Congenital biliary dilatation necessitates timely intervention owing to potential complications. This study endeavors to enhance diagnostic precision using quantitative three-dimensional morphological characteristics. Objectives involve developing models to differentiate congenital from secondary biliary dilatation and identify intrahepatic involvement. Employing machine learning, robust diagnostic models aim to elevate clinical detection rates and improve accuracy.

Congenital biliary dilatation is a primary anomaly affecting the biliary tract. It can involve the extrahepatic bile ducts, intrahepatic bile ducts, or the entire biliary system, including the common bile duct. Patients with congenital biliary dilatation exhibit abnormal expansion of the bile duct system, which can lead to complications such as bile duct stones, pancreatic inflammation, and even bile duct cancer. Timely and accurate diagnosis, followed by surgical intervention to remove the dilated bile duct lesion, is crucial for the treatment of choledochal dilation. However, the differentiation of congenital biliary dilatation in clinical practice poses challenges, primarily due to the limitations of subjective physician experience and macroscopic imaging features, making it difficult to achieve high sensitivity in discerning congenital biliary dilatation. Particularly, in distinguishing between congenital biliary dilatation and secondary biliary dilatation, the similarities of the bile ducts limit the precision of clinical decisions. Therefore, this study aims to address the current challenges in the differential diagnosis of congenital biliary dilatation and secondary biliary dilatation by quantitatively describing the morphology of dilated bile ducts. Moreover, this study plans to build a predictive model of intrahepatic bile duct dilatation to provide more comprehensive clinical support. Specifically, the research objectives are outlined as follows:

1. Establish a diagnostic model for congenital biliary dilatation utilizing three-dimensional morphological characteristics, especially quantitative shape- and diameter-based characteristics, to enhance the accurate discrimination between congenital biliary dilatation and secondary biliary dilatation.
2. Develop a model for identifying intrahepatic involvement of congenital biliary dilatation, aiming to provide more precise information for surgical planning and supportive treatment.
3. Construct robust diagnostic models using machine learning with quantitative three-dimensional morphological characteristics, aiming to increase clinical detection rates and accuracy, thereby achieving risk stratification for patients with biliary dilatation.

Inclusion Criteria:

• The patients with biliary dilation who underwent contrast-enhanced computed tomography (CECT) at Beijing Tsinghua Chang Gung Hospital from 2014 to 2022.

Exclusion Criteria:

• Patients without pre-operative CECT scans or developing cholangiocarcinoma due to congenital biliary dilatation.

Patients with congenital biliary dilatation diagnosed according to the Japanese Study Group on Congenital Biliary Dilatation (JSCBD) guideline or secondary bile duct dilatation due to choledocholithiasis or malignancies such as hilar cholangiocarcinoma, pancreatic carcinoma, and distal cholangiocarcinoma who underwent contrast-enhanced computed tomography (CECT) at Beijing Tsinghua Chang Gung Hospital from 2014 to 2022.