This study aimed to investigate the long-term motor outcome according to early diffusion tensor tractography findings for the affected corticospinal tract (CST) in poststroke hemiplegic patients. A total of 48 supratentorial subacute patients after stroke were enrolled, who had a brain MRI scan within 6 weeks from onset, and no stroke recurrence reported within the 2-year follow-up period. Diffusion tensor images were obtained and CSTs were reconstructed. The participants were classified into three groups: type A, the CST originating from the primary motor cortex was preserved around the lesion area; type B, the CST was similar to type A, except that the fiber originated from the area adjacent to the primary motor cortex; and type C, the CST was interrupted or not shown. Motor functions using Fugl-Meyer Motor Assessment (FMA), the Box and Block Test (BBT), and Functional Ambulation Category, and cognitive function using Mini-Mental Status Examination (MMSE) were measured at baseline and at 2 years from stroke onset. Changes in FMA and BBT were significantly different according to diffusion tensor tractography type at follow-up (P<0.05), but Functional Ambulation Category and Mini-Mental Status Examination were not. In post-hoc analysis, groups A and B showed greater significant improvements on the BBT and on the upper FMA subscale (shoulder/elbow, wrist, and hand) compared with group C (corrected P<0.05), but did not on lower FMA. This study showed the importance of CST integrity for stoke motor recovery. The early integrity of the CST may be useful in predicting long-term motor outcomes, specifically with motor recovery of the upper extremity and hand function.