An international team of researchers has developed new analytical techniques that consider interactions between three or more regions of the brain – providing a more in-depth understanding of human brain activity than conventional analysis. Led by Andrea Santoro at the Neuro-X Institute in Geneva and Enrico Amico at the UK’s University of Birmingham, the team hopes its results could help neurologists identify a vast array of new patterns in human brain data.
To study the structure and function of the brain, researchers often rely on network models. In these, nodes represent specific groups of neurons in the brain and edges represent the electrical connections between neurons using statistical correlations.
Within these models, brain activity has often been represented as pairwise interactions between two specific regions. Yet as the latest advances in neurology have clearly shown, the real picture is far more complex.
“To better analyse how our brains work, we need to look at how several areas interact at the same time,” Santoro explains. “Just as multiple weather factors – like temperature, humidity, and atmospheric pressure – combine to create complex patterns, looking at how groups of brain regions work together can reveal a richer picture of brain function.”
Higher-order interactions
Yet with the mathematical techniques applied in previous studies, researchers have not confirmed whether network models incorporating these higher-order interactions between three or more brain regions could really be more accurate than simpler models, which only account for pairwise interactions.
To shed new light on this question, Santoro’s team built upon their previous analysis of functional MRI (fMRI) data, which identify brain activity by measuring changes in blood flow.
Their approach combined two powerful tools. One is topological data analysis. This identifies patterns within complex datasets like fMRI, where each data point depends on a large number of interconnected variables. The other is time series analysis, which is used to identify patterns in brain activity which emerge over time. Together, these tools allowed the researchers to identify complex patterns of activity occurring across three or more brain regions simultaneously.
To test their approach, the team applied it to fMRI data taken from 100 healthy participants in the Human Connectome Project. “By applying these tools to brain scan data, we were able to detect when multiple regions of the brain were interacting at the same time, rather than only looking at pairs of brain regions,” Santoro explains. “This approach let us uncover patterns that might otherwise stay hidden, giving us a clearer view of how the brain’s complex network operates as a whole.”
Just as they hoped, this analysis of higher-order interactions provided far deeper insights into the participants’ brain activity compared with traditional pairwise methods. “Specifically, we were better able to figure out what type of task a person was performing, and even uniquely identify them based on the patterns of their brain activity,” Santoro continues.
Distinguishing between tasks
With its combination of topological and time series analysis, the team’s method could distinguish between a wide variety of tasks in the participants: including their expression of emotion, use of language, and social interactions.
By building further on their approach, Santoro and colleagues are hopeful it could eventually be used to uncover a vast space of as-yet unexplored patterns within human brain data.
By tailoring the approach to the brains of individual patients, this could ultimately enable researchers to draw direct links between brain activity and physical actions.
“Down the road, the same approach might help us detect subtle brain changes that occur in conditions like Alzheimer’s disease – possibly before symptoms become obvious – and could guide better therapies and earlier interventions,” Santoro predicts.
The research is described in Nature Communications.
The post Higher-order brain function revealed by new analysis of fMRI data appeared first on Physics World.
