AI Detects Pancreatic Cancer Earlier Than Doctors

A new artificial intelligence system called REDMOD may be able to detect pancreatic cancer earlier than current medical methods, according to research published in Gut.

The model identifies subtle tissue changes linked to pancreatic ductal adenocarcinoma, the most common and deadliest form of the disease, which are typically not visible on routine scans.

Pancreatic ductal adenocarcinoma has a low survival rate because it is often diagnosed at an advanced stage. Early disease rarely produces symptoms or visible changes, making timely detection difficult.

To address this, researchers developed the Radiomics-based Early Detection Model, known as REDMOD.

REDMOD analyzes patterns in tissue texture, referred to as radiomics, to detect early cancer signals that are not visible on standard CT scans.

The system also uses automated pancreatic segmentation to accurately outline the pancreas and separate it from surrounding tissues, removing the need for manual input.

The model was tested using abdominal CT scans from 219 patients across multiple hospitals who were initially considered disease-free but were later diagnosed with pancreatic cancer.

Among them, 87 patients had scans taken 3 to 12 months before diagnosis, 76 had scans from 12 to 24 months earlier, and 56 had scans taken more than 24 months before diagnosis, up to about three years.

In 64% of cases, the cancer was located in the head of the pancreas. Results were compared with scans from 1,243 individuals who did not develop pancreatic cancer within three years.

The average age of patients later diagnosed was 69, while the comparison group averaged 64.

REDMOD identified early signs of cancer an average of 475 days before clinical diagnosis. Researchers said this time window could significantly improve treatment outcomes. The model showed higher sensitivity than radiologists, correctly identifying 73% of true cases compared to 39%.

For cases detected more than two years before diagnosis, REDMOD achieved 68% accuracy compared to 23% for radiologists.

In further testing, REDMOD correctly classified just over 81% of scans from an independent group of 539 patients as cancer-free.

It also reached 87.5% accuracy in a dataset from the National Institutes of Health, which included 80 patients. The model produced consistent results in 90 to 92% of cases when earlier scans were reviewed. Researchers noted limitations, including limited ethnic diversity among participants.

The study concludes that REDMOD can identify imaging signatures of stage 0 pancreatic cancer in otherwise normal pancreatic tissue, with performance exceeding that of expert radiologists.

However, further testing in high-risk patients is required before it can be used widely in clinical practice.

Researchers said the system could help shift diagnosis from late-stage detection to earlier intervention, improving survival outcomes.