Building a Science of Individual Differences from fMRI

doi:10.1016/j.tics.2016.03.014

Trends in Cognitive Sciences

Volume 20, Issue 6, June 2016, Pages 425-443

https://doi.org/10.1016/j.tics.2016.03.014 Get rights and content

Trends

Interpretation of fMRI data at the level of individual brains is essential for characterizing brain function in health and disease.

Two core challenges are validity (do we measure what we intend to measure?) and reliability (is our measure stable in the face of variations that should not matter?) of fMRI-derived individual differences; these challenges can be partly addressed with recent tools.

Interpretation of single-subject fMRI measures relies on establishing a relationship with an independent measure in the same subjects. Out-of-sample prediction should be used over correlation analysis.

Accumulation of large samples through consortia and data sharing, as well as careful attention to statistical power issues, are crucial for reproducible research.

Whole-brain characterization in naturalistic conditions, such as while watching a movie or listening to a story, may provide an alternative to resting-state data that permits a rich link to sensory and semantic stimulus variables.

To date, fMRI research has been concerned primarily with evincing generic principles of brain function through averaging data from multiple subjects. Given rapid developments in both hardware and analysis tools, the field is now poised to study fMRI-derived measures in individual subjects, and to relate these to psychological traits or genetic variations. We discuss issues of validity, reliability and statistical assessment that arise when the focus shifts to individual subjects and that are applicable also to other imaging modalities. We emphasize that individual assessment of neural function with fMRI presents specific challenges and necessitates careful consideration of anatomical and vascular between-subject variability as well as sources of within-subject variability.

Section snippets

From the Group to the Individual

Brain imaging with blood oxygen level-dependent functional magnetic resonance imaging (BOLD fMRI) has been used extensively since the early 1990s to understand generic aspects of brain function, typically by averaging data across individuals to improve the signal-to-noise ratio (SNR). The statistical benefits of averaging across subjects have also been leveraged in group comparisons, for example in studies of clinical populations. However, these studies have historically fallen short of a

A Common Space for Mapping Function

How can we tell whether individual differences in metrics such as BOLD activations or functional connectivity are actually related to differences in the underlying neural activity or communication, respectively? One ubiquitous problem arises in matching different brains such that functionally meaningful comparisons across subjects are possible in the first place. In a typical fMRI analysis pipeline, both structural and functional data from individual brains are spatially warped to a common

Reliability: Individual Differences or Unmodeled Noise?

Once validity is maximized (inasmuch as current technology allows), it is also crucial to ensure that individual differences measured with fMRI are not merely attributable to unaccounted-for noise in the measurements. The reliability of fMRI has been inspected closely in recent years 75, 76, 77. Of course there is no such thing as ‘the’ reliability of fMRI because different derived statistics are differentially affected by noise in the raw data and thus have different reliabilities [78] (for an

Choosing Prediction over Correlation

Currently the only way to interpret a fMRI-derived statistic is to relate it to another individual measure in the same set of subjects, such as their age, gender, test scores indexing aspects of intelligence and personality, or other measures of behavior (Table S2). By far the majority of fMRI studies of individual differences use correlation analysis to establish such a relationship. Correlation analysis relies on in-sample population inference and does not directly ensure the generalizability

Concluding Remarks

Individual differences in brain function are key to understanding healthy differences based on personality, gender, age, or culture. They are also crucial for personalized medicine approaches to neuropsychiatric diseases. Recent technical advances have increased the sensitivity of functional MRI and set the stage for a characterization of brain activity at the level of momentary mental events in individual subjects. We now face key challenges of reliability and validity on the path to an

Acknowledgments

This work was funded in part by a NARSAD grant from the Brain and Behavior Research Foundation (to J.D.) and a Conte Center from the National Institute of Mental Health (NIMH) (to R.A.). The authors thank Rebecca Schwarzlose, Michael Miller, Alex Huth, Swaroop Guntupalli and two anonymous reviewers for useful comments on the manuscript.

Glossary

Cluster: a contiguous set of voxels or vertices whose value in a statistical parametric map exceeds the cluster-forming threshold.
Echo planar imaging: any rapid gradient-echo or spin-echo sequence in which k-space is traversed in one (single-shot) or a small number of excitations (multi-shot). Gradient-echo EPI is the workhorse of fMRI.
General linear model: a generalization of the multiple linear regression model to the case of more than one dependent variable. The GLM attempts to explain the BOLD