Pancreatic cancer is one of the most difficult cancers to detect early. In many cases, the disease does not cause clear symptoms in its early stages, and by the time doctors discover it, the cancer may already have progressed. Now, researchers are testing a new approach that could help change that. Scientists have developed an artificial intelligence system that can look at routine CT scans and identify very subtle changes in the pancreas that may be linked to cancer—even when those changes are not obvious to doctors.
The technology is called REDMOD, short for Radiomics-based Early Detection Model. Researchers say it could eventually help identify pancreatic cancer months or even years before a patient receives a clinical diagnosis.
Why Early Detection Matters
Pancreatic ductal adenocarcinoma, or PDAC, is the most common form of pancreatic cancer. One of the biggest challenges is that early pancreatic cancer can be difficult to see and may not produce obvious warning signs. This means that patients are often diagnosed when the disease is already more advanced. Earlier detection could make an important difference because treatment options are generally more useful when cancer is found before it has spread.
The researchers behind REDMOD are therefore asking an important question: What if computers could detect tiny changes that the human eye cannot see?
How Does REDMOD Work?
REDMOD uses a field of technology called radiomics. Instead of simply looking at a CT scan and asking whether a tumor is visible, radiomics analyzes detailed patterns within medical images. These patterns can include subtle differences in tissue texture and structure. Think of it like looking at a photograph.
A person might look at the picture and see a normal-looking area. A computer, however, can analyze thousands of tiny patterns that are difficult for humans to notice. REDMOD also automatically identifies and outlines the pancreas in the CT scan. This is important because it reduces the need for doctors to manually define the organ before the computer analyzes it.
The AI Found Signs Before Cancer Was Diagnosed
Researchers tested the system using CT scans from people who were initially considered to have no visible pancreatic cancer but were later diagnosed with the disease. The scans were taken at different points before diagnosis. Some were performed between three and 12 months before the cancer was diagnosed. Others were taken one to two years earlier, while some were more than two years before diagnosis.
According to the study, REDMOD identified subtle imaging patterns associated with future pancreatic cancer an average of 475 days before clinical diagnosis. That is potentially significant. It suggests that medical images taken for other reasons could contain information about early disease that conventional examination may overlook.
AI Compared With Expert Radiologists
The researchers also compared REDMOD’s performance with experienced radiologists. In their testing, the AI system detected more of the cancer cases than the radiologists when looking for these very early signs.
REDMOD achieved a sensitivity of about 73%, compared with 39% for the radiologists in the study. For cases where the CT scan was taken more than two years before diagnosis, the difference was also notable: the model achieved about 68% sensitivity, compared with 23% for the radiologists.
These numbers do not mean that AI can replace doctors. Instead, they show how artificial intelligence could potentially become another tool that helps medical professionals identify patients who may need closer examination.
There Is Still a Long Way to Go
Despite the promising results, REDMOD is not yet a standard diagnostic tool. The researchers say more testing is necessary, particularly in groups of people who may have a higher risk of pancreatic cancer. The study also had limitations, including a lack of ethnic diversity among participants. More research will be needed to determine how well the system works across different populations and healthcare settings.
Most importantly, a computer prediction would still need to be evaluated by medical professionals. AI can identify patterns, but additional medical tests would be needed before a cancer diagnosis could be confirmed.
A Possible New Role for AI in Medicine
The most interesting part of this research is not simply that an AI system can “see” cancer. It is that AI may be able to recognize early biological changes before cancer becomes obvious on a conventional scan.
If future studies confirm these findings, systems like REDMOD could potentially be used alongside existing medical imaging to help doctors identify patients who require further investigation. This could move pancreatic cancer detection toward a more proactive approach—looking for warning signs before the disease becomes advanced.
For now, the technology remains under investigation. But the research highlights a growing role for artificial intelligence in medicine: helping doctors find information hidden inside data that the human eye may not easily recognize. And when it comes to diseases where timing matters, finding a warning sign earlier could make a meaningful difference.
Read the Original: Researchers Test a New Way To Find Pancreatic Cancer Before It’s Too Late
Read Related: A man who received the world’s first double arm and shoulder transplant amazed doctors with his recovery
