Cerebrovascular reactivity (CVR) can be mapped using BOLD fMRI to provide a clinical insight into vascular health that can be used to diagnose cerebrovascular disease. Breath-holds are a readily accessible method for producing the required arterial CO2 increases but their implementation into clinical studies is limited by concerns that patients will demonstrate highly variable performance of breath-hold challenges. This study assesses the repeatability of CVR measurements despite poor task performance, to determine if and how robust results could be achieved with breath-holds in patients. Twelve healthy volunteers were scanned at 3 T. Six functional scans were acquired, each consisting of 6 breath-hold challenges (10, 15, or 20 s duration) interleaved with periods of paced breathing. These scans simulated the varying breath-hold consistency and ability levels that may occur in patient data. Uniform ramps, time-scaled ramps, and end-tidal CO2 data were used as regressors in a general linear model in order to measure CVR at the grey matter, regional, and voxelwise level. The intraclass correlation coefficient (ICC) quantified the repeatability of the CVR measurement for each breath-hold regressor type and scale of interest across the variable task performances. The ramp regressors did not fully account for variability in breath-hold performance and did not achieve acceptable repeatability (ICC0.4). Further analysis of intra-subject CVR variability across the brain (ICCspatial and voxelwise correlation) supported the use of end-tidal CO2 data to extract robust whole-brain CVR maps, despite variability in breath-hold performance. We conclude that the incorporation of end-tidal CO2 monitoring into scanning enables robust, repeatable measurement of CVR that makes breath-hold challenges suitable for routine clinical practice.

Noise correction is a critical step towards accurate mapping of resting state BOLD fMRI connectivity. Noise sources related to head motion or physiology are typically modelled by nuisance regressors, and a generalised linear model is applied to regress out the associated signal variance. In this study, we use independent component analysis (ICA) to characterise the data variance typically discarded in this pre-processing stage in a cohort of 12 healthy volunteers. The signal variance removed by 24, 12, 6, or only 3 head motion parameters demonstrated network structure typically associated with functional connectivity, and certain networks were discernable in the variance extracted by as few as 2 physiologic regressors. Simulated nuisance regressors, unrelated to the true data noise, also removed variance with network structure, indicating that any group of regressors that randomly sample variance may remove highly structured "signal" as well as "noise." Furthermore, to support this we demonstrate that random sampling of the original data variance continues to exhibit robust network structure, even when as few as 10% of the original volumes are considered. Finally, we examine the diminishing returns of increasing the number of nuisance regressors used in pre-processing, showing that excessive use of motion regressors may do little better than chance in removing variance within a functional network. It remains an open challenge to understand the balance between the benefits and confounds of noise correction using nuisance regressors.

In resting state fMRI, it is necessary to remove signal variance associated with noise sources, leaving cleaned fMRI time-series that more accurately reflect the underlying intrinsic brain fluctuations of interest. This is commonly achieved through nuisance regression, in which the fit is calculated of a noise model of head motion and physiological processes to the fMRI data in a General Linear Model, and the "cleaned" residuals of this fit are used in further analysis. We examine the statistical assumptions and requirements of the General Linear Model, and whether these are met during nuisance regression of resting state fMRI data. Using toy examples and real data we show how pre-whitening, temporal filtering and temporal shifting of regressors impact model fit. Based on our own observations, existing literature, and statistical theory, we make the following recommendations when employing nuisance regression: pre-whitening should be applied to achieve valid statistical inference of the noise model fit parameters; temporal filtering should be incorporated into the noise model to best account for changes in degrees of freedom; temporal shifting of regressors, although merited, should be achieved via optimisation and validation of a single temporal shift. We encourage all readers to make simple, practical changes to their fMRI denoising pipeline, and to regularly assess the appropriateness of the noise model used. By negotiating the potential pitfalls described in this paper, and by clearly reporting the details of nuisance regression in future manuscripts, we hope that the field will achieve more accurate and precise noise models for cleaning the resting state fMRI time-series.

We present the first evidence for vascular regulation driving fMRI signals in specific functional brain networks. Using concurrent neuronal and vascular stimuli, we collected 30 BOLD fMRI datasets in 10 healthy individuals: a working memory task, flashing checkerboard stimulus, and CO₂ inhalation challenge were delivered in concurrent but orthogonal paradigms. The resulting imaging data were averaged together and decomposed using independent component analysis, and three "neuronal networks" were identified as demonstrating maximum temporal correlation with the neuronal stimulus paradigms: Default Mode Network, Task Positive Network, and Visual Network. For each of these, we observed a second network component with high spatial overlap. Using dual regression in the original 30 datasets, we extracted the time-series associated with these network pairs and calculated the percent of variance explained by the neuronal or vascular stimuli using a normalized R² parameter. In each pairing, one network was dominated by the appropriate neuronal stimulus, and the other was dominated by the vascular stimulus as represented by the end-tidal CO₂ time-series recorded in each scan. We acquired a second dataset in 8 of the original participants, where no CO₂ challenge was delivered and CO₂ levels fluctuated naturally with breathing variations. Although splitting of functional networks was not robust in these data, performing dual regression with the network maps from the original analysis in this new dataset successfully replicated our observations. Thus, in addition to responding to localized metabolic changes, the brain's vasculature may be regulated in a coordinated manner that mimics (and potentially supports) specific functional brain networks. Multi-modal imaging and advances in fMRI acquisition and analysis could facilitate further study of the dual nature of functional brain networks. It will be critical to understand network-specific vascular function, and the behavior of a coupled vascular-neural network, in future studies of brain pathology.