Fig. 1

Overview of data acquisition and analyses. a Multi-channel EEG were recorded at term-equivalent age from 46 infants born extremely preterm (EP) and 67 full-term healthy controls (HC). 5-min-long epochs were selected from two different vigilance states: active sleep (AS) and quiet sleep (QS). EEG during AS is continuous with relative lower amplitude fluctuations, whereas EEG during QS presents as discontinuous (trace alternant), with high amplitude signal bursts. Image reproduced with permission from Tokariev et al. (2019). b Using an infant head model, cortical source signals were computed from band-pass filtered EEG. The parcellation scheme comprised 58 regions (29 bilaterally symmetric pairs). To assess functional interactions in the brain, we computed correlation coefficients between amplitude envelopes (red lines) of parcel signals (gray lines). This led to connectivity matrices for every infant for both sleep states and for each frequency band (lower panel). c Network-based statistics were used to detect patterns of connectivity that statistically differ depending on group or sleep state or both factors. The change of the connectivity strength in the cortical patterns that showed significant interaction was regressed against key neurodevelopmental outcomes of preterm-born infants at 2 years. d The geometry of the infant cortical surface (upper panel) was used to compute cortical eigenmodes (lower panel), whose dynamics shape the organization of high-order cortical connectivity