Generative Network Model Reveals Different Trajectories in Brain Networks of PSA and Stroke Patients

Post-stroke aphasia (PSA), induced by acute brain injury, is an acquired language disorder resulting from stroke, primarily characterized by impairments across multiple linguistic functions including spontaneous speech, auditory comprehension, repetition, naming, reading, and writing. Previous studies have demonstrated that the topological features of healthy brains align with complex networks, whereas key topological features in PSA patients (e.g., interhemispheric connectivity, functional connectivity (FC) of language networks) undergo significant alterations due to acute brain injury. However, traditional graph-theoretical approaches fail to elucidate the dynamic evolutionary patterns underlying functional reorganization in brain networks. Moreover, existing research lacks systematic exploration of trajectory characteristics and economic cost regulation mechanisms in network generation among PSA patients. To address these gaps, this study introduces a framework based on generative network modeling, integrating non-geometric rules (power-law functions of topological relationships) and geometric rules (connection distance calculations) to simulate the formation process of functional brain networks. By parametrically modulating the balance between nodal connection propensity and distance cost, and comparing the optimal matching between simulated and observed networks, we explored the evolutionary mechanism of brain networks in PSA patients. Key findings include: For the FC matrix with 10% sparsity, 1) The homogeneous model combined with geometric distance-based economic costs generates optimal simulated networks; 2) PSA patients exhibit significantly higher absolute values of parameter $\eta $ compared to general stroke patients ( ${p}\lt {0}.{05}$ ), indicating increased economic costs for connections with distal nodes; 3) PSA patients show the highest $\gamma $ values, with significant reduction in inter-nodal connection propensity versus healthy controls ( ${p}\lt {0}.{05}$ ), suggesting impaired network integration efficiency; 4) Trajectory analysis reveals decreased parametric values in thalamus-related regions but elevated values in occipital and cerebellar regions among PSA patients, with distance costs showing negative correlation with stroke patients ( ${R}^{{2}}={0}.{86}$ ), uncovering region-specific trajectories of functional reorganization around lesions. By constructing a computational model incorporating economic clustering rules, this study clarifies differential network evolution patterns between PSA and general stroke patients, provides theoretical foundations for targeted neuromodulation and intervention strategy optimization, and addresses the limitations of traditional graph theory in dynamic mechanism analysis.