A Disease That Arrives Silently

Pancreatic cancer is among the most aggressive and hardest-to-detect oncological diagnoses. According to NBC News, approximately 80 percent of patients are diagnosed only at an advanced stage, when treatment options are fundamentally limited. The five-year survival rate hovers at just around 13 percent.

The problem lies in the anatomy itself. The pancreas is located deep within the abdominal cavity, which complicates both physical examinations and early tumor detection on imaging methods. Unlike breast or colon cancer, there is no routine screening for the healthy population. Typical symptoms, such as abdominal pain or significant weight loss, often appear only when the tumor has already spread to other organs. According to estimates, pancreatic cancer could become the second most common cause of cancer-related death by 2030.

How the Mayo Clinic Model Works

Researchers from the prestigious Mayo Clinic in Rochester, Minnesota developed an artificial intelligence system that analyzes patient CT scans. A study published in the professional journal Gut shows that the model was able to identify abnormalities in patients who underwent examinations for other health issues and were only later diagnosed with pancreatic cancer.

The scientists trained the system on retrospective data — scans for which the final diagnosis was already known. They then compared the AI's results with radiologists' evaluations. The result was clear: artificial intelligence was approximately three times more successful at detecting early signs of the disease than the human eye.

What AI Sees and Humans Don't

One of the key discoveries was that the model could recognize specific abnormal cells in the pancreas that help the tumor evade the immune system. Scientists already knew these cells existed, but identifying them on routine CT scans had previously been a major problem. "We knew the signal was there. We just had to find a way to detect it," radiologist and study co-author Dr. Ajit Goenka told NBC News.

In practice, this means that AI detects changes in tissue that precede a visible tumor. In many cases, patients' scans appear completely normal even six months before diagnosis. A head start of up to three years could fundamentally change patients' prognoses.

Why This Is a Breakthrough

Early detection is critical in the case of pancreatic cancer. If doctors identify the disease before it spreads to surrounding organs or blood vessels, the number of patients eligible for surgery, chemotherapy, or radiotherapy increases significantly. "This could be a truly fundamental change in the area of early detection," said surgical oncologist Dr. Pam Hodulová from Moffitt Cancer Center.

This research comes at a time when the fight against pancreatic cancer is seeing several parallel breakthroughs. In recent weeks, attention was drawn to results of an experimental mRNA vaccine that extended patient survival, and the drug daraxonrasib, which in studies doubled the lifespan of patients with advanced disease. The U.S. Food and Drug Administration (FDA) even granted expanded access to this drug under a controlled regimen. AI diagnostics could thus form a key component of a comprehensive strategy — from prevention through early detection to targeted treatment.

Current Limitations and Future Applications

The Mayo Clinic model is not intended for mass population screening. It should primarily serve people at increased risk of developing pancreatic cancer — for example, patients with a family history, type 2 diabetes, or smokers. A positive finding in such a case would lead to further blood tests or more detailed imaging examinations.

"If a patient has symptoms and it really is pancreatic cancer, you don't need artificial intelligence," Dr. Goenka explained. The goal is therefore to catch the disease in asymptomatic individuals at a point when it is still treatable.

Czech and European Context

While the study is taking place in the United States, its impact could extend to Czech and European healthcare as well. In the Czech Republic, pancreatic cancer is among the oncological diseases with an unfavorable prognosis and overall occurs more frequently than in some other European countries. Although the Mayo Clinic system is not yet commercially available, the principles it is based on could be applied to data from European healthcare facilities.

At the same time, Europe is undergoing intensive development of regulations for artificial intelligence in healthcare. The European Parliament approved the AI Act last year, which classifies healthcare AI applications as high-risk systems. This means that any model intended for cancer diagnosis must undergo strict conformity assessment before being placed on the market. This process is lengthy, but it increases patient and physician trust in the technology.

Czech hospitals and research centers are already experimenting with artificial intelligence in oncology — for example, in detecting skin tumors or analyzing histological samples. Integration of advanced models for CT scan analysis would therefore not be a distant future, if their reliability in real-world operation is proven.

When Will the Technology Reach Practice

Despite promising results, experts are careful to set realistic expectations. The Mayo Clinic model itself is not yet heading into routine practice. The clinical study will monitor patients for several more years so researchers can verify how accurately the system works outside laboratory conditions. "In a disease where we've been wandering in the dark for decades, this is a milestone that shows us the finish line. But we still have to work toward it," said Dr. Goenka.

If clinical testing is successful, the first deployment could take place within specialized oncology centers. For ordinary patients, however, this means one thing: early diagnosis of pancreatic cancer might not be a matter of chance in a few years, but rather the result of systematic monitoring of risk groups using artificial intelligence.