On spurious and real fluctuations of dynamic functional connectivity during rest

Abstract

Functional brain networks reconfigure spontaneously during rest. Such network dynamics can be studied by dynamic functional connectivity (dynFC); i.e., sliding-window correlations between regional brain activity. Key parameters-such as window length and cut-off frequencies for filtering-are not yet systematically studied. In this letter we provide the fundamental theory from signal processing to address these parameter choices when estimating and interpreting dynFC. We guide the reader through several illustrative cases, both simple analytical models and experimental fMRI BOLD data. First, we show how spurious fluctuations in dynFC can arise due to the estimation method when the window length is shorter than the largest wavelength present in both signals, even for deterministic signals with a fixed relationship. Second, we study how real fluctuations of dynFC can be explained using a frequency-based view, which is particularly instructive for signals with multiple frequency components such as fMRI BOLD, demonstrating that fluctuations in sliding-window correlation emerge by interaction between frequency components similar to the phenomenon of beat frequencies. We conclude with practical guidelines for the choice and impact of the window length.

Brain dynamics in ASD during movie‐watching show idiosyncratic functional integration and segregation

“…In the resting‐state literature, several approaches have been suggested for this purpose, from the computation of FC across separate subjects (Keilholz, Magnuson, Pan, Willis, & Thompson, 2013) to the generation of appropriate null data with disrupted dynamics (Betzel, Fukushima, He, Zuo, & Sporns, 2016; Leonardi et al, 2013). Here, to extract significant ISFC excursions, we combined two approaches: the high‐pass filtering of preprocessed regional time courses with the inverse of the window length (set here to 20 s), which greatly limits the impact of spurious FC fluctuations (Leonardi & Van De Ville, 2015; Zalesky & Breakspear, 2015), and the use, as null data to threshold the movie‐watching ISFC time courses under the null hypothesis of no movie‐driven FC changes, of resting‐state segments from the same subjects, with identical acquisition parameters.…”

“…For each region pair, connectivity was iteratively computed over a gradually shifted rectangular temporal window of length W  = 20 s, the inverse of the lowest frequency still present in the data (Leonardi & Van De Ville, 2015), with a step size of 2 s. We used Pearson correlation coefficient as a measure of connectivity.…”

Cerebral functional connectivity periodically (de)synchronizes with anatomical constraints

“…However, it has been shown recently that spurious correlations can arise from this classical approach when short windows are used. To limit this confound, we high-pass filtered the fMRI time series with a cutoff frequency of 1/w, where w is the width of the truncated portions (Leonardi and Van De Ville 2015;Zalesky and Breakspear 2015). Denoting by T the number of volumes in the fMRI time series and considering a window width w, we computed T À w þ 1 successive FC matrices from the truncated fMRI time series in each particular window, each one being shifted forward by one TR with respect to the previous one ( Fig.…”

“…In this work, we study the dynamical correlation between dFC and SC using a sliding window approach and explore the role of anatomy in the fluctuations of dFC. The first part of the paper addresses the issue of selecting a proper time window, leading to confirmatory, yet original and independent results (Allen et al 2012;Hutchison et al 2013;Leonardi and Van De Ville 2015). Next, motivated by recent work on the dynamic functional connectivity repertoire (Yang et al 2014;Sidlauskaite et al 2014) and the influence of the underlying architecture on the flexibility of dFC (Gollo et al 2015), we explore the role of anatomy in the shaping of different FC patterns, the transition between these states, and their structure.…”

Time‐resolved detection of stimulus/task‐related networks, via clustering of transient intersubject synchronization

“…These settings did not affect the sustained signal associated with the task‐blocks (cf., Simulations section) and were found not to produce any false positives in simulated datasets, irrespective of the presence/absence of low‐frequency drifts (not shown). However, higher cut‐off may be considered [i.e., related to the length of the sliding‐window: e.g., “1/window‐length,” see Leonardi and Van De Ville, 2015] particularly if the dataset is not expected to contain any regions with sustained periods of activity. At this stage, other covariates of no‐interest could be added, such as the global signal in each volume or covariates related to the individual behavior (e.g., eye‐movements recorded during fMRI scanning).…”