To determine the accuracy of magnetic resonance (MR) imaging in comparison to cerebral angiography after radiosurgery for an arteriovenous malformation (AVM), the authors reviewed the records of patients who underwent radiosurgery at the University of Pittsburgh Medical Center before 1992. All patients in the analysis had AVMs in which the flow-void signal was visible on preradiosurgical MR imaging. One hundred sixty-four postradiosurgical angiograms were obtained in 140 patients at a median of 2 months after postradiosurgical MR imaging (median 24 months after radiosurgery). Magnetic resonance imaging correctly predicted patency in 64 of 80 patients in whom patent AVMs were seen on follow-up angiography (sensitivity 80%) and angiographic obliteration in 84 of 84 patients (specificity 100%). Overall, 84 of 100 AVMs in which evidence of obliteration was seen on MR images displayed angiographic obliteration (negative predictive value, 84%). Ten of the 16 patients with false-negative MR images underwent follow-up angiography: in seven the lesions progressed to complete angiographic obliteration without further treatment. Exclusion of these seven patients from the false-negative MR imaging group increases the predictive value of a negative postradiosurgical MR image from 84% to 91%. No AVM hemorrhage was observed in clinical follow up of 135 patients after evidence of obliteration on MR imaging (median follow-up interval 35 months; range 2-96 months; total follow up 382 patient-years). Magnetic resonance imaging proved to be an accurate, noninvasive method for evaluating the patency of AVMs that were identifiable on MR imaging after stereotactic radiosurgery. This imaging modality is less expensive, more acceptable to patients, and does not have the potential for neurological complications that may be associated with cerebral angiography. The risk associated with follow-up cerebral angiography may no longer justify its role in the assessment of radiosurgical results in the treatment of AVMs.