Development and validation of a machine learning-based nomogram for prediction of intrahepatic cholangiocarcinoma in patients with intrahepatic lithiasis

Background: Accurate diagnosis of intrahepatic cholangiocarcinoma (ICC) caused by intrahepatic lithiasis (IHL) is crucial for timely and effective surgical intervention. The aim of the present study was to develop a nomogram to identify ICC associated with IHL (IHL-ICC). Methods: The study included 2,269 patients with IHL, who received pathological diagnosis after hepatectomy or diagnostic biopsy. Machine learning algorithms including Lasso regression and random forest were used to identify important features out of the available features. Univariate and multivariate logistic regression analyses were used to reconfirm the features and develop the nomogram. The nomogram was externally validated in two independent cohorts. Results: The seven potential predictors were revealed for IHL-ICC, including age, abdominal pain, vomiting, comprehensive radiological diagnosis, alkaline phosphatase (ALK), carcinoembryonic antigen (CEA), and cancer antigen (CA) 19-9. The optimal cutoff value was 2.05 mu g/L for serum CEA and 133.65 U/mL for serum CA 19-9. The accuracy of the nomogram in predicting ICC was 82.6%. The area under the curve (AUC) of nomogram in training cohort was 0.867. The AUC for the validation set was 0.881 from The Second Affiliated Hospital of Wenzhou Medical University, and 0.938 from The First Affiliated Hospital of Fujian Medical University. Conclusions: The nomogram holds promise as a novel and accurate tool to predict IHL-ICC, which can identify lesions in IHL in time for hepatectomy or avoid unnecessary surgical resection.