Use of artificial intelligence in the management of cholangiocarcinoma: Current applications, evidence, and future perspectives

Cholangiocarcinoma is a highly aggressive and heterogeneous type of biliary cancer diagnosed late, with limited tissue availability for analysis, multiple staging challenges, and poor prognosis. In the case of ambiguous biliary strictures or in advanced stage disease, management should integrate endoscopic, radiological, pathological, molecular, and clinical information. In recent years artificial intelligence (AI) has been identified as an appropriate decision-support tool in this context. Deep learning approaches can support the recognition of malignant mucosal features during cholangioscopy and may assist endoscopists in targeting biopsies more accurately. In addition, radiomics and deep learning approaches can be used to diagnose the illness, classify tumors, assist in staging, estimate lymph node involvement, and assess recurrence risks with imaging methods including computed tomography (CT), magnetic resonance imaging (MRI)/magnetic resonance cholangiopancreatography (MRCP), and positron emission tomography (PET)/CT. Computational techniques might help to predict actionable genetic alterations, assist with biomarker selection, and integrate imaging analysis with histological and genomic data within pathology and molecular oncology. AI may also contribute to prognostic modeling and personalized treatment selection, including surgery, transplantation protocols, systemic therapy, targeted therapy, immunotherapy, locoregional therapy, biliary drainage, and surveillance planning. However, most of the available studies were retrospective, single-center, and insufficiently validated. Key obstacles consist of limited datasets, variability in disease characteristics, inconsistent reference standards, challenges in model interpretability, difficulties in workflow integration, regulatory concerns, and unpredictable clinical outcomes. This review outlines both current and developing uses of AI in cholangiocarcinoma, ranging from ambiguous biliary strictures to tailored treatment options, compares progress in this field with the more mature experience of radiology, pathology, and gastrointestinal endoscopy, and summarizes emerging guidance for the safe development, reporting, and governance of clinical AI. It emphasizes the necessity for prospective, externally validated multimodal AI systems that enhance multidisciplinary clinical decision-making.