We'll never know her name or what she did for a living. All we know is her age at death—65—and that she had no known neurological diseases. Now, after 10 years of painstaking study of this woman's brain, scientists present her parting gift to science: BigBrain, the first 3D digital atlas that reveals the organ in microscopic detail.
For more than a century, neuroscientists have largely relied on 2D anatomical drawings of the brain based on work completed in the early 1900s, says lead author Katrin Amunts, a neuroanatomist at the Research Centre Jülich in Germany. Although some digitized, 3D reference models are available, they don't have the resolution needed to show brain tissue at the neuronal level, and they often have gaps, adds Joseph Masdeu, a neuroimaging expert at the National Institutes of Health in Bethesda, Maryland, who wasn't involved in the research. He compares previous atlases to an understocked library because too often "the book you need is not there."
Dissecting and then digitally reconstructing a human brain is a long and arduous process. In recent decades, neuroscientists' interest has shifted from classical anatomy to physiological studies, Amunts says, so few labs still have the expertise or tools to make, scan, and digitize the hair-thin slices of brain tissue required to capture details at the cellular level. Thicker slices mean fewer details and bigger gaps of data between tissue slabs, Amunts says.
So in 2003, Amunts and colleagues began the BigBrain atlas. The researchers selected the brain of the 65-year-old female donor as the basis for the atlas because it had no obvious signs of degenerative disease or other damage. After preserving the brain in formalin, a chemical fixative, and embedding it in wax for several months, they began to cut it using a knife called a microtome, which carves linked sections onto a conveyer belt like a deli slicer cutting turkey breast. They had to be very careful that slices didn't fall off of the belt during this stage, Amunts says. "Someone opens a door, and whoosh!"
Next, the researchers mounted each slice on a microscope slide, stained it to make the cell bodies of neurons visible, and scanned it. Overall, it took about 1000 hours of nearly continuous labor to prepare and scan each brain slice and the researchers obtained more than 7400 slices in all, they report online today in Science. Even taking lunch breaks could disrupt the delicate task, Amunts says.
Although the researchers took the utmost care in handling the fragile slices, inevitable warps and tears occur when brain tissue is stained and cut, says co-author Alan Evans, a neuroimaging expert at McGill University in Montreal, Canada. "Think of over 7000 sections of Saran Wrap that have been ripped and distorted," he says. Evans's job was to reconstruct the brain into a coherent whole, correcting errors by manually shifting misaligned scans back into place and using corrective software. Processing the quantity of data involved in BigBrain required a high-performance computing grid distributed across Canada.
At 50 times the spatial resolution of previous models, BigBrain shows individual neurons and the connections between them, an unprecedented level of detail for a whole-brain model. "What this brain does is give us nearly all the neurons in a single space" Masdeu says. "It's a wonderful resource." The finished product, which is part of the European Human Brain Project, a €1 billion effort to make a computer model of human brain function over the next 10 years, will soon be available for free through a web portal called CBRAIN. With mere clicks of the mouse, Masdeu says, researchers will be able to combine data from living brains, such as brain activity scans, with detailed anatomical information at the cellular level.
The BigBrain approach isn't perfect. In addition to the errors introduced by slicing, BigBrain doesn't capture the considerable variability between individual brains, notes neuroscientist John Mazziotta of the University of California, Los Angeles, who wasn't involved in the research. One solution to that problem might be to combine BigBrain with data gleaned from vast numbers of lower-resolution brain scans that show where different brain structures are likely to be, he says.
A new method that makes brain tissue transparent, called CLARITY, holds promise for future brain-imaging studies because researchers would not need to cut the brain, but at present it can be applied to only small amounts of tissue, equivalent to a mouse brain, Mazziotta says.
Next, Amunts wants to create BigBrain atlases of a male brain and a younger person's brain to capture potential sex and developmental differences. Now that the first map is complete, Mazziotta predicts that the process will get easier and faster over time, allowing researchers to look at the brains of large groups of people with disorders such as autism on a molecular level. "Eventually we'll have populations of brains done this way," he says.