Helping Your Brain Be All That It Can Be: Neuro Research Works to Combine Human Brain Capabilities with Computer Data Crunching Power (3/29/2008)

As fantastic as computers are at processing mountains of data at lightning speed, they still can't beat the human brain at, for instance, detecting a friend's face in a crowded airport terminal, or even at more specialized tasks such as detecting suspicious objects in X-ray scans of checked baggage. The human brain can do these tasks not just very well but also very rapidly, within a fraction of a second, says Georgetown University Medical Center neuroscientist Maximilian Riesenhuber, Ph.D., of the Laboratory for Computational Cognitive Neuroscience. However, while the brain's visual system can perform the actual detection task very rapidly, even for multiple images presented at the same time, turning these decisions into behavioral responses is much slower.

So Riesenhuber is working on circumventing the response "bottleneck" by building hybrid brain-machine object detection systems that marry the brain's sensory and cognitive processing speed with a computer's ability to quickly process large amounts of data. In his research, he's sifting through brain signals recorded via electroencephalography (EEG) to better understand the brain's potential and perform visual detection at the speed of thought.

"In a previous project we showed that a hybrid brain-machine system could analyze satellite images several-fold faster than humans could." A new project, that has attracted interest from defense contractors and the Department of Defense, aims to build on this success by developing neurally assisted "binoculars" that can identify objects of interest at large distances and over a wide viewing angle by combining a computer vision system that identifies potentially interesting image snippets with a neurally based system that rapidly flashes these snippets at the viewer and analyzes the resulting brain activity to decide which parts of the environment merit a closer look.

"We are now trying to figure out how fast we can push things to fully leverage the brain's processing power, including its ability to perform multiple tasks at once. There is a critical mass building in this field to push the envelope and explore how we can 'augment cognition' by combining the best of both worlds, neuroscience and engineering," Riesenhuber says. The resulting technology, he adds, will be of interest for a variety of applications."

Riesenhuber is fascinated by the exceedingly complex question of how a human brain assigns meaning to all of the sensory input it receives. Vision alone involves millions of photoreceptors and dozens of different brain areas, all of which are being continually activated as the eyes are being used. He uses computational analyses, along with functional MRI and EEG, to help understand - and model - how the brain recognizes objects and learns from experience.

Since coming to Georgetown University in 2003 from the Massachusetts Institute of Technology, Riesenhuber has forged links with other interdisciplinary scientists, such as from Physiology and Biophysics and Neurology, in order to build a more accurate and comprehensive model of the brain. This work has applications in a wide variety of areas. For instance, the NIH is funding work in his lab on testing hypotheses for the neural bases of behavioral differences in autism, and NSF is funding projects on object recognition in complex scenes as well as on learning in vision and audition.

"The brain does an excellent job of solving a lot of very hard computational problems," he says. "Figuring out the underlying neural mechanisms is not only crucial for making progress in understanding mental disorders but also for eventually building more intelligent machines."

Note: This story has been adapted from a news release issued by Georgetown University Medical Center

Post Comments: