This study was undertaken to determine the accuracy of color Doppler imaging in the diagnosis of portal vein thrombosis. Two hundred fifteen patients were studied with color Doppler imaging to determine patency of the main portal vein. Sonographic findings were confirmed in 75 patients, aged 19 to 66 years. Correlation with angiography was obtained in 13 patients, and surgical correlation was obtained in the remaining 62. Nine patients had portal vein thrombosis on the basis of these gold standards. Sonograms were classified as showing either patency or thrombosis, depending on the ability to show color flow within the main portal vein. Agreement between sonography and angiography or surgery was found in 69 patients (61 patent, eight thrombosed). One patient with a patent portal vein at sonography was found to have a thrombosed vessel at surgery, whereas five patients without portal venous flow at sonography had patent vessels at angiography (one patient) or surgery (four patients). Overall sensitivity and specificity for detection of portal vein thrombosis were 89% and 92%, with an accuracy of 92%, a false-negative rate of 0.11, a negative predictive value of 0.98, and a positive predictive value of 0.62. We postulate that the majority of errors in our study occurred in vessels that, although patent, had only sluggish flow, which could not be resolved because of technical limitations. We conclude that color Doppler imaging is a valuable screening procedure for the assessment of portal vein patency. If the sonogram shows a patent portal vein, no further studies are required. However, a lack of demonstrable flow does not always indicate thrombosis, and other imaging studies should be performed for confirmation.