Artificial intelligence in chronic and autoimmune pancreatitis: diagnosis, prognosis, and personalized management
Xiaoming Xu, Hualei Chen, Xi Zhang, Xuetao Wang, Yuanyuan Ding, Guobin Wang, Zhaoran Zhang
Chronic pancreatitis (CP) and autoimmune pancreatitis (AIP) have overlapping clinical and imaging features with pancreatic ductal adenocarcinoma (PDAC), resulting in frequent misdiagnosis and improper clinical treatment, and conventional diagnostic methods are subject to subjective factors, low accuracy and sampling errors. A review of the paradigm shift brought about by artificial intelligence (AI) in the diagnosis, prognosis and individualized management of CP and AIP. Based on AI, including deep learning and radiomics, has achieved a high-precision differential diagnosis and severity grading of CP and AIP by analyzing endoscopic ultrasound, CT, MRI and other imaging modalities, and integrates multi-source clinical, serological and omics data to further improve diagnostic efficiency. AI-powered digital pathology has realized quantitative histological analysis, and prognosis AI models can help predict complications such as exocrine pancreatic insufficiency and treatment non-adherence to support early intervention. In addition, we also address the current problems of AI in clinical translation, such as model overfitting and the “black box” issue, and indicate that prospective multicentre studies, explainable AI, and multimodal data integration will be the primary directions for future research, thereby promoting the development of precision medicine in pancreatology.