Dynamic Network Reconfiguration during Attention and Working Memory: Integrating Neuroimaging and Computational Modeling for Cognitive Profiling

Understanding how large-scale brain networks dynamically reorganize to support attention and working memory remains a central challenge in cognitive neuroscience. Emerging evidence suggests that the prefrontal–parietal network flexibly interacts with subcortical and default mode regions to optimize cognitive control, yet the temporal mechanisms underlying this reconfiguration remain unclear.

This study employed a multi-modal experimental design integrating functional MRI, electroencephalography (EEG), and computational modeling to investigate dynamic network transitions during attentional load and memory manipulation tasks. Eighty healthy adults completed a parametric n-back paradigm with real-time neuroimaging. Dynamic causal modeling (DCM) quantified directed connectivity, while graph-theoretical metrics assessed modularity and integration across cortical systems. Additionally, recurrent neural network (RNN) simulations were trained to reproduce observed neural trajectories and predict behavioral accuracy.

Results revealed a robust task-dependent reorganization of the frontoparietal control system, with transient decoupling from the default mode network (DMN) during high-load trials. EEG phase-synchrony analyses indicated theta–gamma coupling between dorsolateral prefrontal cortex and intraparietal sulcus as a predictor of task performance (r = 0.67, p < 0.001). Computational models recapitulated these oscillatory dynamics, suggesting that recurrent feedback mechanisms enable efficient information maintenance.

Our findings provide convergent neurobiological and computational evidence that cognitive flexibility arises from transient, hierarchical synchronization across distributed neural systems. These results advance a mechanistic framework for understanding attention–memory interactions and may inform neuroadaptive interventions for cognitive decline.