Diagnosis, differentiation and prevention in pancreatic diseases

Abstract

Pancreatic cancer is one of the deadliest cancers, largely because it is often diagnosed at a late stage. This dissertation investigates whether tumors are already visible on CT and MRI scans performed before diagnosis. In about half of these pre-diagnostic scans, suspicious abnormalities were retrospectively identified, while such findings were rare in healthy controls. However, these warning signs were frequently overlooked, often due to distracting findings or misinterpretation. This highlights the potential role of artificial intelligence (AI) as a “second pair of eyes.” A pilot study developed an AI model capable of automatically segmenting the pancreas on MRI, reducing analysis time sixteenfold while maintaining accuracy comparable to current literature. In addition, biomarkers for early detection were evaluated. A combination of methylated DNA markers from pancreatic juice and the blood marker CA19-9 showed high accuracy in detecting pancreatic cancer, including early-stage disease.

The second part focuses on predicting and preventing post-ERCP pancreatitis. Although combining a urinary test (trypsinogen-2) with clinical risk factors was not sufficient for clinical implementation, nationwide surveys show improved risk awareness and preventive strategies. Moreover, additional imaging (MRI or EUS) can prevent unnecessary ERCPs, reducing complications and improving patient safety.