Corticospinal tract integrity after stroke has been widely investigated through the evaluation of fibres descending from the primary motor cortex. However, about half of the corticospinal tract is composed by sub-pathways descending from premotor and parietal areas, to which damage may play a more specific role in motor impairment and recovery, particularly post-stroke. Therefore, the main aim of this study was to investigate lesion load within corticospinal tract sub-pathways as predictors of upper limb motor impairment after stroke. Motor impairment (Fugl-Meyer Upper Extremity score) was evaluated in 27 participants at one week and six months after stroke, together with other clinical and demographic data. Neuroimaging data were obtained within the first week after stroke. Univariate regression analysis indicated that among all neural correlates, lesion load within premotor fibres explained the most variance in motor impairment at six months (R2 = 0.44, p < 0.001). Multivariable regression analysis resulted in three independent, significant variables explaining motor impairment at six months; Fugl-Meyer Upper Extremity score at one week, premotor dorsal fibre lesion load at one week, and age below or above 70 years (total R2 = 0.81; p < 0.001). Early examination of premotor dorsal fibre integrity may be a promising biomarker of upper limb motor impairment after stroke.