An artificial intelligence framework named REDMOD can identify the earliest signs of pancreatic cancer on routine CT scans an average of 475 days before typical clinical diagnosis, according to a study published in the journal Gut. Developed by researchers at Mayo Clinic, REDMOD detects subtle tissue texture changes associated with pancreatic ductal adenocarcinoma that are not visible to the human eye or traditional imaging techniques.
REDMOD exhibits nearly twice the sensitivity of experienced radiologists, achieving an accuracy rate of 73% in detecting preclinical disease, compared to 39% for radiologists. For cases identified more than two years before clinical diagnosis, REDMOD’s accuracy soars to 68%, while radiologists’ accuracy drops to 23%.
The model also classified over 81% of scans as cancer-free in an independent cohort of 539 patients and recorded an accuracy of 87.5% in the NIH-PCT dataset of 80 patients. Additionally, REDMOD demonstrated consistent results upon retesting the same patients months later, achieving consistency rates of 90 to 92%.
Pancreatic cancer remains a critical health issue, being the deadliest major cancer in the United States with a five-year survival rate of just 13%. It is the third-leading cause of cancer-related deaths, with an estimated 67,530 new diagnoses and 52,740 deaths expected in 2023. If detected at a localized stage, the five-year survival rate increases significantly to 44%.
The researchers emphasize that REDMOD’s ability to detect cancer early could greatly enhance the probability of cure and improve survival outcomes. “This temporal window holds profound significance, as attaining such early detection would substantially augment the probability of cure and improved survival,” the researchers stated.
REDMOD utilizes automated pancreatic segmentation, reducing the reliance on radiologists to manually outline pancreatic borders, a process that is often time-consuming. However, the team noted that prospective validation in high-risk patients is necessary before REDMOD can be broadly implemented in clinical practice. They referred to this tool as a significant advance in shifting pancreatic cancer diagnosis from late-stage symptomatic detection to early pre-clinical interception.
