Often nicknamed “silent cancer”, pancreatic cancer remains one of the deadliest: 13% of patients survive five years after diagnosis. The cause is very late detection, when the tumor has already grown.
A Taiwanese team claims to have taken a key step to change this scenario. She describes in the magazine Nature Communications A AI-boosted blood test
able to spot a pancreatic cancer at an early stage, with accuracy up to 94%.
An AI blood test for difficult-to-screen pancreatic cancer
The pancreas lies deep in the abdomen and its tumors cause few specific symptoms for a long time. Doctors have a blood marker, CA19‑9, but it lacks sensitivity and specificity for reliable screening in the early stages of the disease. As Dr. Yu-Ting Chang, professor of internal medicine (gastroenterology and hepatology) at National Taiwan University, points out, the biggest obstacle to recovery remains until then “to discover it too late“.
To address this gap, the National Taiwan University Hospital and Academia Sinica have developed the PanMETAI platform, which combines nuclear magnetic resonance (NMR) spectroscopy and artificial intelligence. From around 110 microliters of blood, it extracts 260,000 metabolic signals analyzed by the algorithm.
How PanMETAI uses AI to analyze blood
Concretely, the machine reads this chemical “fingerprint” of the blood by combining it with some simple clinical data: patient age, CA19‑9 level and level of a protein called Activin A. The TabPFN model then learns to distinguish healthy and sick subjects. “Our study shows how a triple interaction between clinical expertise, basic cancer research and advanced AI can bridge the gap between laboratory discovery and diagnostic application” said Dr. Chun-Mei Hu, assistant researcher at the Genomics Research Center at Academia Sinica and co-lead author of the study.
Researchers report high performance: in a blinded test conducted in Taipei, PanMETAI showed a sensitivity of 93% and a specificity of 94% in high-risk patients. During external validation on 322 people monitored in Lithuania, the test detected approximately 90% of cancers and 83% of non-diseased subjects. For stages I and II, the algorithm maintains an overall accuracy close to 92%.
What future for this AI blood test for pancreatic cancer?
For researchers, this type of test could serve as a non-invasive filter in people considered at high risk, before more serious imaging examinations. “By combining the power of AI and the wealth of metabolic information captured by NMR, we have created a tool capable of detecting pancreatic cancer at its earliest and most treatable stages“, detailed Dr. Yu‑Ting Chang.
If these results are confirmed in other studies and in real conditions, this type of tool could become a valuable screening filter for people at risk before more serious examinations. For the moment, the PanMETAI technology remains at the research stage, but its designers already hope to adapt it to other cancers, from the liver to the colon, with the ambition of detecting tumors well before they become visible or symptomatic.