Endoscopic Ultrasound-Based Artificial Intelligence Diagnosis of Pancreatic Cystic Neoplasms

Abstract

Pancreatic cystic neoplasms (PCNs) are precursors of pancreatic cancer, and the rate of their incidental detection has gradually increased recently with a reported prevalence from 2.4 to 13.5%. However, accurate diagnosis can be challenging because PCNs have morphologies ranging from benign to malignant disease, and as for other cancers, precise and timely management of premalignant PCN is essential to prevent malignant transformation. Endoscopic ultrasound (EUS) is a useful tool for the differential diagnosis PCN and treatment decision-making because its imaging features predict malignant transformation. However, its performance is suboptimal, and its accuracy for differentiating mucinous pancreatic cysts and other PCNs is only 65-75%, which has increased interest in the application of artificial intelligence (AI). AI has already provided tools that have improved diagnostic accuracies for many cancers, including colon, lung, and breast cancer, and recent studies have shown AI has the potential to differentiate mucinous and non-mucinous tumors and stratify the malignant potentials of PCNs. This article provides a review of the literature on EUS-based AI studies of PCNs.