We all feel unique, believing that our inner lives and our physical selves are different somehow than those of others. Various methods for identifying an individual confirm our intuitions, whether through inspecting the whorls of a fingerprint or by sequencing a strand of DNA.
Scientists also look to the source of that feeling of being special. But a common form of brain imaging used over the past few decades—functional magnetic resonance imaging (fMRI)—has been unable to provide the desired specificity to derive an individual “brainprint.” Most studies examine differences between the brains of entire groups who differ in some important way. Researchers compare the average readings for certain brain areas in a group diagnosed with schizophrenia, for example, with the averages in a healthy control group.
Neuroscience has really not had much choice. Brain-imaging technologies are actually fairly crude tools; fMRI is a technique that measures changes in levels of oxygen in the blood flowing inside the brain. It indicates which parts of the brain work hardest by looking at the areas that demand more oxygen in order to fuel metabolism. Changes in oxygen levels occur slowly compared with the rapid-fire pace of the electrical impulses zipping throughout the brain. What’s more, the signals that the scanner receives are noisy because of the subject’s respiration and heartbeat. As well as being noisy, the anatomical and physical details can also vary greatly from person to person. So imaging studies usually average their results from the scans of many people to uncover meaningful information about how brains work.
But this still begs the question: Just how variable are brains really? Variable enough to extract useful information about an individual from a single scan? Are they sufficiently distinct to identify that individual? Are they, in fact, unique? And could this variation contain useful information about how and why we differ in terms of cognitive performance, or mental health?
A study published online this week in Nature Neuroscience takes a stab at answering some of these questions. The researchers, led by Emily Finn and Xilin Shen of Yale University used a specialized form of fMRI called functional connectivity MRI (fcMRI) that maps how similar the activity is in different parts of the brain. The researchers deployed fcMRI to identify a given individual with an accuracy of up to 99 percent from a group of 126 healthy young adults. The scans were gathered as part of the Human Connectome Project (HCP), a U.S. collaboration led by Washington University in Saint Louis and the University of Minnesota that aims to map the wiring of the human brain. (Scientific American is part of Nature Publishing Group.)
The technique involves taking tiny fMRI imaging elements called voxels whose activity is largely synchronized and grouping them into what are called nodes. Researchers then compare how well the average activity in one node is synchronized with that of another—a measure of how well connected the two areas are.
In this study, the investigators created a network of 268 nodes and assessed how connected each node was to all the other nodes. The end result produced a "connectivity profile" of 35,778 connections. This enabled the researchers to identify individuals by comparing a pattern of neural connections from a scan for an individual in one session to scans of all the study participants in a second session. The scans were taken when the participants were at rest or engaged in one of several tasks. When comparing a rest scan for an individual against a second round of scans for the whole group, the technique identified the correct person 93 to 94 percent of the time. When comparing a rest scan with an activity scan or considering two activity tasks, performance was lower but still ranged between 54 to 87 percent, suggesting that the pattern of connections in an individual's brain remains distinctive even as she engages in different activities. “What we've shown is that the same brain doing two different things looks more similar than two different brains doing the same thing,” Finn says.
The researchers also performed another level of analysis in which they found certain groups of nodes that correspond to well-established brain networks dedicated to visual, motor or other tasks. They then asked whether some of these networks reveal more about a brain’s distinctiveness than others. One performed best—the frontoparietal network, involved with controlling attention and other cognitive functions. It identified an individual 98 to 99 percent of the time during rest and 80 to 90 percent for most other comparisons.
Finally, the team showed a relationship between the participants' fluid intelligence (Gf, a measure of reasoning and problem-solving in novel situations) and brain connectivity. This link was not strong enough to predict participants' Gf to a level that could be used by a psychologist, but that's not surprising—fluid intelligence is likely determined by many different factors.
The same frontoparietal network that was most useful in identifying people was also most strongly linked to Gf. The nodes of this network control the switching of connections to cope with changing tasks. “Those are the most recently evolved and sophisticated regions, involved in the higher-order functions we're so interested in, like attention, memory and language.” Finn says. This finding supports a number of ideas about what makes us unique. These areas may be more shaped by experience whereas sensory and motor networks are more hardwired. “We all can see the rock falling and get out of the way,” says cognitive neuroscientist Michael Gazzaniga of the University of California, Santa Barbara, who was not part of this study. “But some of us are better at figuring out why it fell in the first place.”
The authors don't suggest actually using these techniques to identify people. “We don't need to put people in a scanner to know who they are,” Finn says. “We can identify people by looking at them or fingerprinting them.” Co-author Shen agrees: “It's just a proof of principle to show there's sufficient information in these scans to tell the difference between people.” This has important implications for developing clinically useful applications, however. “Starting to focus on individuals is a fantastic idea, that's really going to be the future,” says neuroimaging expert Cameron Craddock of the Child Mind Institute and the Nathan S. Kline Institute for Psychiatric Research, who was not involved with the research. “It could be a fingerprint for mental health or treatment outcomes,” he says. “That's where the big payoff will be.” It could even be useful for estimating who might respond best to educational programmes.
The researchers also do not advocate using the technique to estimate intelligence. Finn says she does not think brain imaging will ever fully replace an IQ test or other behavioral measures that are easier, cheaper and more accurate. “But for things we can't tell just by looking at people or giving them a test—like who might develop Alzheimer's in a few years or respond to an intervention—those are the types of things this could be useful for.” The team has already started to work with data from a group of adolescents at high risk for schizophrenia to see if they can predict who ends up getting the full-blown illness. “Those types of predictions are what I'm interested in going after next,” Finn says.