Diagnostic accuracy of pancreatic computed tomography (CT) using a 3 s scanner was compared to our previously published results using an 18 s scanner. Technically unsuccessful examinations have decreased to less than 1%. Pancreatic masses were detected prospectively in 83% of patients with pancreatic carcinoma and retrospectively in all such patients. False negative diagnoses of normal pancreas in proven pancreatic carcinoma were reduced fourfold down to 6%. False positive diagnoses of pancreatic carcinoma remained less than 1% but were not significantly reduced compared to our second year's experience using an 18 s scanner. Interpretation errors and errors due to limitations of CT as a diagnostic technique are discussed. © Raven Press, New York.
