Spatial variation of changes in test–retest reliability of functional connectivity after global signal regression: The effect of considering hemodynamic delay

Abstract

Global signal regression (GSR) is a controversial analysis method, since its removal of signal has been observed to reduce the reliability of functional connectivity estimates. Here, we used test–retest reliability to characterize potential differences in spatial patterns between conventional, static GSR (sGSR) and a novel dynamic form of GSR (dGSR). In contrast with sGSR, dGSR models the global signal at a time delay to correct for blood arrival time. Thus, dGSR accounts for greater variation in global signal, removes blood‐flow‐related nuisance signal, and leaves higher quality neuronal signal remaining. We used intraclass correlation coefficients (ICCs) to estimate the reliability of functional connectivity in 462 healthy controls from the Human Connectome Project. We tested across two factors: denoising method used (control, sGSR, and dGSR), and interacquisition interval (between days, or within session while varying phase encoding direction). Reliability was estimated regionally to identify topographic patterns for each condition. sGSR and dGSR provided global reductions in reliability compared with the non‐GSR control. Test–retest reliability was highest in the frontoparietal and default mode regions, and lowest in sensorimotor cortex for all conditions. dGSR provides more effective denoising in regions where both strategies greatly reduce reliability. Both GSR methods substantially reduced test–retest reliability, which was most evident in brain regions that had low reliability prior to denoising. These findings suggest that reliability of interregional correlation is likely inflated by the global signal, which is thought to primarily reflect dynamic blood flow.

Brain-wide functional connectivity artifactually inflates throughout fMRI scans: a problem and solution

“…Spatial ICA denoising methods like ICA-AROMA and ICA-FIX are not well suited to capturing noise signals that appear in different brain regions at different times, and thus the sLFO signal is one that would evade detection and removal by these denoising methods. Moreover, spatial maps specifying the contribution of the sLFO to BOLD signal variance in different areas of the brain align with the spatial distribution of FC inflation observed here, with peaks in the occipital lobe, sensory-motor strip and superior temporal lobe 32,35,36 . The sLFO signal would thus satisfy the constraints specified above.…”

“…S4c). This is consistent with prior reports showing that while physiological denoising methods increase the validity of fMRI data and increase betweenscan reproducibility, they tend to decrease subject discriminability 39,45 via the removal of ICCinflating noise. The relative insensitivity of between-scan reproducibility and subject discriminability to FC inflation can be understood when considering that these metrics use timeaveraged connectivity values as inputs, which would obscure the impact of temporal changes in connectivity that happen within a scan.…”

“…As such, we investigated the temporal evolution of sLFO signal strength during fMRI scanning in the HCP resting-state acquisitions. To isolate and measure the sLFO signal, we used Regressor Interpolation at Progressive Time Delays (RIPTiDe) as implemented by the software package rapidtide 37,38 , which has been used extensively to both characterize and remove physiologically based sLFO signals 32,39-44…”

“…As such, we investigated the temporal evolution of sLFO signal strength during fMRI scanning in the HCP resting-state acquisitions. To isolate and measure the sLFO signal, we used Regressor Interpolation at Progressive Time Delays (RIPTiDe) as implemented by the software package rapidtide 37,38 , which has been used extensively to both characterize and remove physiologically based sLFO signals 32,[39][40][41][42][43][44] The RIPTiDe analysis revealed that the proportion of BOLD signal variance attributable to sLFO noise increased monotonically with time within both the HCP's first and second resting state acquisitions of a given scanning session, and also increased across these acquisitions (Fig. 5a).…”

Aberrant flexibility of dynamic brain network in patients with autism spectrum disorder

“…Notably, the ABIDE database provided both the preprocessed data with and without the step of global signal regression (GSR). In the current study, we used the data without GSR since the usage of GSR is still controversial (42,43). Furthermore, to control for confounding effects caused by differences in data acquisition and preprocessing procedures between different site, we included "site" as covariates (using dummy coding) in all analyses (see details in the followed sections).…”