Pancreatic cancer is one of the most dangerous forms of cancer in the world. Many patients do not know they have the disease until it has already spread. Because of this, the survival rate is very low.
Only about 13% of patients live for five years after diagnosis. Doctors and scientists have long searched for a reliable way to detect pancreatic cancer earlier, when treatment has a much better chance of working.
Now a research team in Taiwan has developed a new artificial intelligence system that may help solve this problem. The system, called PanMETAI, can analyze the chemical signals in a small blood sample and detect pancreatic cancer at a very early stage.
In testing, the tool achieved diagnostic accuracy of up to 94%. The research describing this technology was published in the scientific journal Nature Communications.
Pancreatic cancer begins in the tissues of the pancreas, an organ located behind the stomach. The pancreas plays an important role in digestion and blood sugar control. It releases enzymes that help break down food and hormones such as insulin that regulate glucose levels in the body.
When cancer forms in the pancreas, it often grows silently for a long time without causing clear symptoms. By the time people feel pain or notice changes in their health, the disease may already be advanced.
Doctors currently use several methods to look for pancreatic cancer. Imaging tests such as CT scans or MRI scans can reveal tumors, but these are usually used after symptoms appear. Blood tests are also used, especially a marker called CA19-9.
However, this marker is not very reliable for early detection. Some patients with cancer do not show high levels of CA19-9, while others without cancer may show elevated levels. Because of these limitations, researchers have been looking for better tools.
The new PanMETAI system was developed by scientists from National Taiwan University Hospital and Academia Sinica. The team combined knowledge from medicine, chemistry, and artificial intelligence.
Their idea was to study the body’s metabolism, which includes all the chemical reactions that happen inside cells. These reactions produce many small molecules known as metabolites. Changes in these molecules can sometimes reveal the presence of disease before obvious symptoms appear.
To examine these chemical signals, the researchers used a technology called nuclear magnetic resonance spectroscopy, or NMR. This technique can measure many different metabolites in a blood sample at the same time. Each person’s blood contains a complex pattern of these chemicals, often called a metabolic fingerprint. Diseases such as cancer can change this fingerprint in subtle ways.
The PanMETAI system analyzes these metabolic fingerprints using artificial intelligence. In the study, the researchers used only about 500 microliters of blood serum, which is a very small amount.
From this sample, the NMR system captured more than 260,000 metabolic signals. These signals were then analyzed using a powerful machine learning model known as TabPFN.
The AI system did not rely only on metabolic data. The researchers also included important clinical information such as the patient’s age, levels of the CA19-9 marker, and another protein biomarker called Activin A. By combining all of this information, the system was able to recognize patterns linked to pancreatic cancer.
When the researchers tested the system on patients in Taiwan, the results were extremely strong. The model achieved an area under the curve, or AUC, of 0.99.
This means the system could almost perfectly distinguish people with pancreatic cancer from those without the disease. A value of 1.0 would represent perfect accuracy, so this result shows very high diagnostic performance.
To make sure the system was not limited to one population, the researchers also tested it on an independent group of patients in Lithuania. This group included 322 participants from the Lithuanian University of Health Sciences.
Even though the population had a different genetic background and lived in a different region of the world, the system still performed well. In this test, it achieved an AUC of 0.93, showing that the method can work across diverse populations.
One of the most important achievements of the study was the system’s ability to detect early-stage pancreatic cancer. Early-stage disease, known as Stage I or Stage II, is much easier to treat than advanced cancer.
However, detecting it has been extremely difficult. The researchers found that the metabolic information captured by NMR played a key role in identifying these early changes.
The study showed that several metabolites changed in patients with early cancer. For example, levels of HDL cholesterol and glutamine were lower than normal. At the same time, levels of lactic acid, glucose, and glutamic acid were higher. These metabolic shifts may reflect how cancer cells alter the body’s energy use and chemical balance.
Another advantage of the PanMETAI system is that it does not require extremely large datasets to train the AI model. In many medical AI projects, researchers need thousands of patient samples to achieve reliable results.
However, this platform was able to reach stable accuracy of about 90% with only around 50 training cases. This feature makes the technology more practical for hospitals or research centers that may not have access to large patient databases.
The researchers believe the system could become a useful screening tool in the future. A simple blood test could be used to identify people who are at high risk of pancreatic cancer.
Those individuals could then receive further medical testing, such as imaging scans, to confirm the diagnosis. Detecting the disease earlier could allow doctors to start treatment sooner and improve survival rates.
Dr. Chun-Mei Hu from the Genomics Research Center at Academia Sinica explained that the project brought together several different areas of expertise. Clinical doctors provided knowledge about patient care and disease patterns.
Cancer researchers studied metabolic changes in tumors. AI specialists developed advanced machine learning models to analyze the complex data.
Dr. Chao-Ping Hsu from the Institute of Chemistry at Academia Sinica noted that the work highlights how artificial intelligence can help scientists understand extremely complex biological data. By combining advanced computation with medical knowledge, researchers may be able to develop powerful new diagnostic tools.
Professor Yu-Ting Chang from National Taiwan University also emphasized that the goal is to translate the technology into real clinical practice. If the system can be widely used in hospitals, it could help doctors detect pancreatic cancer much earlier and provide patients with better chances for treatment.
Overall, the study represents an important step forward in the effort to detect pancreatic cancer earlier. The combination of AI, metabolic analysis, and clinical data shows strong promise as a new diagnostic strategy.
However, the research is still at an early stage. Larger studies involving more patients from different regions will be needed before the technology can become a standard medical test.
Even so, the results are encouraging. If future studies confirm these findings, a simple blood test powered by artificial intelligence could one day help doctors detect pancreatic cancer before it becomes life‑threatening.
Such progress could dramatically improve survival rates and change the way this devastating disease is diagnosed and treated.
If you care about cancer, please read studies that a low-carb diet could increase overall cancer risk, and berry that can prevent cancer, diabetes, and obesity.
For more health information, please see recent studies about how drinking milk affects the risks of heart disease and cancer and results showing vitamin D supplements could strongly reduce cancer death.
Copyright © 2026 Knowridge Science Report. All rights reserved.
