A key challenge in the effort to link brain activity with behavior is that brain activity, measured by functional magnetic resonance imaging (fMRI), for instance, is extraordinarily complex. That complexity can make it difficult to find recurring activity patterns across different people or within individuals.
In a new study, Yale researchers were able to take fMRI data, reduce its complexity, and in doing so, uncover stable patterns of activity shared across more than 300 different people. The findings, researchers say, are a promising step forward in uncovering biomarkers for psychiatric disorders.
The study was published Sept. 24 in the journal PLOS Biology.
"Because human brain activity is so complex, it can be unreliable, particularly when you're aiming for reproducibility," said Kangjoo Lee, lead author of the study and an associate research scientist in the Department of Psychiatry at Yale School of Medicine. "In this study, we wanted to capture features of brain activity that were linked to features of human behavior and were also consistent across different people."
Researchers revealed brain patterns by using an approach similar to one that represents the complex patterns of a dance sequence through a small number of basic movements.
To do this, the researchers used fMRI data from 337 healthy young adults, each of whom underwent four 15-minute fMRI scans.
"These images, which were taken during a resting state, were essentially snapshots of brain activity," said Lee. "So over time, we were able to observe moment-to-moment changes in brain activity."
Each snapshot represented activity occurring throughout the brain in a particular moment, which included many different brain networks engaged in many different processes, contributing to the data's complexity.
To reveal shared patterns in the data, the researchers applied an approach called data dimension reduction, which, Lee says, essentially takes high-dimensional, complex data like brain activity and reduces it to a lower dimensional space. The idea is similar to representing the complex pattern of a dance sequence through a small number of basic movements.
After reducing the complexity of the data, the researchers uncovered three shared patterns of brain activity that were "highly recurring across participants and within participants," Lee said.
Further, while these patterns were found among all participants, there were also differences among the individuals. For instance, the researchers observed differences in terms of which of the three states different individuals spent the most time, how long they lingered in particular states, and which states people transitioned between. The findings suggest these types of patterns could reveal information about something shared across different people - such as a behavior - as well as individual differences related to those behaviors or how they change over time.
The researchers are now considering how this approach could be applied to psychiatric disorders.
"Here we looked at healthy adults, but if we ran a similar analysis in a clinical population, we may find recurring brain patterns that are shared among that population but not among healthy individuals," said co-senior author John Murray, a former professor of psychiatry and physics at Yale, now a professor at Dartmouth College. "Therefore, these shared patterns could represent biomarkers of psychiatric illness that are useful in clinical settings."
The current study lends support to that idea. The researchers found that the patterns in which people spent more time and the patterns that they transitioned to was associated with cognitive function, emotion regulation, and alcohol and substance use.
"Uncovering recurring brain patterns in clinical populations could tell us something about the neural activity associated with specific symptoms and how it differs between individuals," said Lee.
