The situation or the person? Individual and task‐evoked differences in BOLD activity

Abstract

Investigations of between‐person variability are enjoying a recent resurgence in functional magnetic resonance imaging (fMRI) research. Several recent studies have found persistent between‐person differences in blood‐oxygenated‐level dependent (BOLD) activation patterns and resting‐state functional connectivity. Conflicting findings have been reported regarding the extent to which (a) between‐person or (b) within‐person cognitive state differences explain differences in BOLD activation patterns. These discrepancies may arise due to statistical analysis choices, parcellation resolution, and limited sampling of task‐fMRI datasets. We attempt to address these issues in a large‐scale analysis of several task‐fMRI paradigms. Using a novel application of multivariate distance matrix regression, we examine between‐person and task‐condition variability estimates across varying levels of “resolution”, from a coarse region‐of‐interest level to the vertex‐level, and across different distance metrics. These analyses revealed that under most circumstances, differences in task conditions explained a greater amount of variance in activation map differences than between‐person differences. However, this finding was reversed when comparing activation maps at a “high‐resolution” vertex level. More generally, we observed that when moving from “low” to “high” resolutions, the variance explained by between‐person differences increased while variance explained by task conditions decreased. We further analyzed the relationships among subject‐level activation maps across all task‐conditions using an unsupervised clustering approach and identified a superordinate task structure. This structure went beyond conventional task labels and highlighted those experimental manipulations across task conditions that produce contrasting versus similar whole‐brain activation patterns. Overall, these analyses suggest that the question of the subject‐ versus task‐effects on BOLD activation patterns is nontrivial, and depends on the comparison “resolution,” choice of distance metric, and the coding of task‐conditions.

Intrinsic organization of cortical networks predicts state anxiety: an functional near-infrared spectroscopy (fNIRS) study

“…The connectivity-based predictive model characterizes individual-specific trait profiles in a way as fingerprint 59 , which has been widely applied to cognitive functions 24,26 , personality trait 28,60 , and clinical disorders 61 . However, extensive evidence also shows indispensable contributions of state-dependent functional architecture to Individualized brain organization 62,63 . Brain state manipulations have been shown to improve the prediction of individual traits by amplifying individual variations in specific state 64 .…”

Individual differences in delay discounting are associated with dorsal prefrontal cortex connectivity in children, adolescents, and adults

“…We conducted a connectome-wide analysis using MDMR as described in detail in previous studies ( Shehzad et al, 2014 , Satterthwaite et al, 2015 , Sharma et al, 2017 ) with the goal of understanding how the temporal discounting rate was associated with individual differences in the multivariate pattern of FC at each voxel. Matrix regression is a class of statistical methods that have been widely used in genomics and ecology research ( Minas and Montana, 2014 , Zakharov et al, 2013 ) and have been increasingly used for analysis of high-dimensional imaging data ( Chen et al, 2022 , Kim et al, 2022 , Bolt et al, 2019 ). In this approach, the outcome of the regression equation is not a single vector of values, but rather a distance matrix between participants (see Fig.…”

Improving fMRI-based prediction of individual pain ratings by accounting for interindividual variance

“…Our results allow for a more optimistic conclusion, demonstrating evidence for predicting interindividual differences in pain perception. One possible explanation for the absence of a relationship between individual pain rating and the corresponding BOLD readout is the between-person variability of BOLD activity, which has been explored in previous studies [2][3][4] . To address this issue, several solutions have been proposed.…”