The structural integrity of the corticospinal tract (CST) is an important biomarker of poststroke upper limb recovery. Injured CST undergoes Wallerian degeneration rostrocaudally during the first few months. However, there is no standardized measurement of the structural integrity of the CST. This study aimed to determine the measurement accuracy of the structural integrity of the CST. This cross-sectional study included 50 consecutive patients with middle cerebral artery stroke who underwent diffusion tensor imaging upon transfer from the acute stroke unit to the inpatient rehabilitation facility (2018-2022). We evaluated hemiplegic upper limb function using Shoulder Abduction and Finger Extension (SAFE) scores. Fractional anisotropy values representing the structural integrity of the CST were evaluated using 4 region of interest-based and 2 tract-based measurements, including the posterior limb of internal capsule, cerebral peduncle, pons, pontomedullary junction, entire CST, and CST in the brainstem. Multivariate linear regression models and the area under the curve (AUC) were used to determine measurement accuracy for hemiplegic upper limb function. The structural integrity of the CST at the pontomedullary junction showed the highest explanatory power, followed by the entire CST, in the multivariate linear regression models (adjusted R2=0.459 and 0.425, respectively). The structural integrity of the CST at the pontomedullary junction also showed the highest AUC, followed by the entire CST, in discriminating patients with a SAFE score of <8 or 5 from those with SAFE ≥8 or 5 (SAFE <8: AUC, 0.90 [95% CI, 0.80-1.00]; AUC, 0.83 [0.66-0.99]; SAFE <5: AUC, 0.87 [0.77-0.96]; AUC, 0.83, [0.72-0.95], respectively). The structural integrity of the CST measured at the pontomedullary junction or entire CST demonstrated the highest accuracy for hemiplegic upper limb function in the subacute phase of stroke.