A new artificial intelligence system may significantly advance the early detection of pancreatic cancer, identifying subtle warning signs in CT scans that were previously considered normal.
The model, known as REDMOD, was developed by researchers at Mayo Clinic and the University of Texas MD Anderson Cancer Center. According to the study, it successfully detected the disease in nearly three out of four cases, on average about 16 months before diagnosis.
Pancreatic cancer remains one of the deadliest forms of cancer, largely due to late detection. In the United States, it is projected to become the second leading cause of cancer-related deaths by 2030, as most cases are diagnosed at an advanced stage.
REDMOD relies on radiomic analysis techniques and was trained on hundreds of pancreatic CT scans. Rather than searching for obvious tumors, it focuses on subtle changes in tissue texture and structure – patterns often too faint for the human eye to detect.
During testing, the system was evaluated using scans from patients who were later diagnosed with pancreatic cancer, as well as from healthy individuals. It achieved a detection rate of 73%, demonstrating its ability to identify early suspicious patterns.
Notably, all of these scans had initially been deemed clear by radiologists. Even when re-examined with AI assistance, human detection rates remained significantly lower than those of the model.
In some cases, REDMOD identified signs of the disease more than two years before diagnosis. Researchers believe that with further development, the system could potentially detect pancreatic cancer up to three years in advance.
While the findings are promising, experts emphasize the need for larger and more diverse studies before the tool can be integrated into clinical practice. If validated, it could leverage routine imaging performed for other reasons, enabling earlier diagnosis and significantly improving treatment outcomes.