Edge-Community Entropy is a Novel Neural Correlate of Aging and Moderator of Fluid Cognition

Decreased neuronal specificity of the brain in response to cognitive demands (i.e., neural dedifferentiation) has been implicated in age-related cognitive decline. Investigations into functional connectivity analogues of these processes have focused primarily on measuring segregation of nonoverlapping networks at rest. Here, we used an edge-centric network approach to derive entropy, a measure of specialization, from spatially overlapping communities during cognitive task fMRI. Using Human Connectome Project Lifespan data (713 participants, 36-100 years old, 55.7% female), we characterized a pattern of nodal despecialization differentially affecting the medial temporal lobe and limbic, visual, and subcortical systems. At the whole-brain level, global entropy moderated declines in fluid cognition across the lifespan and uniquely covaried with age when controlling for the network segregation metric modularity. Importantly, relationships between both metrics (entropy and modularity) and fluid cognition were age-dependent, although entropy's relationship with cognition was specific to older adults. These results suggest entropy is a potentially important metric for examining how neurological processes in aging affect functional specialization at the nodal, network, and whole-brain level. Significant Statement Many potential clinical applications of fMRI necessitate examining localized function to assess and guide behavioral, pharmacological, surgical, and neuromodulatory interventions. Edge-community entropy offers the mathematical and conceptual advantage of calculating the functional specialization of individual nodes within an overlapping and interdependent system of communities using whole-brain, temporally-informed data. We suggest entropy during cognitive tasks may be a more conceptually appropriate functional connectivity analogue of age-related neural dedifferentiation than network segregation metrics, and demonstrate entropy strongly correlates with age and moderates cognition. Notably, entropy was strongly related to age at the whole-brain level and more strongly related than participation coefficient at the nodal level, offering greater potential to interrogate how local age-related dysfunction influences distributed systems supporting cognition and behavior.