Disrupted resting-sate brain network dynamics in children born extremely preterm

Abstract

The developing brain has to adapt to environmental and intrinsic insults after extremely preterm (EPT) birth. Ongoing maturational processes maximize their fit to the environment and this can provide a substrate for neurodevelopmental failures. Resting-state functional magnetic resonance imaging was used to scan 33 children born EPT, at < 27 weeks of gestational age, and 26 full-term controls at 10 years of age. We studied the capability of a brain area to propagate neural information (intrinsic ignition) and its variability across time (node-metastability). This framework was computed for the dorsal attention network (DAN), frontoparietal, default-mode network (DMN), and the salience, limbic, visual, and somatosensory networks. The EPT group showed reduced intrinsic ignition in the DMN and DAN, compared with the controls, and reduced node-metastability in the DMN, DAN, and salience networks. Intrinsic ignition and node-metastability values correlated with cognitive performance at 12 years of age in both groups, but only survived in the term group after adjustment. Preterm birth disturbed the signatures of functional brain organization at rest in 3 core high-order networks: DMN, salience, and DAN. Identifying vulnerable resting-state networks after EPT birth may lead to interventions that aim to rebalance brain function.

Keywords: brain development; brain network dynamic; cognitive neurodevelopment; extreme prematurity.

Fig. 1

Intrinsic ignition framework. 1) We extracted the BOLD time series and computed the phase space of the BOLD signal for each of the 100 brain areas. 1A) the time series for each brain area was extracted using a resting-state atlas (Schaefer et al. 2018), and 1B) then, the phase space of the BOLD signals for each brain area was assessed by calculating the Hilbert transform. The BOLD signal (red) was filtered between 0.01 and 0.07 Hz (blue) and converted with the Hilbert transform into an analytical signal, represented by its instantaneous amplitude and its phase (with real and imaginary components). The phase dynamics are represented in the complex plane as (black bold line), with representing the imaginary part (black dotted lines) and representing the real part. The purple arrows denote the Hilbert phases for a given brain area over time. 2) Intrinsic ignition. 2A) Events were obtained using a threshold (Tagliazucchi et al. 2012) (green area), and the activity in the rest of the network was calculated for each ignition event crossing the threshold (red stippled area), in the 4TR time window (gray area). 2B) A binarized phase-lock matrix was obtained from the time window. 2C) The integration was obtained from this phase-lock matrix by calculating the largest subcomponent (Deco et al. 2017). Finally, we repeated the process for each driving event and the framework returned the ignition and node-metastability for each brain area across the network. Figure adapted from 2 studies (Deco and Kringelbach 2017; Escrichs et al. 2019).

Fig. 2

a) Ignition measure. The EPT group exhibited significantly lower ignition values in the DMN and DAN than the control group. Rendered brains showed the absolute difference between the groups, with the regions in yellow representing the largest differences. The boxplots show the ignition values for each group and resting-state network. b) Node-metastability measure. The EPT group exhibited lower values of node-metastability in the DMN, DAN, and salience networks than the control group. P-values are based on the Wilcoxon rank-sum test.