Regression analyses for the association between different gFA-based CST measures and ARAT motor scores. (A) Mean gFA values within the ipsilesional CST explained motor impairment for the entire cohort and to an even higher degree for the non-fully recovered subgroup. (B) By obtaining the mean gFA value from one-directional voxels within the ipsilesional CST, it becomes evident that the association between anisotropy and motor performance was driven by descending fibres. Note that the mean taken from two-directional voxels did not significantly explain motor performance (P > 0.2, FDR-corrected). (C) To improve the signal-to-noise ratio, we calculated the asymmetry index that accounts for the subject-specific level of atrophy by considering both the affected and unaffected CST. Across all CST voxels, we observed a significant association with the ARAT score for the entire sample and the non-fully recovered subgroup. (D) Entering the asymmetry index from one-directional voxels as the predictor into the regression models yielded a similar ratio of explained variance as in (C).