Prediction of behavioral deficits in acute stroke from lesion and structural disconnection mapping.

Abstract

Prediction of behavioral deficits in stroke relies on understanding the distribution of focal damage as well as the distribution of the underlying functional anatomy. Using structural or functional magnetic resonance imaging, previous studies investigated the predictive performance of imaging biomarkers for behavioral deficits in stroke patients. However, only focal lesion information or functional connectivity information alone was used in the modelling, with a small sample size and on a specific behavioral deficit domain. In this study, we investigated the prediction of behavioral deficits in acute stroke using both focal lesion patterns and structural disconnection mapping on a cohort of 551 ischemic stroke patients within one week post symptom onset. Five behavioral deficits domains, including motor, cognitive, visual, somatosensory and coordination deficits, were investigated. A probabilistic map of lesion-induced structural "disconnectome" map was created to estimate the degree of structural disconnection due to lesions. In the predictive modelling, both lesion volume and location and distant structural disconnections were included in combination with the clinical information. The results showed that improved prediction performance was achieved when considering both focal lesion patterns and global lesion-induced structural disconnections for all five behavioral deficits groups. Distinct lesion maps were obtained for each behavioral deficit, providing insights into neurobiological mechanisms of stroke functional impairment.