AI Helps Detect Pancreatic Cancer (IMAGE) Radiological Society of North America Caption False-negative (A, B) and false-positive (C, D) tumor segmentation by the segmentation model. Blue and yellow outlines indicate normal pancreas and tumor segmented with the segmentation model, respectively. Images in the left column are original unannotated CT scans. (A, B) Tumors (red outline) were not segmented by the segmentation convolutional neural network. The upstream pancreas shows secondary signs of pancreatic cancer, including dilation of the pancreatic duct with abrupt cutoff (arrowhead in A) and parenchymal atrophy with dilation of the pancreatic duct (arrowhead in B). (C) Collateral veins secondary to idiopathic portal vein thrombosis were incorrectly segmented as tumor by the segmentation model. (D) Pancreatic parenchyma adjacent to biliary stents (arrowhead) placed for relieving obstructive jaundice from hepatocellular carcinoma was incorrectly segmented as tumor by the segmentation model. Credit Radiological Society of North America Usage Restrictions May use with credit. License Licensed content Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.