AI system REDMOD detects pancreatic cancer signs 15 months early

A new artificial intelligence system named REDMOD has demonstrated the ability to identify subtle signs of pancreatic ductal adenocarcinoma in standard abdominal CT scans, often before they are visible to human experts. According to a study published in the journal Gut, the AI detected abnormalities an average of 475 days, or approximately 15 months, before a formal clinical diagnosis.

In comparative tests, REDMOD achieved a sensitivity rate of 73%, significantly higher than the 39% sensitivity rate observed in experienced radiologists. When analyzing scans taken more than two years prior to diagnosis, the AI maintained a 68% detection rate, nearly triple the 23% rate achieved by physicians. The system utilizes radiomics to analyze minute texture changes in pancreatic tissue that are typically imperceptible during conventional manual reviews.

While the results are promising, researchers noted that the technology requires further validation. Current studies have had limited population diversity, and future research must focus on prospective testing within broader populations and high-risk groups, such as individuals experiencing unexplained weight loss or new-onset diabetes, before it can be implemented in routine clinical practice.