Proposal of a diagnostic algorithm for intraductal ultrasonography to distinguish between benign and malignant biliary strictures

Abstract

To reexamine the recognizability of intraductal ultrasonography (IDUS) findings from an imaging database and propose a novel algorithm for clinical application. IDUS images of 102 patients who had undergone IDUS examinations for indeterminate causes of common bile duct dilation were independently reviewed by two endoscopists. The strength of the inter-rater agreement between the endoscopists was analyzed using Cohen's kappa (κ). An algorithm was implemented by arranging the IDUS characteristics according to their recognizability. The proposed algorithm was evaluated by examining the inter-rater agreement and diagnostic accuracy before and after the use of the algorithm. The strength of the inter-rater agreement was good for common bile duct stones with or without acoustic shadowing; intraluminal tumors; or bile duct wall thicknesses of more than or equal to 9 mm (κ > 0.8); followed by intraluminal hypoechoic nodules without common bile duct stone characteristics (κ = 0.771); and finally eccentric wall thickening, outer layer disruption, irregular mucosa, and destructed mural layers (κ: 0.595-0.419). Our algorithm improved the strength of inter-rater agreement with a diagnostic accuracy of 81.4%. We proposed an algorithm according to the recognizability of IDUS characteristics, and it can be used by endoscopists to evaluate such characteristics and determine the cause of biliary obstruction.