Early Detection of Pancreatic Cancer: Opportunities Provided by Cancer-induced Paraneoplastic Phenomena and Artificial Intelligence

Abstract Objective: Pancreatic ductal adenocarcinoma (PDAC) is the most lethal cancer, with a 5-year survival rate of only 11%. Surgery is the only potential cure for PDAC, but approximately 85% of patients present with unresectable tumors at diagnosis. The difficulty in early detection is attributed to the fact that early PDACs cause few or nonspecific symptoms and are frequently obscure or even invisible in imaging studies such as computed tomography (CT). This review aims to briefly summarize the status of screening/surveillance for PDAC and elaborate on the potential windows of opportunity for early detection through PDAC-induced paraneoplastic phenomena and artificial intelligence (AI)-augmented image analysis. Data Sources: Relevant studies and review articles were searched in PubMed. Study Selection: Studies and articles on human subjects were selected. Results: Surveillance for high-risk individuals with imaging-based tools (endoscopic ultrasound and magnetic resonance image) is now advocated, whereas screening for asymptomatic general populations is not warranted at present. Paraneoplastic syndromes, including pancreatic cancer-associated diabetes and cachexia, are prevalent in PDAC patients and may provide windows of opportunity for early detection. S100A9 and galectin-3 are novel PDAC-derived factors mediating pancreatic cancer-associated diabetes and have shown promise in facilitating the early detection of PDAC. Novel computer-aided detection tools based on AI technologies, including deep learning and radiomic analysis with machine learning, have achieved accurate detection and might supplement human interpretation to improve the sensitivity for early PDAC on CT images. Conclusion: Novel blood-based biomarkers and AI-augmented image analysis may be complementary and hold promise for the early detection of PDAC.