Introduction: Structural changes post-stroke have frequently been assessed using diffusion tensor imaging (DTI). While the microstructural integrity of various parts of the corticospinal tract (CST) has been linked to motor impairment 1 , it remains unclear whether the reduction in anisotropy can be attributed to descending or crossing fibers within the CST. To address this question, we used diffusion spectrum imaging (DSI), an advanced imaging approach that overcomes the shortcomings of DTI, and delineated the influence of CST-inherent crossing and descending fibers. Patients and Methods: DSI scans of 25 chronic stroke patients (mean age=66.7, sd=11.2) and 22 age-matched controls were corrected for motion-and distortion artifacts, reconstructed, and normalized using QSIPrep 2 . Lesion and individual white matter (WM) masks were applied to the resulting generalized fractional anisotropy (gfa)-maps, focusing the analysis on secondary degeneration in WM-tracts. A deterministic mask containing the number of trackable directions per voxel 3 was applied to compartmentalize whole-brain gfa-values. Compartment-wise mean gfa was extracted from motor tracts as defined in the HCP tractography atlas 4 . Using these values, we tested for differences between patients and controls and for an association with upper and lower limb motor impairment. Results: Patients featured reduced gfa-values within the affected CST compared to age-matched controls (t(45)=-3.35, p=.002). This difference was mainly attributable to voxels containing only one fiber direction (t(45)=-3.92, p<.001), i.e., descending fibers, while voxels containing two directions, i.e., crossing fibers, showed no difference (p>0.4). Mean gfa of the affected CST significantly explained upper limb impairment (R 2 =0.30, p=.005), especially for patients suffering from persistent motor deficits (R 2 =0.43, p=.01). This association was driven by unidirectional (R 2 =0.26, p=.009) and not multidirectional voxels (p>0.3). Mean-gfa within the rubrospinal tract was related to persisting lower limb deficits. Conclusion: Anisotropy measures of the CST derived from DSI were significantly reduced and explained upper limb motor impairment in chronic stroke patients. These findings were driven by voxels containing a single fiber direction suggesting that changes in microstructural integrity of motor pathways assessed via diffusion imaging originate from Wallerian degeneration of descending fibers. In non-fully recovered stroke patients with CST damage, extrapyramidal pathways such as the rubrospinal tract may contribute to leg movements, potentially reflecting a compensatory mechanism after CST damage. Our findings underline the potential of DSI to predict motor outcome, especially when focusing the analysis on descending fibers via compartmentalization. References 1 Koch et al., 2016, Annals Clin Transl Neurol. 2 Cieslak et al., 2021, Nature Methods. 3 Volz et al., 2018, Brain Struct Funct. 4 Yeh et al., 2018, NeuroImage.