To enhance reproducibility, RSNs were defined in each participant using the functional templates available through the Functional Imaging in Neuropsychiatric Disorders Lab at Stanford University, USA (
https://findlab.stanford.edu/functional_ROIs.html) (Supplementary Figure S1 and Table S1) (Shirer et al.
2012). We specifically examined the dorsal attention network (DAN), the central executive network (CEN), the salience network (SAL), the somatosensory network (SMN), the visual network (VN), the auditory network (AN), and the language network (LAN). We considered the default mode network (DMN) in terms of its sub-divisions into the dorsal DMN (dDMN), the ventral DMN (vDMN) and the precuneus network (PN) because prior evidence indicates that each DMN sub-division may support different cognitive processes (Andrews-Hanna et al.
2010) and may have a different functional impact on whole brain organization (Doucet et al.
2011). In each participant, we calculated the average time-series of all the voxels in each region of each RSN, and then applied the bandpass filtering (0.01–0.1 Hz) to isolate low-frequency resting-state blood oxygen-level dependent (BOLD) signal fluctuations (Cordes et al.
2001). The Hilbert transform was applied to the bandpass-filtered fMRI signals to compute the associated analytical signals. The analytic signal represents a narrowband signal,
\(s\left(t\right)\), in the time domain as a rotating vector with an instantaneous phase,
\(\phi \left(t\right)\), and an instantaneous amplitude,
\(A\left(t\right)\), i.e.,
\(s\left(t\right)=A\left(t\right)\text{cos}\left(\phi \left(t\right)\right)\). The phase and the amplitude are given by the argument and the modulus, respectively, of the complex signal
\(z\left(t\right)\), given by
\(z\left(t\right)=s\left(t\right)+i.\text{H}\left[s\left(t\right)\right]\), where
\(i\) is the imaginary unit and
\(\text{H}\left[s\left(t\right)\right]\) is the Hilbert transform of
\(s\left(t\right)\) (Glerean et al.
2012). To evaluate the dynamic properties of each RSN, we computed the Kuramoto order parameter
\(R\left(t\right)\), defined as
$$R\left( t \right)=\frac{1}{N}\left| {~\mathop \sum \limits_{{n=1}}^{N} {e^{i{\varphi _n}(t)}}~} \right|$$
where
N is the total number of regions within each RSN and
\({\phi }_{n}\left(t\right)\) is the instantaneous phase of the BOLD signal at region
n of each RSN. For each RSN, metastability was defined as the standard deviation of the Kuramoto order parameter
\(R\left(t\right)\) over time (Cabral et al.
2011; Lee et al.
2017; Lee and Frangou
2017; Shanahan
2010).