In a scientific accomplishment many once thought impossible, an international team of researchers has successfully created a highly detailed three-dimensional map of brain tissue from a mouse. Using just one cubic millimeter of the brain’s visual cortex, scientists were able to chart 84,000 neurons, more than 500 million synapses, and 200,000 brain cells. This microscopic sample, no bigger than a grain of sand, generated a staggering 1.6 petabytes of data, the equivalent of 22 years of nonstop high-definition video.
A Decade of Work by Global Experts
This achievement is the result of nearly ten years of work under the Machine Intelligence from Cortical Networks program, known as MICrONS. More than 150 scientists across 22 institutions contributed, with key leadership from the Allen Institute for Brain Science in Seattle, Baylor College of Medicine in Houston, and Princeton University. Funding came primarily from the U.S. Office of the Director of National Intelligence and the National Institutes of Health.
Clay Reid, a senior investigator at the Allen Institute, said, “It really has been one of the holy grails of the field from the beginning. There are many thousands of neuroscientists who study the cerebral cortex, and pretty much everyone who studies the cerebral cortex would like to be able to know what are the sources of inputs to any given cell within the cortex, and what are the outputs of that cell.”
Watching Movies to Light Up the Brain
To capture brain activity, researchers at Baylor recorded signals from the visual cortex of an awake mouse. The mouse was placed on a treadmill and shown short video clips from movies such as “The Matrix” and “Mad Max: Fury Road,” along with YouTube footage of extreme sports like motocross and luge. This helped researchers track how neurons fired in real time.
After the mouse was euthanized, scientists at the Allen Institute sliced the brain tissue into more than 28,000 layers, each one thinner than a human hair. These slices were photographed using electron microscopes, and the images were stitched back together to create a complete digital map.
Nuno Maçarico da Costa, associate investigator at the Allen Institute, explained the process: “That took us about 12 days and 12 nights with the team taking shifts around the clock. Not because we were cutting it by hand – it’s a machine that is automated. We needed to be there to stop at any point in time if we thought we’re going to lose more than a section in a row.”
AI and Human Effort Combined
At Princeton University, artificial intelligence was used to trace the structure of each neuron. The process, called segmentation, colored each neuron to make its branches and pathways visible. Scientists continue to manually verify the AI’s work to ensure accuracy.
Dr. Sebastian Seung, a professor of neuroscience and computer science at Princeton, called the effort a digital transformation for brain science. “With a few keystrokes you can search for information and get the results in seconds. Some of that information would have taken a whole Ph.D. thesis to get before,” he said.
The final product is known as a “connectome,” a complete wiring diagram of the brain tissue showing how neurons are physically connected and how they behave.
New Insights into How the Brain Works
One major finding involved inhibitory neurons, which control the activity of other brain cells. Reid said, “They are not just on-off switches for the entire circuit. Different types of inhibitory neurons inhibit different elements within the circuit. They don’t turn on and off every light in the building.”
This level of detail could help scientists better understand neurological conditions like autism, schizophrenia, Parkinson’s disease, and Alzheimer’s disease. Comparing healthy brain wiring with disease models may reveal the root causes of these disorders.
“If you have a broken radio and you have the circuit diagram, you’ll be in a better position to fix it,” said da Costa.
Advancing Artificial Intelligence
The MICrONS project was also designed to benefit the field of artificial intelligence. The Intelligence Advanced Research Projects Activity (IARPA) backed the project to explore how the brain’s structure could inspire more efficient machine learning systems. “Why don’t we use the most complete and detailed characterization of a cortical circuit perhaps as inspiration for new architectures for machine learning?” Reid asked.
A Milestone Compared to the Human Genome Project
Experts have compared the mouse brain map to the Human Genome Project. David Markowitz, a former IARPA program manager, called it a “watershed moment for neuroscience, comparable to the Human Genome Project in their transformative potential.”
Even though this map represents only 1/1,000 of a mouse brain, it is about 20 times larger than the entire brain of a fruit fly and far more complex. The researchers hope to map an entire mouse brain in the near future, but doing the same for a human brain would be a far greater challenge.
“Because of the size of the human brain, it is unimaginable, and I would say impossible in any reasonable future, to map the entire human brain at the level that one did for this cubic millimeter,” Reid explained.
A Turning Point for Brain Science
The techniques used in this project have already been adopted by other researchers working on the brains of insects, monkeys, and even humans. The tools and processes developed through this effort are expected to dramatically accelerate discoveries in neuroscience.
Thomas Macrina, who worked on the project as a graduate student and now leads Zetta AI, a company that assists with brain mapping, said, “We think that every neuroscience experiment should in some ways be referencing a connectome.”
For those who once thought it couldn’t be done, this project proves otherwise. In 1979, Nobel-winning biologist Francis Crick wrote, “It is no use asking for the impossible, such as, say, the exact wiring diagram for a cubic millimeter of brain tissue and the way all its neurons are firing.”
But as Reid pointed out, “Crick never set an expiration date for his pronouncement.” Forty-six years later, what once seemed impossible has become a reality, paving the way for deeper understanding of the brain and the potential to one day simulate its functions in machines.
