The recent failure of many clinical trials of neuroprotective compounds may be due in part to poor animal models of human stroke. We have developed an endovascular stroke model in nonhuman primates that is compatible with serial magnetic resonance imaging (MRI) monitoring. Using cynomologous macaques (n = 4), a microcatheter was navigated transarterially (under fluoroscopic guidance) from the femoral artery to the middle cerebral artery (MCA). The microcatheter was wedged in a branch of the MCA for 3 hours to cause focal cerebral ischemia, as verified angiographically. During occlusion and/or reperfusion, animals were scanned with MRI, and imaging findings were compared with the stained brain sections. All animals demonstrated small stroke lesions in the expected vascular territory, as seen on diffusion-weighted MRI and confirmed by postmortem examination. Reperfusion after 3 hours was confirmed angiographically (n = 2) and also by MRI (n = 4). The mean initial lesion volume, measured on the postreperfusion MRI scans, was 2.3 +/- 1.3 mL (n = 4). There was good agreement between anatomic location of the lesion on MRI and postmortem histological staining (n = 3). A "minimally invasive" primate model of focal cerebral ischemia was developed that is ideally suited to MRI studies of both acute and chronic stroke. By using serial MRI scans to measure changes in lesion size over time, we will be able to control for variability in lesion size/location. This model should prove useful as a test bed for new stroke therapies, in which noninvasive imaging findings are readily comparable to human stroke.