Understanding variability in the BOLD signal and why it matters for aging

Abstract

Recent work in neuroscience supports the idea that variability in brain function is necessary for optimal brain responsivity to a changing environment. In this review, we discuss a series of functional magnetic resonance imaging (fMRI) studies in younger and older adults to assess age-related differences in variability of the fMRI signal. This work shows that moment-to-moment brain signal variability represents an important "signal" within what is typically considered measurement-related "noise" in fMRI. This accumulation of evidence suggests that moving beyond the mean will provide a complementary window into aging-related neural processes.

Figure 1

PLS brain patterns from the first study (Garrett et al. 2010). (a) Yellow/red regions indicate robust age-related increases, and blue regions indicate age-related decreases, in BOLD SDs. (b) Yellow/red regions indicate robust age-related increases, and blue regions indicate age-related decreases, in BOLD means. In both (a) and (b), darker colors indicate greater robustness (i.e., a stronger contribution to the pattern of age differences). In this and subsequent figures, the threshold for coloured regions was set to a bootstrap ratio of ± 3, i.e. where the weight for each voxel was 3 times an estimate of its standard error.

Figure 2

PLS results from the second study (Garrett et al. 2011). (a) Blue regions indicate greater variability with younger age, and faster and more consistent RT performance. Yellow/red regions show the opposite pattern of an association between more brain variability and older age, and slower and less consistent RT performance. Darker colors indicate greater robustness. (b) The graph shows the correlations (Pearson r) between age, mean RT, and within-subject variability of RT (ISDRT) across tasks and pattern of BOLD SD seen in (a). PMT = perceptual matching. ATT = attentional cueing. DMS = delayed match-to-sample. Error bars represent bootstrapped 95% confidence intervals. (c) The graph shows levels of brain variability in robust blue and yellow/red regions seen in (a). “Fast” and “slow” refer to greater than −1 and greater than +1 SD from the sample mean RT across tasks; “Consistent” and “inconsistent” refer to greater than −1 and greater than +1 SD from the sample average ISDRT across tasks. Younger, faster, more consistent individuals exhibited considerably more variability across brain regions than older, slower, more inconsistent adults.

Figure 3

Brain scores (a) and brain regions showing increased SD from fixation to task (b) from the third study (Garrett et al. 2013). Brain scores (a) represent the degree to which young and older adults, as a group, show increased SD in the regions seen in (b). Error bars represent bootstrapped 95% confidence intervals. Non-overlapping confidence intervals indicate that the increase in SD from fixation to tasks is larger in the younger adults. In the brain regions seen in (b), darker colours (i.e., more red) indicate regions with a more robust increase in variability on task compared to fixation. Fix = fixation. PMT = perceptual matching. ATT = attentional cueing. DMS = delayed match-to-sample.

Figure 4

The data in Figure 3 were analyzed in each age group separately. Young adults (a) showed a robust increase in BOLD variability from fixation to tasks. Older adults showed a similar but more muted response (b) both in terms of magnitude of the increase and extent of brain regions showing the effect. Abbreviations and interpretation of the graphs are the same as in Figure 3.

Figure 5

Three dimensions of age- and performance-related group differences in dynamic range supported by the present studies. (a) Within-region, within-condition: example that captures group differences in moment-to-moment variability in signal. (b) Within-region, across conditions: example that captures group differences in signal variability as the brain transitions from fixation to task, or from one task to another. (c) Between-region, within task: example that captures group differences in the extent of region differentiation in SD level.