Non-invasive techniques in the diagnosis of jaundice--ultrasound and computer

This study describes the use of ultrasound in 84 consecutive patients to distinguish between medical (intrahepatic) and surgical (extrahepatic) types of jaundice. Accuracy rates in the 84 patients investigated were compared with those found in 169 patients using the computer-aided diagnostic model for jaundice. In 50 patients in whom both techniques were carried out direct comparison was possible. Ultrasound failed to give adequate visualisation for technical reasons in two patients, but 75 of the remaining 82 (91.5%) were correctly separated into medical and surgical categories. A similar percentage (90%, 152 of 169 patients) were correctly classified by the computer-assisted model, and in the 50 patients assessed by both techniques correct diagnostic separation was achieved in 43 (86%) by ultrasound and in 42 (84%) by computer. However, in this latter group two of the errors with computer diagnosis were falsely positive for a surgical jaundice (extrahepatic obstruction), and a surgical exploration performed on the basis of this test could have been unnecessary. This was not found with ultrasound in these same patients, all the errors being false negative. A high degree of confidence can therefore be attached to the demonstration by ultrasound of a dilated biliary tree.