Artificial Intelligence Detects Pancreatic Cancer Years Before Diagnosis

According to research published in the journal Gut, an artificial intelligence system called Radiomics-based Early Detection MODel (REDMOD) can detect early-stage signs of pancreatic ductal adenocarcinoma an average of 475 days before clinical diagnosis. Outperforming expert radiologists, the system analyzes subtle tissue patterns that go unnoticed in conventional CT scans.

According to research published in the journal Gut, an artificial intelligence model called REDMOD can detect early signs of pancreatic ductal adenocarcinoma, the most common type of pancreatic cancer, by detecting very subtle changes in pancreatic tissue. The system captured early warning signs an average of 475 days before patients received a clinical diagnosis, demonstrating a higher success rate than experienced radiologists. Researchers note that this technology could increase treatment chances by raising early-stage detection rates.

The system examines radiomic tissue texture patterns that are not easily visible with standard methods in computed tomography scans. It also automatically segments the borders of the pancreas, eliminating the need for experts to perform manual drawings. In an examination of historical scanned data from 219 patients, REDMOD achieved a sensitivity rate of 73%, while the sensitivity of radiologists remained at 39%. In scans taken more than two years prior, the artificial intelligence's success rate reached nearly three times that of radiologists.

Researchers emphasize that the technology must be tested prospectively in broader populations with ethnic diversity and in high-risk patients before entering routine clinical practice. While the study confirms that REDMOD holds potential in catching early-stage signs during the asymptomatic period, it indicates that additional research is needed to confirm its clinical utility.