With help from engineers and computer scientists, two Harvard neuroscientists are setting out to catalogue all quadrillion of the brain’s connections.

Constanza Villalba

Neuroscientists spend a lot of time figuring out which parts of the brain connect with which, and which networks of neurons control specific tasks. They would love to have a more detailed map of the brain, but most would say that deciphering the complete wiring diagram of the human brain, which includes 100 billion neurons and a quadrillion connections, is too formidable a task. Two Harvard neuroscientists, Jeff Lichtman and Clay Reid, disagree.

Lichtman and Reid have teamed up with Hanspeter Pfister, a computer scientist with the Mitsubishi Electric Research Laboratories in Cambridge and Harvard’s Initiative for Innovative Computing, to catalogue the connections between each of the brain’s cells.

The goal of the project—called the Connectome —is to build a database of high-resolution images of every cubic millimeter of the mammalian nervous system, mostly using electron microscopes.

Once completed, the connectome could become a valuable tool for neuroscientists working on a wide range of problems. Reid, for example, would use it to follow the neural pathways that visual information takes through the brain. The connectome would help Lichtman study how connections between cells form and dissolve according to changes in an animal’s experience or context.

The project is in its infancy. The researchers have started with images from mice and rats, but they hope someday to include humans.

“To get an image of every part of the human brain at a level of resolution high enough to see every connection would involve gathering approximately one million petabytes of data,” says Lichtman. “That’s more information than in all the libraries in the world.”

Data overload

While no one is prepared to tackle that volume of data just yet, Lichtman is moving in that direction. He has already generated a connectome for one of the tiny muscles that controls ear movements in the mouse.

One of the limiting factors in expanding the connectome has been the speed at which researchers can acquire brain images. One of the slowest and more problematic steps is preparing the tissue for microscopy; researchers must slice the brain into 50 nanometer-thick sections and mount them. The process can take hours and the tissue is so fragile that often it is destroyed before it can be mounted.

To speed up the process and reduce error, Lichtman enlisted the help of Ken Hayworth, a University of Southern California neuroscience graduate student who developed a machine that automates the sectioning of brain tissue.

Hayworth’s device is called the ATLUM. Rather than cut tissue into discrete sections that have to be individually and manually mounted, it rotates the tissue, cutting it into long strips, and automatically mounts the strips onto carbon-covered tape that can be read by an electron microscope.

Brain panorama

Reid is taking a different approach to speeding up image acquisition. He sticks with standard tissue-mounting techniques but uses four cameras instead of one to photograph a larger area of tissue at one time. The result is hundreds of pictures—each taken from a slightly different vantage point.

“It’s like you’re taking a panorama with a [conventional] camera,” says Reid, “and it doesn’t all fit into your field of view.”

The next step is to align the individual images so that they can be seamlessly stitched together into one picture. Pfister is collaborating with Microsoft to develop software for that task.

Pfister’s next project will be to find ways to stack and merge brain slice images to create a three-dimensional model of the brain and its connections.

Stephen Smith, of Stanford University, who is both a scientific competitor and a friend of Lichtman’s, says that Lichtman and Reid are “good people to be leading the charge on the Connectome.” But he says they’ll need to work with neuroscientists using different imaging approaches, because no single strategy can answer all the questions raised by the Connectome.