AI-Assisted Differentiation of Autoimmune Pancreatitis and Pancreatic Cancer Using EUS: A Multicenter Ambispective Study

Focal autoimmune pancreatitis (f-AIP) and pancreatic ductal adenocarcinoma (PDAC) exhibit substantial overlap in clinical manifestations and imaging features, posing a major diagnostic challenge in pancreatic diseases. Given the marked differences in treatment strategies and prognosis, accurate differentiation is essential to avoid unnecessary surgery and ensure timely initiation of anticancer treatment. This study aims to collect endoscopic ultrasound (EUS) images from patients with confirmed f-AIP or PDAC across multiple centers, using a previously established multicenter database. Data collected before January 2026 will be used to develop an artificial intelligence (AI) model based on EUS radiomics for differential diagnosis. Subsequently, prospectively collected data from participating centers from 2026 onward will be used for model testing and validation. The diagnostic performance of the model will be evaluated and compared with that of endoscopists with different levels of experience. Furthermore, this study will investigate the clinical value of interpretable radiomic features, with particular emphasis on challenging diagnostic scenarios, including IgG4-negative AIP, AIP with elevated carbohydrate antigen 19-9 (CA19-9) levels, and PDAC with imaging features resembling those of AIP. The ultimate goal is to provide a reliable tool for accurate, noninvasive preoperative differential diagnosis while promoting the standardization of multicenter AIP databases and the sharing and utilization of imaging data resources.