Artificial intelligence for pancreatic cancer guided by clinical need

M. Schuurmans

• Promotor: G. Litjens and H. Huisman
• Copromotor: J. Hermans
• Graduation year: 2025
• Radboud University

Abstract

This thesis focuses on the diagnostic phase of pancreatic cancer care, where AI has the potential to enhance detection and disease stratification. We explore the current state of AI in pancreatic cancer and identify gaps in clinical relevance. Additionally, we develop a stateof-the-art pancreatic cancer detection model for CECT and benchmark it against the largest international cohort of radiologists published to date. In parallel, we analyze diagnostic uncertainty from both AI and human readers to enhance the safe integration of informed clinical decisions. Finally, we evaluate whether AI-based stratification better reflects patient outcomes than conventional staging, with the aim of supporting more personalized treatment planning.