Blood oxygen level-dependent signal variability is more than just noise

Abstract

Functional magnetic resonance imaging (fMRI) research often attributes blood oxygen level-dependent (BOLD) signal variance to measurement-related confounds. However, what is typically considered "noise" variance in data may be a vital feature of brain function. We examined fMRI signal variability during fixation baseline periods, and then compared SD- and mean-based spatial patterns and their relations with chronological age (20-85 years). We found that not only was the SD-based pattern robust, it differed greatly, both spatially and statistically, from the mean-based pattern. Notably, the unique age-predictive power of the SD-based pattern was more than five times that of the mean-based pattern. This reliable SD-based pattern of activity highlights an important "signal" within what is often considered measurement-related "noise." We suggest that examination of BOLD signal variability may reveal a host of novel brain-related effects not previously considered in neuroimaging research.

Figures

Figure 1.

Conceptual comparison between fMRI signal mean and variability for a random brain voxel.

Figure 2.

Example result of block normalization on a single voxel time series (shown in gray), obtained by concatenating 16 fixation blocks from two randomly selected runs (horizontal black lines represent mean block levels). Our experiment contained 32 blocks, but 16 are shown here for descriptive purposes, both before (a) and after (b) block normalization.

Figure 3.

Zero-order relations between age and brain scores from SD- (a) and mean- (b) based analyses. We have deliberately used correlational analyses here, despite what appears to be an extreme group design. In initial model runs, we ran SD- and mean-based analyses with a dichotomous young/old variable, instead of a continuous measure of age. The use of dichotomous and continuous age variables yielded nearly identical results for each brain measure (within an R2 of 1.00%). Thus, we elected to maintain the use of continuous age to allow better visualization of scatter around lines of best fit.

Figure 4.

PLS brain patterns and overlay plots. 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, all robust areas surpassed a thresholded bootstrap ratio (salience/SE) of ≥3.00 (for yellow/red regions) or ≤−3.00 (for blue regions). Darker colors indicate greater robustness. c, Overlay plot highlighting differences between age-based SD- and mean-brain spatial patterns. Red, Mean increase, no SD effect; blue, mean decrease, no SD effect; green, SD increase, no mean effect; yellow, SD decrease, no mean effect. d, Overlay plot highlighting similarities between age-based SD- and mean-brain spatial patterns. Blue, mean and SD both decrease with age; green, mean decrease, SD increase. All images represent every other slice in z-direction.

Figure 5.

Relative contributions of SD- and mean-based brain measures for predicting chronological age. Values represent unique percentage variance accounted for (unique R2 × 100) in chronological age. “Shared” represents predictive overlap between mean- and SD-based measures; “Unknown” represents variance not accounted for by either mean- or SD-based measures. We found no interaction between the effects of mean and SD on age.