Structurally informed resting-state effective connectivity recapitulates cortical hierarchy

Abstract

Neuronal communication relies on the anatomy of the brain, yet it remains unclear whether, at the macroscale, structural (or anatomical) connectivity provides useful constraints for modeling effective connectivity. Here, we assess a hierarchical empirical Bayes model that builds on a well-established dynamic causal model by integrating structural connectivity into resting-state effective connectivity via priors. In silico analyses show that the model successfully recovers ground-truth effective connectivity and compares favorably with a popular alternative. Analyses of empirical data reveal that a positive, monotonic relationship between structural connectivity and the prior variance of group-level effective connectivity generalizes across sessions and samples. Finally, attesting to the models biological plausibility, we show that inter-network differences in the coupling between structural and effective connectivity recapitulate a well-known unimodal- transmodal hierarchy. These findings underscore the value of integrating structural and effective connectivity to enhance understanding of functional integration, with implications for health and disease. Significance statementTo advance the understanding of how neuronal populations interact in vivo, it is essential to develop models that integrate neuroimaging modalities. Here, we show that integrating structural connectivity into dynamic causal modeling of resting-state effective connectivity substantially improves model evidence, yields reliable inferences, and demonstrates face and construct validity in silico. Furthermore, this integration reveals that structural connectivitys influence on effective connectivity varies along an established unimodal-transmodal cortical hierarchy. This finding provides the first evidence of a network-dependent modulation of the relationship between structural and effective connectivity in humans. Against the backdrop of sustained and widespread interest in dynamic causal modeling, this study highlights the added value of integrating structural connectivity-based constraints, offering a more biologically grounded account of brain dynamics.