BACKGROUND AND PURPOSE This study was designed to assess the accuracy of transcranial color-coded sonography (TCCS) as compared to magnetic resonance angiography (MRA) for detecting intracranial arterial stenosis in patients with acute cerebral ischemia. METHODS The authors prospectively identified 120 consecutive patients admitted with acute ischemic stroke and performed both TCCS and MRA with a mean interval of 1 day. TCCS data (sampling depth, peak systolic and end diastolic angle-corrected velocity, mean angle-corrected velocity, and pulsatility index) for middle cerebral arteries (MCAs) were compared to MRA data and classified into 4 grades: normal (grade 1): normal caliber and signal; mild stenosis (grade 2): irregular lumen with reduced signal; severe stenosis (grade 3): absent signal in the stenotic segment (flow gap) and reconstituted distal signal; and possible occlusion (grade 4): absent signal. The cutoffs were chosen to maximize diagnostic accuracy. RESULTS Interobserver agreement for MRA grading resulted in a weighted-kappa value of 0.776. The rate of poor temporal window was 37% (89/240). Doppler signals were obtained in 135 vessels, and the angle-corrected velocities (peak systolic, end diastolic, mean) were significantly different (P = .001, P = .006, and P < .001) among the MRA grades: grade 1 (100, 47, 68 cm/s), grade 2 (171, 72, 110 cm/s), grade 3 (226, 79, 134 cm/s), grade 4 (61, 26, 39 cm/s). Additionally, an angle-corrected MCA peak systolic velocity > or = 120 cm/s correlates with intracranial stenosis on MRA (grade 2 or worse) with high specificity (90.5%; 95% confidence interval = 78.5%-96.8%) and positive predictive value (93.9%) but relatively low sensitivity (66.7%; 95% confidence interval = 61.2%-69.5%) and negative predictive value (55.1%). CONCLUSION Elevated MCA velocities on TCCS correlate with intracranial stenosis detected on MRA. An angle-corrected peak systolic velocity > or = 120 cm/s is highly specific for detecting intracranial stenosis as defined by significant MRA abnormality.