Faculty Member Turns an Early Career Award into a Gold Mine

• RSS
• Print

Max Riesenhuber, PhD, had a problem. He needed time – a lot of it - on a functional Magnetic Resonance Imaging (fMRI) brain scanner to test his new method for investigating how the brain recognizes objects and learns from experience. But said time was expensive.

It was 2003, and Riesenhuber had just been recruited to Georgetown University Medical Center’s Department of Neuroscience from the Massachusetts Institute of Technology, where he did his graduate and postgraduate research. He had developed a computational model of object recognition in the brain that had helped to interpret data from animal electrophysiology experiments, and for his pioneering work Riesenhuber had been named as one of Technology Review magazine’s “100 innovators 35 or younger whose technologies are poised to make a dramatic impact on our world.”

He now wanted to use that method as a framework to investigate how the human brain recognizes objects and learns to assign meaning to images. But to do this, Riesenhuber needed to use the then-new state-of-the-art functional MRI, which measures which parts of a brain are active while a person is doing particular tasks, at GUMC’s new Center for Functional and Molecular Imaging.

He put his concept into a grant and submitted it to the National Science Foundation (NSF). The NSF answered back with a Faculty Early Career Development award of $742,000.

With one grant, Riesenhuber turned the funds into a treasure trove of new brain science findings. In April, he published the third of three research studies on how the brain handles object recognition, all in Neuron, a high-profile neuroscience journal.

These papers show that the brain appears to use the same computational strategies to recognize different objects, whether faces, words, or more specialized objects such as cars, but also that brain cells are “tuned” in different ways in order to discriminate between similar objects – that is, how selective neurons need to be if a certain object is to learned and recognized again.

In facial recognition, Riesenhuber says, the neurons have to be selective enough that together, they can represent particular faces. However, their tuning should not be so specific that they would not respond to unfamiliar but similar faces, allowing the brain to learn new faces more efficiently. He's using these principles for various practical applications as well, such as enhanced binoculars that he's developing for the military.

“If neurons are too broadly tuned, the ability to discriminate between faces will be affected,” Riesenhuber says. “That is what we think is going on in autism, which includes a deficit in recognizing faces.”

Whereas faces appear to be learned and recognized by a group of neurons on the right side of the brain, the ability to learn and utilize words occurs on the left side, and is a substantially more selective process, he says.

In fact, his latest research, published on April 30, shows that there is virtually a dictionary in the brain. Successful reading requires the brain to correctly recognize printed individual words, and Riesenhuber and his colleagues found that neurons seem to be specialized to process words as whole-word units. In other words, groups of neurons fire in response to a single word, and not to any other.

That may be of interest for the study of reading disorders such as dyslexia, because it would be harder to recognize words if the neurons responsible for the brain dictionary are not tuned as tightly as they need to be, Riesenhuber says.

What these findings suggest is that the brain uses very flexible and general, yet powerful, mechanisms to learn to recognize objects.

Riesenhuber promises more discoveries to come – after all, he still has a year to go on his NSF award.

By Renee Twombly, GUMC Communications