S3339 Enhancing Diagnosis of Indeterminate Biliary Strictures: A Case Series Using AI–Enhanced Cholangioscopy

Introduction: The evaluation of indeterminate biliary strictures (BSs) remains clinically challenging due to innaccurate visual assessment, high interobserver variability and suboptimal biopsy yield during digital single-operator cholangioscopy (DSOC). Artificial intelligence (AI), particularly convolutional neural networks (CNNs), offers a promising support real-time diagnostic accuracy in this setting. This case series aims to describe examples of how a previously published AI-based tool can be integrated during DSOC for real-time characterization of BSs across global high-volume centers and case types. Predictions were generated over 30-second intervals and interpreted alongside visual findings and histology. AI results were not used to guide clinical decisions at the time of care. Case Description/Methods: In Case 1, a 71-year-old man with hilar thickening showed high AI-predicted malignancy (>85%), corroborated by biopsy-confirmed cholangiocarcinoma. In Case 2, a 75-year-old man with choledocholithiasis and distal duct inflammation had a low malignancy prediction (<5%) and benign histology, with no recurrence after 12 months. In Case 3, a 72-year-old man with obstructive jaundice, presented a nodular lesion with neovascularity; AI output showed >90% malignancy probability, confirmed as adenocarcinoma. In Case 4, a 69-year-old woman with autoimmune background and ambiguous DSOC appearance had a benign AI prediction, later supported by steroid-responsive IgG4-related cholangitis. In Case 5 involved a 64-year-old man with primary sclerosing cholangitis. AI yielded intermediate malignancy risk (∼55%) in a dominant stricture. Surgical resection revealed early cholangiocarcinoma. Discussion: This cases series from multiple centers demonstrated how integrating AI in real-time during cholangioscopy could enhance diagnostic performance, particularly when diagnoses are uncertain or high-risk cases. Although AI-assisted DSOC is not intended to replace histologic confirmation, it can aid in clinical decision-making by reducing ambiguity in interpretation and enhancing detection of lesions at earlier phases, although further prospective validation studies are still pending.