Abstract: Objective: The use of diffusion-weighted imaging (DWI) to define irreversibly damaged infarct core is challenged by data suggesting potential partial reversal of DWI abnormalities. However, previous studies have not considered infarct involution. We investigated the prevalence of DWI lesion reversal in the EPITHET Trial. Methods: EPITHET randomized patients 3-6 hours from onset of acute ischemic stroke to tissue plasminogen activator (tPA) or placebo. Pretreatment DWI and day 90 T2-weighted images were coregistered. Apparent reversal of the acute ischemic lesion was defined as DWI lesion not incorporated into the final infarct. Voxels of CSF at follow-up were subtracted from regions of apparent DWI lesion reversal to adjust for infarct atrophy. All cases were visually cross-checked to exclude volume loss and coregistration inaccuracies. Results: In 60 patients, apparent reversal involved a median 46% of the baseline DWI lesion (median volume 4.9 mL, interquartile range 2.6-9.5 mL) and was associated with less severe baseline hypoperfusion (p < 0.001). Apparent reversal was increased by reperfusion, regardless of the severity of baseline hypoperfusion (p = 0.02). However, the median volume of apparent reversal was reduced by 45% when CSF voxels were subtracted (2.7 mL, interquartile range 1.6-6.2 mL, p < 0.001). Perfusion-diffusion mismatch classification only rarely altered after adjusting the baseline DWI volume for apparent reversal. Visual comparison of acute DWI to subacute DWI or day 90 T2 identified minor regions of true DWI lesion reversal in only 6 of 93 patients. Conclusions: True DWI lesion reversal is uncommon in ischemic stroke patients. The volume of apparent lesion reversal is small and would rarely affect treatment decisions based on perfusion-diffusion mismatch.