Extent to Which Network Hubs Are Affected by Ischemic Stroke Predicts Cognitive Recovery

Abstract

Background and Purpose— It is uncertain what determines the potential for cognitive recovery after ischemic stroke. The extent to which strategic areas of the brain network, so-called hubs, are affected by the infarct could be a key factor. We developed a lesion impact score, which estimates the damage to network hubs by integrating information on infarct size with healthy brain network topology. We verified whether the lesion impact score indeed reflects global network disturbances in patients and assessed if it could predict cognitive recovery. Methods— Seventy-five ischemic stroke patients without signs of a prestroke cognitive disorder were included, all with evidence of a cognitive disorder during hospitalization. A brain magnetic resonance imaging and neuropsychological assessment were performed 5 weeks (±1 week) after stroke. Neuropsychological testing was repeated after 1 year to assess cognitive recovery. Brain networks were reconstructed from diffusion-weighted data and consisted of 90 gray matter regions (ie, network nodes). A standard brain network map, indicating the hub-score of each node, was obtained from network data of 44 cognitively healthy adults. For each patient, we calculated the lesion impact score by multiplying the percentage of node volume affected by the infarct with the node’s corresponding hub-score. The patients’ maximum lesion impact score was used as outcome predictor. Results— A higher lesion impact score in patients, indicating an increasing infarct size in nodes with a higher hub-score, was related to lower global brain network efficiency (β=−0.528 [−0.776 to −0.277]; P <0.001), independent of age, brain volume, infarct volume, and white matter hyperintensity severity. A lower lesion impact score, however, was an independent predictor of cognitive recovery 1 year after stroke (odds ratio=0.434 [0.193–0.978]; P =0.044). Conclusions— We introduced a lesion impact score that combines information on infarct size and network topology to predict long-term recovery after stroke. This score can potentially be used in a clinical setting, also without availability of high-resolution diffusion-weighted magnetic resonance imaging.