Google helps Harvard scientists map the human brain

The result is the most detailed digital map, or “neural network,” of the human brain ever created.

On Thursday, Lichtman and his partners revealed results For their efforts in the prestigious journal Science, and also published in Internet renderings Of the human brain unlike ever before. It came complete with software that allowed viewers to move through a microscopic space landscape so detailed that Lichtman couldn’t help but get poetic when talking about it.

“It’s a strange world inside your head,” he said. “The neurons themselves are really stunningly beautiful. There’s no two ways about it.”

It is true that the insights derived from the small sample have not yet revealed the secrets of autism, schizophrenia, or depression. They cannot yet explain the mechanisms of human learning, memory, and personality at the cellular level. But it represents an important first step in this direction, and provides a tantalizing glimpse into the kind of visions we might see in the coming decades.

In this complex landscape are strange, never-before-seen, textbook-less structures, including, in Lichtman’s words, “fantastically strange.” Neurons that point in only one direction, directly opposite each other. Axons, the brain’s long-range fiber optic cables, veer from straight to odd lines “Swirls” Which look like turbans – then unravel and return to straight lines. The purpose of Many of these strange anomalies remain the subject of future study.

Some are already generating potential paradigm-shifting theories that may reveal fundamental new insights into how the brain works. The highlight, Lichtman said, is the discovery of what appears to be a new and extremely rare type of “superconnection” that connects individuals. Neurons to axonal fibers that carry information and that intersect with the brain. Each superconnection has a combination of about 50 or so protrusions where there is usually only one. Lichtman hypothesizes that these structures may help explain how learned habits, such as stopping at a red light without thinking, remain engraved in the physical structure of the brain.

“Probably 99% of the connections between axons and individual brain cells are these very weak connections,” Lichtman says. But “those strong connections are so strong that information can flow very efficiently. This might be a way to explain the fact that after you learn something, you have the automatic ability to do it.”

The new study is part of a much larger series of projects funded by the BRAIN Initiative, a massive scientific effort launched by the Obama administration in 2013 to uncover fundamental insights about the human brain.

“It’s a pretty big deal,” said Ed Lin, a neuroscientist at the Allen Institute for Brain Science in Seattle, who was not involved in the study. The mapping is a “human first.”

Len, who helps drive Another component of the BRAIN initiative, He said Lichtman’s work could help change our understanding of the human brain and dramatically improve our ability to treat diseases.

“We have a very poor understanding of these circuits,” he said. “Imagine your cell phone broke and you don’t know anything about the components of the cell phone or how they’re connected together, and you’re trying to fix it. If we don’t understand how things are connected together at all, our chances of being able to fix it are slim.

Lichtman’s project was originally funded by part of a five-year, $7 million grant from the National Institutes of Health, and recently received an additional $30 million. More than five years of relevant NIH program. The federal agency’s goal is to improve our understanding of diseases that affect cognition and emotion.

The money is also financing another Related projectsis supplemented by the pro bono collaborative efforts of Google, which provided the computational power and engineering labor needed to run the project.

After the human brain sample was colored, cut, and photographed In Lichtman laboratoryGoogle engineers applied machine learning to connect those chips back together and applied colors to make the wires visible to the naked eye.

The scope of the challenge of recreating that 1 cubic millimeter sample From the human brain In digital form it was so great that the effort to continue imaging an entire human brain would have to wait. Accurate picture of The entire human brain is in This scale, Lichtman says, would roughly equal the amount of data produced in the entire world in one year.

Which is why the next effort will be more modest: Over the next five years, Lichtman and his collaborators aim to photograph the first 10-cubic-millimeter section. from U.S Mouse brain. The project is a proof of concept with the ultimate goal: a complete mouse brain, 50 times larger.

“The human brain will be another factor a thousand times larger than the mouse brain,” Lichtman says. “We don’t have the ability to store that information.”

The gains from all these efforts could ultimately be huge. Google and others expect to use the results to improve their ability to invent artificial intelligence algorithms modeled after the human brain.

For his part, Lichtman hopes to answer fundamental questions about the human mind: How can representations of the world become imprinted inside our heads? What are the material foundations of knowledge?

The project has already taken him into intellectual territory he never expected to enter. He describes the experience of sitting in his office with a new “neuroglass” tool that allowed him to maneuver through the visual landscape of a neural network as “wonderful,” “magical,” and “like a fantasy.” He wanted to click on each cell individually.

Invoking the names of Magellan, Amerigo Vespucci and other famous explorers, Lichtman praised the excitement of the discovery.

“This is very similar to using the Hubble Telescope or the James Webb Telescope,” Lichtman says. “But it’s not a telescope, it’s a microscope that allows us to look inside. There are certainly all kinds of things we’ve never seen before. We’re exploring terra incognita.”

Adam Piore can be reached at [email protected].