In a groundbreaking advancement for oncology and radiology, researchers at the Mayo Clinic have unveiled a next-generation artificial intelligence (AI) model that can detect pancreatic cancer on routine abdominal CT scans long before traditional clinical diagnosis. This model identifies subtle biomarkers and minute changes in pancreatic tissue architecture that are imperceptible to human radiologists, offering a potential paradigm shift in early pancreatic cancer detection. According to findings published recently in the esteemed journal Gut, the AI system, designated the Radiomics-based Early Detection Model (REDMOD), demonstrated remarkable proficiency in identifying malignancy signatures up to three years prior to the clinical manifestation of visible tumors.
Pancreatic cancer has long been notorious for its elusive early presentation and poor prognosis, with the majority of patients diagnosed only once the disease has metastasized. This late detection contributes to dismal survival statistics, with five-year survival rates languishing below 15%. The urgency of developing earlier diagnostic tools is underscored by projections indicating pancreatic cancer will become the second leading cause of cancer-related deaths in the United States by 2030. The Mayo Clinic’s REDMOD AI system directly addresses this unmet need by leveraging advanced computational techniques, including radiomics, which quantitatively analyze tissue texture, heterogeneity, and morphology from cross-sectional imaging.
The innovation embedded in REDMOD lies in its ability to extract and interpret hundreds of quantitative imaging features that capture the biological microenvironmental alterations preceding macroscopic tumor formation. Essentially, this approach translates the scanning data into high-dimensional datasets that delineate subtle phenotypic shifts in pancreatic tissue. Importantly, the AI model was trained and validated on nearly 2,000 CT scans sourced from multiple institutions, encompassing various imaging hardware and acquisition protocols. This rigorous multi-institutional validation ensures that the AI’s diagnostic precision is not limited by scanner variability or institutional bias, highlighting its potential for broad clinical deployment.
Furthermore, the Mayo Clinic team emphasized REDMOD’s longitudinal stability — the model generates consistent diagnostic outputs across sequential scans taken months apart from the same patient. This feature is critical for monitoring high-risk individuals over time and could facilitate ongoing surveillance strategies to detect the earliest signs of neoplastic transformation. Of particular note, REDMOD automatically processes routine abdominal CT images obtained for disparate clinical indications, obviating the need for specialized imaging or labor-intensive manual annotations. This design inherently supports integration into existing clinical workflows, enabling opportunistic screening and risk stratification.
The model’s performance was starkly superior to that of expert radiologists when analyzing scans originally interpreted as normal. REDMOD identified 73% of pancreatic cancers that were retrospectively known to be present up to 16 months before clinical diagnosis, effectively doubling the detection rate achievable through conventional human review alone. The detection advantage expanded further at earlier time points; more than two years prior to diagnosis, the AI model uncovered nearly three times as many hidden cancers as specialists. These compelling metrics underscore the transformative potential of AI-enhanced radiomics in addressing the longstanding challenge of occult pancreatic cancer.
REDMOD’s application could carry particular significance for high-risk populations, such as individuals with new-onset diabetes or a familial predisposition to pancreatic cancer. These groups often undergo abdominal imaging for various indications, creating an opportunity for “second-look” AI analysis to identify early cancer signatures. By flagging cases with elevated risk before overt tumor development, REDMOD raises the prospect of earlier intervention and potentially curative treatment, heralding a shift from reactive to predictive oncology.
Coupled with clinical expertise, the AI model is currently undergoing prospective clinical testing within the Artificial Intelligence for Pancreatic Cancer Early Detection (AI-PACED) study. This initiative evaluates how REDMOD-guided interpretation can be integrated into patient management, measuring outcomes such as early detection rates, false-positive frequency, and downstream clinical benefit. The prospective design and real-world setting serve to validate the utility of AI not just as a retrospective research tool but as a practical adjunct in routine clinical care.
This research is a flagship component of the Mayo Clinic’s Precure initiative, which aims to preempt disease through early biological detection and intervention, fundamentally reshaping the future of healthcare. It also aligns closely with Mayo Clinic’s Clinical Impact strategy focused on accelerating the translation of scientific discoveries into tangible patient benefits. Such institutional commitment underscores the transformative nature of this AI development and its consonance with the evolving landscape of precision medicine.
Importantly, the Mayo Clinic team has leveraged substantial funding and collaborative infrastructure, receiving support from the National Institutes of Health and distinguished philanthropic foundations. This robust backing facilitated the extensive multi-institutional datasets essential for rigorous training and validation of AI algorithms, preserving generalizability and minimizing overfitting. Such thorough experimental design is paramount to ensuring that AI tools meet the stringent standards required for clinical acceptance and regulatory approval.
The implications of REDMOD extend beyond pancreatic cancer. This approach exemplifies the power of integrating advanced imaging informatics and machine learning to reveal “invisible” disease phenotypes at presymptomatic stages. Other malignancies and chronic diseases with heterogeneous tissue alterations may similarly benefit from AI analytics applied to routine imaging, heralding a new era where artificial intelligence fundamentally enhances radiologic interpretation.
In conclusion, Mayo Clinic’s REDMOD represents a landmark leap toward transforming pancreatic cancer from a disease diagnosed too late for cure into one detectable in its silent inception. By extracting rich radiomic data imperceptible to humans, this AI model offers a new window into early disease states. As clinical testing progresses, the hope is that this technology will integrate seamlessly with standard care, enabling timely intervention and ultimately improving survival outcomes in what has historically been one of the most challenging cancers to diagnose and treat.
Subject of Research: Early detection of pancreatic cancer using AI-enhanced radiomics on routine CT scans
Article Title: Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability
News Publication Date: April 22, 2026
Web References:
- Mayo Clinic: https://www.mayoclinic.org/about-mayo-clinic
- Pancreatic cancer information: https://www.mayoclinic.org/diseases-conditions/pancreatic-cancer/symptoms-causes/syc-20355421
- Published study: https://gut.bmj.com/content/early/2026/04/22/gutjnl-2025-337266
References: Refer to the Gut journal publication: “Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability”
Image Credits: Not provided
Keywords: Pancreatic cancer, early detection, artificial intelligence, radiomics, CT imaging, AI-PACED study, Mayo Clinic, radiology, oncology, machine learning, precision medicine, diagnostic innovation
