Unveiling the drivers of individual critical synchronization dynamics with personalized whole-brain computational modeling

Brain activity exhibits critical-like dynamics, which is fundamental for cognitive functions. Whole-brain computational modeling has become indispensable in studying the systems-level mechanisms that govern the dynamics of large-scale brain activity in health and disease. However, previous approaches have focused on modeling either functional connectivity or criticality, and capturing both simultaneously in whole-brain systems remains a challenge. Here, we advance a new approach to modeling multi-scale synchronization dynamics with the Hierarchical Kuramoto model. At two levels of hierarchy, the model comprises a network of nodes that each encapsulate a large number of coupled oscillators, which enables dissection of local and inter-areal coupling and dynamics. We found the model to produce oscillations with critical-like dynamics, including emergent long-range temporal correlations (LRTCs) and inter-areal order correlations at the critical transition between asynchronous and synchronous phases. The model predicted unique structure-function relationships for functional observables so that coupling with structure peaked at criticality for LRTCs and order correlations but collapsed for inter-areal phase synchronization. Comparison of the model dynamics with human resting-state magnetoencephalography (MEG) showed that model-MEG correlations were maximized in the subcritical side of an extended critical regime for all observables and across the model parameter space. Finally, we found that the model may also yield realistic-like multi-peak power spectra and that also their similarity with MEG data is maximized below peak criticality. In conclusion, Hierarchical Kuramoto modeling yields realistic and separable local and global synchronization dynamics that are directly comparable with neuroimaging, which opens new avenues for understanding collectivity in brain activity.