Introduction Dynamic functional connectivity (dFC) is a pervasive feature of brain activity, even at rest, but its functional role remains debated. We ask whether temporal reconfiguration of functional links can be advantageous when maintaining links is costly. We hypothesize that dFC helps resolve the trade-off between large-scale integration and transient local segregation by reusing a limited communication budget over time.
Methods Resting-state fMRI dFC was modeled as a cost-constrained temporal communication network. Sliding-window functional-connectivity frames were binarized at different densities and compared with equal-cost static architectures and temporal null models. Information dispatch was quantified using smart and random walkers, measuring irrigation reach, penalized latency, temporal clustering, return latency and neighborhood retention. A connectome-based mean-field model was used as a mechanistic benchmark.
Results Empirical dFC outperformed equal-cost static networks in sparse, high-cost regimes, allowing information to reach more nodes and reducing penalized latency. However, more randomized temporal nulls often spread information even more efficiently, showing that empirical dFC is not optimized for diffusion alone. Empirical networks also preserved strong spatial and temporal clustering, rapid return to source nodes and high neighborhood retention, supporting transient local segregation.
Discussion These findings suggest that resting-state dFC is neither a mere by-product of neural dynamics nor a simple maximizer of global spreading. Instead, it reflects a structured regime of controlled persistence and renewal: local neighborhoods remain stable long enough for transient recirculation, before broader network-wide spreading occurs. dFC may therefore be a resource-efficient solution to competing demands for integration and segregation in brain communication.
References Mengiste, S.A., and Battaglia, D. (2026). Dynamic Functional Connectivity Resolves Brain Integration-Segregation Trade-off Under Costly Links. arXiv. https://doi.org/10.48550/arXiv.2604.11608.
Acknowledgement This work was supported by the PEPR Sané Numérique program (France 2030), project “Brain Health Trajectories (BHT)”, implemented by the Agence Nationale de la Recherche (ANR) under grant number ANR-22-PESN-0012-BHT. We wish to thank Alain Barrat, Caio Seguin and Sinisa Pajevic for inspiring discussions and Anagh Pathak for sharing time-series from connectome-based simulations.
