Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Big Data Functional Interactions in Human Brain


Science, November 2013: Vol. 342 no. 6158 pp. 580-584

Functional Interactions as Big Data in the Human Brain

Nicholas B. Turk-Browne

Department of Psychology and Princeton Neuroscience Institute, Princeton University, Princeton, NJ, 08540, USA.


Noninvasive studies of human brain function hold great potential to unlock mysteries of the human mind. The complexity of data generated by such studies, however, has prompted various simplifying assumptions during analysis. Although this has enabled considerable progress, our current understanding is partly contingent upon these assumptions. An emerging approach embraces the complexity, accounting for the fact that neural representations are widely distributed, neural processes involve interactions between regions, interactions vary by cognitive state, and the space of interactions is massive. Because what you see depends on how you look, such unbiased approaches provide the greatest flexibility for discovery.

Understanding how the brain works could help uncover the fundamental principles of cognition and behavior.

The development of magnetic resonance imaging (MRI) began a new era in cognitive neuroscience. Exploiting differences in magnetic susceptibility between oxygenated and deoxygenated blood [blood oxygenation level–dependent (BOLD) contrast], functional MRI (fMRI) detects metabolic activity, and by inference, neuronal activity, noninvasively throughout the brain. This technique generates complex data sets: ~100,000 locations, measured simultaneously hundreds of times, resulting in billions of pairwise relations, collected in multiple experimental conditions, and from dozens of participants per study. With this powerful technology in widespread use, data analysis has become the bottleneck for progress. What is the best way to find the mind in brain data?

This review is organized around four desiderata for examining the mind with fMRI, each embracing a different aspect of the nature and complexity of human brain function: (i) neural representations are widely distributed within and across brain regions, (ii) neural processes depend on dynamic interactions between regions, (iii) these interactions vary systematically by cognitive state, and (iv) the space of possible interactions has high dimensionality. All four complexities can be accounted for by harnessing recent advances in large-scale computing. Such unbiased approaches are beginning to reveal how disparate parts of the brain work in concert to orchestrate the mind.

In fMRI, brain activity is not measured at the level of regions but rather in terms of volumetric pixels (voxels). The average amplitude of BOLD activity evoked by trials relative to baseline (“activation”) identifies voxels that are responsive to the function engaged by that trial type.

Multivariate pattern analysis (MVPA) relies on tools from machine learning to decode patterns of activation across voxels. One of the first discoveries enabled by MVPA was that information about a category is present throughout visual cortex, beyond voxels with the strongest activation to that category.

How can we hold vivid images in our mind’s eye? Frontal and parietal regions that help maintain information in working memory lack detailed visual selectivity, and visual areas with the needed selectivity show little delay-period activation in working memory tasks. Despite this weak activation, however, MVPA of visual cortex can successfully decode what information is being held in mind --revealing that sensory machinery is recruited for working memory.

The advent of MVPA eliminated a bias to interpret brain regions as having homogeneous and discrete functions. This approach helped capture another core aspect of brain function: Regions do not work in isolation, with computation depending on local and long-range interactions. This can be reflected in fMRI coactivation: Voxels containing interacting neurons are more likely to activate together, which could produce distributed patterns visible to MVPA.

However, a limitation of most uses of MVPA is that they focus on (patterns of) activation and are thus blind to certain kinds of interactions. Voxels need not vary in activation to have selectivity: Neuronal populations may generally be active, with their function defined on the basis of which specific neurons are communicating with each other.

The most common application of functional connectivity is examining intrinsic correlations while participants rest, typically by modeling whole-brain BOLD activity with the time course from a seed region. This approach has helped characterize the functional architecture of the brain, namely, how regions group together into broader systems. One such system is the “default network,” a set of regions that are robustly correlated at rest.

A brain with N = 50,000 voxels contains N(N – 1)/2 = 1,249,975,000 unique voxel pairs,

The full correlation matrix can be represented as a six-dimensional (6-D) autocorrelation field: For each voxel in the 3-D brain, there is a 3-D brain of functional connectivity with every other voxel. Computing all pairwise correlations was prohibitively slow in the past—up to hours or days. Matrix multiplication can be used for drastically improved computational speed: If each voxel’s time course is mean-centered and the result is divided by its root sum of squares, the Pearson correlation of any two voxels is reduced to the sum of pointwise products over time (the dot product), and the full matrix of coefficients is obtained by the product of a voxels-by-time matrix and its transpose. Technological advances can reduce such largematrix multiplication operations to less than 1 s.

Analysis of the full correlation matrix during rest has started yielding insights into the topology and dynamics of human brain networks. If each voxel is treated as a node, and all correlations between that and other nodes above some threshold are treated as edges, then the resulting binary matrix generates a graph. These voxelwise graphs can be characterized quantitatively with network measures, including degree, number of edges for a node; modularity, density of edges within versus between node clusters; path length, minimum number of edges between nodes; and centrality, proportion of shortest paths passing through a node.

In this lexicon, functional brain networks exhibit high modularity and short path lengths. High modularity reflects strong connections between nodes that contribute to the same function, such as in visual cortex, whereas short path lengths reflect connections between these node communities via “hub” nodes that have high centrality and tend to be connected to each other, such as in frontal cortex. These two properties fit the definition of a “small-world” network, an organizational scheme found in many biological and nonbiological complex systems that enables efficient information processing, both locally within modules and globally across the network.

Thinking of brain function as a small-world network has enabled progress on several fronts. For example, it was recently discovered that although voxelwise graphs from infants’ brains also have small-world properties, their cortical hubs are located in different places than adults—unexpectedly, in primary sensorimotor cortex. There is variation in network properties even among adults: Some brains have shorter path lengths, and these individuals score higher on an intelligence test. These studies suggest that investigating how information is integrated across the brain holds particular promise for understanding the origins and limits of cognition.

[end of paraphrase]


Return to — Modeling Brain Functionality of Consciousness