Changes in Neural Dynamics of Brain Activity and Connectivity Independently Predict Post-Stroke Aphasia Recovery

Post-stroke language recovery involves reorganization of left-hemisphere language and bilateral domain-general multiple-demand networks. Neuroimaging predictors of recovery are widely used, yet, they are often grouped into broad predictor families, limiting clinical translation and the identification of specific therapeutic targets. Here, we assess brain region-specific neural predictors of recovery using task-related activity and connectivity within a multivariate, cross-validated predictive framework. Forty-seven patients with aphasia were examined longitudinally in the acute, subacute, and chronic phases after ischemic stroke using a sentence comprehension paradigm during functional neuroimaging, with regions of interest spanning language and multiple-demand networks. Confound-controlled multivariate regularized regression with feature ablation tested whether predictor families explained language recovery beyond lesion, age, and aphasia severity, and identified the most relevant region-specific predictors. Early recovery was independently predicted by acute and subacute activity and connectivity, whereas later improvement was predicted by activity alone. Subacute connectivity from multiple-demand to language regions predicted early recovery, while acute multiple-demand network activity predicted long-term outcomes. Later subacute-to-chronic recovery was instead predicted by subacute and chronic activity in bilateral inferior frontal regions. These findings identify phase-specific functional biomarkers of aphasia recovery beyond traditional clinical measures, highlighting multiple-demand and inferior frontal regions as candidate targets for phase-adapted neurostimulation approaches.