Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder that often persists into adulthood. One hallmark in the characterization of pathological processing in ADHD is that attention skills are not impaired per se but more inconsistent and with higher variability compared to typically developing children (TDC). Increased variability in ADHD patients has been found in reaction times, as well as resting-state fMRI (rs-fMRI) brain signals. High variability has been assumed to reflect occasional lapses in attention, linked to intrusions of distracting activity during task performance and/or reduced anti-correlation between regions in the DMN and attention networks. Therefore, Dajani et al. (2019) concluded that it is more likely the dynamics between and within neural networks [i.e., the variability of network processing across time- and frequency-scales], that are affected in ADHD, than functional connectivity [in terms of one coefficient describing the (averaged) correlation between two regions over time]. 

We determined wavelet variance to quantify these dynamics. We determined wVar at rest and under task in fMRI timeseries of regions of the DMN and the FPN in three different frequency bands: 0.02 to 0.04Hz, 0.04 to 0.08Hz, and 0.08-0.16Hz.

We found that wVar differed group specifically between rest and task (significant group X condition interaction: whereas wVar was higher at rest compared to task in TDC, wVar was comparable or even decreased at rest in ADHD. For an external validation of group comparisons in wVar at rest, we determined wVar in rs-fMRI timeseries of a subsample of the Child Mind Institute data set (Functional Connectomes Project International Neuroimaging Data-Sharing Initiative http://dx.doi.org/10.15387/CMI_HBN (2017)). Results replicated our findings in terms of no significant group differences in wVar at rest in combination with similar or lower absolute values in ADHD patients compared to control subjects. 

In normal processing, high wVar at rest was interpreted as reflecting free fluctuating brain signaling, in comparison to small wVar under task indicating focussed processing. Thus, we conclude that wVar is a sensitive measure of cognitive processing and is even capable of detecting deviant processing in pathological brain function.

MRI data acquisition: Scanning was performed on a 3 Tesla TIM Trio Scanner at the Institute for Diagnostical and Interventional Neuroradiology at the University Hospital Wuerzburg and on a 3 Tesla PRISMA Scanner (Siemens, Erlangen, Germany) at the Department of Diagnostic and Interventional Radiology at the University Hospital Duesseldorf. Whole-brain T2*-weighted BOLD images were recorded with a simultaneous multi-slice echo-planar imaging sequence (repetition time=800ms, echo time=37ms, 72 slices, 2mm thickness, flip angle=52°, rs-fMRI: 6:58min, 512volumes, task-fMRI: 14.5min, 1069 volumes).

We determined wavelet variance (wVar) at rest and under task in fMRI timeseries of the DMN and the FPN in three different frequency bands: 0.02 to 0.04Hz, 0.04 to 0.08Hz, and 0.08-0.16Hz in children and adolescents with and without ADHD. WVar was determined from fMRI timeseries in regions defined from CONN’s ICA analyses of the HCP dataset (497 subjects) including the default mode network (DMN) and the fronto-parietal attention network (FPN) using wavelet transform from a signal filtering perspective, and the definition of wVar quantity as introduced by Percival (1995).

Additionally, for an external validation, wVar in rs-fMRI timeseries of a subsample of the Child Mind Institute data set (Functional Connectomes Project International Neuroimaging Data-Sharing Initiative http://dx.doi.org/10.15387/CMI_HBN (2017)) was determined.