Abstract
BACKGROUND: Subcortical structures are pivotal for cortico-subcortical communication, yet are highly vulnerable to stroke. Damage to these structures can produce widespread cortical dysfunction, but how subcortical stroke reshapes the macroscale hierarchical organization of cortical networks remains poorly understood. Clarifying this relationship is critical for understanding subcortical contributions to brain-wide communication and for identifying reliable neural markers of clinical outcome. METHODS: We analyzed resting-state fMRI data from two independent cohorts of patients with subcortical stroke (discovery, 55 patients, 49 controls; replication, 23 patients, 26 controls). Cortical functional organization was mapped using diffusion map embedding to derive connectivity gradients, and group differences were quantified at both global and regional levels. Gradient metrics were further compared with conventional functional connectivity strength and tested for predictive utility using multivariate ridge regression against neurological outcome measures. RESULTS: Across cohorts, patients consistently exhibited compression of the principal unimodal-to-transmodal gradient, reflecting diminished hierarchical differentiation between sensory and association cortices. These alterations were stable over the first three months of recovery and spatially aligned with normative maps of dopaminergic, GABAergic, glutamatergic, serotonergic and cannabinoid systems. By contrast, conventional functional connectivity strength metrics showed poor reproducibility. Gradient features outperformed connectivity strength in predicting overall neurological severity (NIHSS scores) and generalized robustly across cohorts, although predictive power for motor-specific deficits (FMA scores) was limited. CONCLUSIONS: Subcortical stroke disrupts cortical functional hierarchy in a reproducible and neurochemically constrained manner, and this disruption serves as a robust biomarker of global neurological impairment. These findings establish cortical gradients as a mechanistically informative and clinically relevant tool for probing subcortical-cortical interactions, with potential to inform prognosis and targeted neurorehabilitation.