📰 Body
A breakthrough study from the Mayo Clinic is bringing new hope to the early detection of pancreatic cancer. The research team’s AI system, REDMOD, can accurately identify pancreatic cancer signatures from routine CT scans an average of 16 months before clinical diagnosis — achieving a detection rate nearly double that of human radiologists.
REDMOD: Redefining Early Cancer Screening
Pancreatic cancer, often called the “silent killer,” remains one of the world’s deadliest malignancies. Because early symptoms are minimal, most patients are diagnosed at advanced stages, with a five-year survival rate below 13%. This grim reality has driven researchers to seek methods for much earlier detection.
REDMOD (Radiomic Early Detection Model) differs from traditional AI diagnostic tools in that it does not look for obvious tumor imaging. Instead, it analyzes radiomic patterns — subtle abnormalities in tissue texture and structure — to catch early cancerous changes invisible to the human eye.
The research team trained REDMOD using 969 pancreatic CT scans, teaching it to recognize the nuanced changes that mark the earliest stages of cancer development. In subsequent testing, the model performed blind evaluations on 63 CT scans from patients who were eventually diagnosed with cancer (but were healthy at scan time) and 430 healthy control samples.
Key Data: Nearly Double the Expert Detection Rate
The results showed that REDMOD successfully identified the most common form of pancreatic cancer in nearly three out of four cases, providing an average warning of approximately 16 months before diagnosis. In some scans, the AI detected suspicious tissue patterns more than two years ahead of diagnosis. The research team believes the system could theoretically extend the detection window to three years.
By comparison, specialist physicians reviewing the same scans without AI assistance achieved a detection rate of only about 40%.
Dr. Ajit Goenka, a radiologist and nuclear medicine specialist at the Mayo Clinic, stated: “The greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable. This AI can now identify the signature of cancer from a normal-appearing pancreas, and it can do so reliably over time and across diverse clinical settings.”
The Power of Radiomics
REDMOD’s core innovation lies in its analytical approach. Most cancers begin when normal cells acquire DNA mutations that affect how they grow and divide, but it can take years for these changes to develop into tumors large enough to cause symptoms or appear clearly on imaging. REDMOD uses deep learning to capture the early impact these mutations have on tissue microstructure, enabling ultra-early warning.
Researchers note that this radiomic approach is not limited to pancreatic cancer and could potentially be extended to other malignancies that are difficult to detect early, such as ovarian cancer and certain types of lung cancer.
Next Steps: Clinical Validation and Rollout
REDMOD is currently in the research phase and requires validation through larger-scale prospective clinical trials before receiving FDA approval for clinical use. The team plans to launch a multi-center trial within the next 12 months, encompassing diverse populations and healthcare facilities.
If REDMOD successfully transitions to clinical practice, it could transform the paradigm of pancreatic cancer care — shifting from “discovered at late stage” to “detected while still curable,” potentially saving tens of thousands of lives each year.
Source: ScienceAlert