Fig. 1

Machine learning pipeline for the prediction of neuropsychological scores from Resting State Functional Connectivity (RSFC) matrices of stroke patients. The mean RSFC matrix (\(324 \times 324\)) across all patients is shown as reference. Parcels in the matrix are sorted in relation to 12 large-scale intrinsic brain networks. Model predictions can be validated against left-out empirical data (top-right panels), while the most predictive edges can be visualized in a brain-like topology to better understand the contribution of different circuits to the behavioral deficits (bottom-right panel)