The location as well as the volume of the ischemic penumbra in human stroke is likely to influence the outcome of therapeutic intervention but its spatial extent is poorly characterized. Based on the observation that infarct expansion progresses from the center to the periphery of the penumbra in animal stroke models, we describe a method of mapping the three-dimensional spatial extent of the penumbra relative to the infarct in a 'Penumbragram'. Central, peripheral and external zones of the final infarct were defined according to median voxel distance from the infarct center (IC) and were further subdivided by coronal, sagittal and axial planes through the IC. In 10 patients with hypoxic, viable (penumbral) tissue identified by (18)F-Fluoromisonidazole positron emission tomography within 48 h of stroke onset, 'Penumbragrams' displaying the percentage of penumbra in each region were generated using anatomically co-registered data sets. The correlation between penumbral percentage and time from stroke onset was negative in the central (P < 0.05) and peripheral (P > 0.05) zones of the infarct and positive in external zones (P < 0.05). The validity of infarct segmentation was assessed by factor analysis with no a priori grouping of regions. Negative and positive correlations of penumbra volume and time from stroke onset were observed in seven (five corresponding to central zone of the infarct) and four (all in external zone) infarct regions and were measured reliably (Cronbach's alpha 0.84 and 0.9, respectively). The 'Penumbragram' is a valid method for objectively mapping the spatial extent of the penumbra, which is applicable to other imaging modalities.