June 25, 2026
Pancreatic cancer remains one of the deadliest cancers today — with projections that it will become the second-leading cause of cancer death in the U.S. by 2030 since it so often goes undetected until its later stages. Artificial intelligence (AI)-powered detection methods under development at Mayo Clinic Comprehensive Cancer Center are changing that approach, helping physicians detect pancreatic cancer up to three years earlier, when it is more treatable.
Ajit H. Goenka, M.D., a nuclear medicine specialist and radiologist at Mayo Clinic in Rochester, Minnesota, leads a group of physicians, data scientists and other experts who developed an AI model to identify pancreatic cancer earlier than previously thought possible. The group's findings were published in a 2026 issue of Gut.
"The greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable," Dr. Goenka says. "This AI can now identify the signature of cancer from a normal-appearing pancreas, and it can do so reliably over time and across diverse clinical settings."
Overcoming barriers to early diagnosis of one of the deadliest cancers
Symptoms of pancreatic cancer often do not appear until later stages of the disease, when it has spread to surrounding organs. According to the National Cancer Institute, more than 85% of patients receive their diagnosis after the cancer has metastasized. Five-year survival rates for these patients remain below 15%.
Much of the challenge of diagnosing pancreatic cancer comes from limitations with standard imaging technology, primarily abdominal CT scans.
"The failure of conventional imaging to detect pancreatic ductal adenocarcinoma (PDA) at its preinvasive stage is a primary barrier to improving its otherwise poor rate of survival," Dr. Goenka says. "This 'invisibility' of early-stage disease on standard imaging, combined with the aggressive and rapid progression of this cancer, mandates the development of innovative solutions."
This disparity led the Mayo Clinic team to develop an AI model called Radiomics-Based Early Detection Model (REDMOD), which identifies subtle signs of disease on routine abdominal CT scans before tumors are visible, when curative treatment may still be possible. REDMOD is an investigational research tool and is not currently approved by the FDA for clinical use or population screening.
REDMOD extracts and analyzes quantitative radiomic features from CT scans, capturing subtle tissue texture changes that reflect early biological remodeling as cancer begins to develop. The model is designed to analyze routine contrast-enhanced CT scans in high-risk patients, such as those with new-onset diabetes, to identify early cancer signatures before any visible mass appears.
In this retrospective validation study, the model identified 73% of cancers a median of about 16 months before diagnosis, a rate nearly double that found by expert radiologists reviewing the same scans without AI assistance. In CT scans from more than two years before diagnosis, REDMOD identified nearly three times as many early cancers that would otherwise go undetected. These predictions also remained stable over time: In patients with multiple scans, REDMOD produced 90% to 92% concordance across serial scans, supporting its use for longitudinal monitoring and early detection.
The team used REDMOD to analyze nearly 2,000 CT scans, including scans that were originally interpreted as normal but were obtained from patients later diagnosed with pancreatic cancer. They validated the model across CT scans from multiple institutions, imaging systems and protocols, demonstrating consistent performance and the ability to correctly identify most patients who did not go on to develop pancreatic cancer (specificity of 81% in the multi-institutional cohort and 87.5% on an independent public NIH dataset). Researchers note that prospective testing and evaluation in more diverse patient populations are needed to confirm clinical utility.
"REDMOD's demonstrated ability to consistently detect these occult signals on a large dataset establishes a robust foundation for AI-augmented early detection," Dr. Goenka says. "This work overcomes key barriers in the field by providing a scalable, objective tool that addresses a critical diagnostic gap, moving from a late-stage symptomatic diagnosis to proactive preclinical interception, offering tangible hope for improving outcomes in this challenging disease."
Improving outcomes in pancreatic cancer detection
In the future, REDMOD could serve as an advanced risk-enrichment tool.
"A 'high-risk' flag from the model could prompt further investigation with molecular imaging," Dr. Goenka says. "Ongoing research is investigating whether the AI signal localizes to the eventual tumor site, which could have implications for targeted tissue sampling in the future."
Tools such as REDMOD also could work hand-in-hand with new treatment options. With rapid evolution in the therapeutic landscape for early-stage PDA, emerging interception strategies, including KRAS-directed therapies and investigational immunologic approaches, reinforce the clinical value of identifying high-risk individuals before a mass forms. REDMOD, which offers a robust risk assessment for the pancreas, could help to support this need.
As a next step, researchers are continuing this work into clinical testing through Artificial Intelligence for Pancreatic Cancer Early Detection (AI-PACED). AI-PACED is the prospective clinical trial designed to evaluate REDMOD in real-time clinical practice. The study enrolls individuals at elevated risk of pancreatic cancer based on new-onset diabetes and validated risk scores and combines serial AI-augmented CT imaging with biobanking to assess detection performance, false-positive rates and clinical outcomes.
For more information
Mukherjee S, et al. Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability. Gut. In press.
Refer a patient to Mayo Clinic.