AI Can Detect Pancreatic Cancer Years Before It Is Diagnosed

Advancements in Early Detection of Pancreatic Cancer

Pancreatic cancer is projected to become the second-leading cause of cancer-related deaths in the United States by 2030. This alarming trend is largely due to the fact that 85 percent of cases are diagnosed only after the disease has already spread. It’s a serious issue—we’re not catching it soon enough.

However, there’s some hope on the horizon with a new AI model developed by researchers at the Mayo Clinic and the University of Texas MD Anderson Cancer Center. This innovative tool, known as REDMOD (radiomics-based early detection model), focuses on analyzing CT scans of patients who were later diagnosed with pancreatic cancer.

The results are promising. REDMOD successfully identified the most common type of pancreatic cancer in nearly 75 percent of cases, roughly 16 months prior to formal diagnosis. That’s nearly double the rate achieved by specialists examining the scans without AI assistance.

Interestingly, there were instances where REDMOD detected concerning tissue patterns over two years ahead of time. The team believes that it could potentially spot cancer up to three years in advance.

Ajit Goenka, a radiologist at the Mayo Clinic, pointed out that “the greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable.” The AI model is capable of identifying the unique signals of cancer even in a pancreas that appears normal over time and across various medical environments.

For training, the researchers used a comprehensive dataset of 969 CT scans to help REDMOD learn how to identify subtle signs of early-stage cancer. Rather than hunting for obvious tumors, it looks for radiomic patterns—disruptions in the texture and structure of tissues that often escape the human eye.

After training, REDMOD was tested on a different set of scans: 63 from patients who went on to have cancer but were scanned pre-diagnosis, alongside 430 healthy individuals for comparison. It flagged 46 of the 63 scans as suspicious—a success rate of about 73 percent. Meanwhile, radiologists reviewing the same scans identified early signs in only about 39 percent of cases.

However, REDMOD did mistakenly categorize 81 of the healthy controls as suspicious, which means they might have received unnecessary additional tests in a real-world scenario. The AI’s performance was consistent across different datasets, too, showing its reliability.

Those with multiple scans found that the AI’s results were largely consistent even when scans were taken several months apart. The goal now is to validate REDMOD’s effectiveness in larger, more diverse patient groups and determine how easily it can be integrated into existing clinical workflows.

Researchers are optimistic about these early findings, believing that with further refinement, REDMOD might become a valuable tool in combatting one of the most lethal forms of cancer. They suggest that if it can analyze routine CT scans, perhaps for other conditions, it could facilitate early detection when curative treatments are still viable.

In summary, while there’s still a way to go before implementation, the potential of this AI system offers a glimmer of hope in the fight against pancreatic cancer, a disease that has long eluded timely diagnosis.