AI Tool Helps Detect, Differentiate Pancreatic Lesions During Endoscopic Ultrasound (Medscape)

At the American College of Gastroenterology (ACG) 2024 Annual Meeting, researchers presented an AI tool that identifies and differentiates pancreatic cystic and solid lesions with remarkable accuracy during endoscopic ultrasound (EUS). Developed collaboratively by teams across the U.S., Portugal, Spain, and Brazil, the AI model utilizes a convolutional neural network trained on over 126,000 still images from 378 EUS exams. Results showed high accuracy, with the AI distinguishing between normal pancreatic tissue and lesions, including 99.1% accuracy for normal tissue, 99-99.8% for cystic lesions (M-PCLs and NM-PCLs), and 94% accuracy for solid lesions like pancreatic ductal adenocarcinoma (P-DAC) and neuroendocrine tumors (P-NETs).