Each person has a unique pattern of white matter connecting different brain regions, a new study suggests.
By Nathan Collins
The human brain is a mass of something like 86 billion neurons, nearly always organized into the same basic structure. But just as nearly all of us have 10 fingers featuring different patterns of swirls on their tips, a new study suggests there are identifiable differences inside our heads. Our brains, in other words, have fingerprints.
Those differences show up in white matter or axons, the long arms of nerve cells that transmit signals from one cell body to the next. To map those connections, Fang-Cheng Yeh, Timothy Verstynen, and their colleagues used a technique called diffusion magnetic resonance imaging (dMRI), a variation on conventional MRI scans. In fact, all MRI really does is detect water and, therefore, blood. dMRI adds the ability to track blood flow and density, with which neuroscientists can construct detailed white matter maps known as connectomes.
But are our connectomes unique? To answer that, the researchers first looked at dMRI density data from 93 people who’d been scanned several times over the course of anywhere from a single day to nearly a year. Those scans showed very few signs of change over time, which meant that the connectome could, in principle, act like a brain fingerprint.
Next, the team expanded the data set to include scans from 120 more people, for a total of 213, then looked at how each person’s white-matter map differed from the next. The differences the researchers found were big enough, they write, that they could almost always tell whether two scans came from the same person at different times or from two different people.
That’s not to say the method is perfect: The team estimates that, for two scans taken from the same person six months or a year apart, they’d make a mistake—that is, they’d identify the scans as belonging to different people—about once out of every 180 such comparisons. That’s possibly because there’s just enough change over time to make an ID difficult. For shorter timeframes, however, the error rate was around one in 200,000 or less.