Network flexibility facilitates treatment-induced recovery in post-stroke aphasia

Predicting the recovery of post-stroke aphasia, particularly in the chronic phase, poses significant challenges. While many studies have focused on static resting state functional connectivity measures as biomarkers for language recovery, there is a significant lack of studies investigating dynamic functional connectivity, which takes into account second to minute timescale variation in connectivity and has been used extensively in other contexts. This retrospective cohort study investigates the predictive value of dynamic functional connectivity for treatment response in post-stroke aphasia. Two temporal metrics, temporal variability and community stability were explored. Temporal variability quantifies the magnitude of variation in the strength of functional connections over time, while community stability measures the mean length of time over which dynamic functional connectivity remains relatively stable. Additionally, a dynamic functional connectivity states analysis, in which time windows are clustered into states using a k-means clustering algorithm, was used to investigate time-varying network properties of dynamic functional connectivity in relation to treatment response. Baseline MRI data were collected for 30 participants with chronic post-stroke aphasia who received a semantic treatment aimed at improving word-finding. Dynamic functional connectivity was generated using a sliding-window approach and the relationships between treatment-induced improvements in naming accuracy and the respective temporal metrics and network properties were evaluated using linear mixed effects models predicting naming accuracy. Specifically, a significant interaction effect of session (i.e. time) and the respective measure was interpreted as a significant relationship between the measure and treatment response. Both temporal metrics were found to be significantly associated with naming improvements, with both higher temporal variability (i.e. magnitude of fluctuations in dynamic functional connectivity; interaction effect: β = 0.45, P = 3.6e-09, η1 = 0.072) and higher community stability (i.e. moment-to-moment stability of dynamic functional connectivity; interaction effect: β = 0.0046, P = 5.4e-06, η ¹ = 0.044) predicting greater treatment gains. These findings suggest that patients who benefit most from treatment have neural dynamics characterized by a large magnitude, but low frequency, of fluctuations in network synchrony. Finally, consistent with previous studies, patients who spent more time in a state characterized by higher modularity showed greater treatment response (interaction effect: β = 7.9e-05, P = 6.2e-04, η ¹ = 0.025). Overall, the findings of this study provide evidence for the utility of dynamic functional connectivity in studying post-stroke language recovery as a measure that captures interindividual difference in the capacity for recovery. They create a new direction for post-stroke aphasia research focusing on temporal dynamics and their relationship to long-term functional network plasticity.