancy. Kreiman even imagines brain-computer interfaces that would give blind people at least partial sight. As for Madsen, he hopes that more precise analysis of brain signals will make epilepsy surgery safer and more effective.
|
|
| |
