OBJECTIVE The purpose of this study was to validate the accuracy and criteria of transcranial color-coded sonography (TCCS) in detecting severe middle cerebral artery (MCA) stenosis. METHODS One hundred ninety-three patients with acute ischemic cerebrovascular disease who received both TCCS and magnetic resonance angiography (MRA) examinations were evaluated. Middle cerebral artery stenosis assessed by MRA was graded as follows: grade 0, normal to mild (< 50%); grade 1, focal severe stenosis (> or = 50% and stenotic length within the M1 prebifurcation segment); and grade 2, diffuse severe stenosis (> or = 50% and stenotic length greater than the M1 prebifurcation segment). The peak systolic velocity (Vs) and mean velocity (Vm) of bilateral MCAs were obtained by TCCS. Estimates of sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were calculated for TCCS relative to MRA. RESULTS After 46 patients were excluded, 309 MCAs (grade 0, 77.3%; grade 1, 15.5%; and grade 2, 7.1%) were evaluated. The optimal diagnostic accuracy of TCCS for the detection of grade 1 MCA stenosis can be reached if Vs is 140 cm/s or higher or Vm is 90 cm/s or higher (sensitivity, 83.3%; specificity, 91.2%). The optimal diagnostic accuracy of grade 2 MCA stenosis can be obtained if V(s) is lower than 50 cm/s or V(s) is 140 cm/s or higher (sensitivity, 81.8%; specificity, 92.1%). In our study, none of the grade 1 but around half of the grade 2 stenosis showed a low flow velocity pattern. CONCLUSIONS Transcranial color-coded sonography is reliable in detecting severe MCA stenosis. Except for high flow velocity, the addition of a low cutoff of normal flow velocity in our criteria not only increases the study sensitivity but also enables the identification of around half of diffuse severe MCA stenosis.