New AI model spots pancreatic cancer up to 3 years earlier than human doctors in test

Researchers have developed an AI tool named REDMOD that analyzes CT scans to identify subtle structural irregularities in the pancreas that precede tumor development. Tested on 2,000 scans previously cleared as normal, the model successfully identified 73% of early-stage cancer cases, performing significantly better than human radiologists at detecting markers years before clinical diagnosis. While the model currently has a higher false-positive rate than human experts, researchers suggest it will serve as a powerful complementary tool to assist physicians in high-risk patient surveillance.

(Source：Live Science)