Accuracy of automated 3D biliary tract reconstruction compared to ERCP to assess Bismuth-Corlette classification in patients with perihilar cholangiocarcinoma

Abstract Introduction Accurate descriptions of the extent of disease in intrahepatic bile ducts are essential for managing patients with perihilar cholangiocarcinoma (pCCA). This study describes and assesses a new, automated Computed Tomography (CT)-scan-based 3D reconstruction process (3DR) of biliary tree in these patients compared with magnetic resonance cholangiopancreatography (MRCP) and endoscopic retrograde cholangiopancreatography (ERCP). Methods We studied consecutive patients with histologically proven hilar cholangiocarcinoma, who were referred for biliary drainage by ERCP in our center between 2015 and 2023. Automated reconstruction of bile ducts from CT images was performed using an artificial intelligence software (Synapse 3D 6.7, Fujifilm, Tokyo, Japan). We compared the Bismuth-Corlette classification assessed by ERCP (gold-standard), MRCP, CT-scan, and 3DR by computing exact percentage and weighted Kappa agreements. Results We included 62 patients. As compared with ERCP, CT-scan without 3DR had an exact agreement of 48% and had a moderate weighted Kappa agreement (weighted Kappa 0.47; 95%CI: 0.27-0.67). The MRCP had an exact agreement of 52% and a moderate weighted Kappa agreement (weighted Kappa 0.57; 95%CI: 0.39-0.76). The 3DR had an exact agreement of 60%, and a strong weighted Kappa agreement (weighted Kappa 0.72; 95%CI: 0.60-0.85). Conclusion 3DR is highly concordant with ERCP in assessing the extent of pCCA.