Advancing post-stroke outcome prediction with movement-specific structural and functional brain atlases.

Stroke is a leading cause of death and disability, with motor deficits contributing significantly to post-stroke disability. Brain atlases hold promise for predicting motor outcomes post-stroke, but existing tools often lack comprehensive coverage of motor-related brain regions and do not integrate both structural and functional measures. This retrospective longitudinal study aimed to develop and evaluate neuroimaging biomarkers for predicting post-stroke motor outcomes by constructing comprehensive sensorimotor brain atlases. We developed two novel atlases: a sensorimotor structural connectivity atlas (SSCA), integrating three existing tractography-based atlases, and a probabilistic sensorimotor activation-based atlas (SMAA), derived from an ALE meta-analysis of 3,252 activation foci related to motor execution and learning. We assessed their predictive value by analysing the relationship between baseline lesion load and Action Research Arm Test scores at 12 weeks post-ischemic stroke in 142 patients, using multivariable linear regression models. Lesion loads from five published atlases were also quantified for comparison. While the SSCA demonstrated moderate predictive performance, it was outperformed by the Sensorimotor Area Tract Template, indicating that broader tract coverage did not improve prediction. Despite comprising only 12.8% of the Brainnetome atlas volume, the SMAA achieved comparable performance with reduced model complexity. Overall, these atlas-based lesion load metrics correlated with upper limb motor outcomes but provided only limited additional predictive value beyond baseline Fugl-Meyer Assessment for Upper Extremity scores, highlighting the need for refinement and future multimodal approaches.