The Pathologist
  • Explore Pathology

    Explore

    • Latest
    • Insights
    • Case Studies
    • Opinion & Personal Narratives
    • Research & Innovations
    • Product Profiles

    Featured Topics

    • Molecular Pathology
    • Infectious Disease
    • Digital Pathology

    Issues

    • Latest Issue
    • Archive
  • Subspecialties
    • Oncology
    • Histology
    • Cytology
    • Hematology
    • Endocrinology
    • Neurology
    • Microbiology & Immunology
    • Forensics
    • Pathologists' Assistants
  • Training & Education

    Career Development

    • Professional Development
    • Career Pathways
    • Workforce Trends

    Educational Resources

    • Guidelines & Recommendations
    • App Notes
    • eBooks

    Events

    • Webinars
    • Live Events
  • Events
    • Live Events
    • Webinars
  • Profiles & Community

    People & Profiles

    • Power List
    • Voices in the Community
    • Authors & Contributors
  • Multimedia
    • Video
    • Pathology Captures
Subscribe
Subscribe

False

The Pathologist / Issues / 2026 / November / AI Detects Hidden Pancreatic Cancer Signs
Endocrinology Biochemistry and molecular biology Clinical care Molecular Pathology

AI Detects Hidden Pancreatic Cancer Signs

An automated CT analysis tool identifies pancreatic cancer more than a year before clinical diagnosis, potentially opening new opportunities for early detection

11/18/2026 News 2 min read
  • Full Article
  • Summary
  • Takeaways
  • Listen
  • Scorecard
  • Quiz
  • Poll

Share

An artificial intelligence (AI) system detected signs of pancreatic cancer in CT scans taken more than a year before clinical diagnosis, identifying abnormalities that were invisible to experienced radiologists, according to research from the Mayo Clinic.

In a study published in Gut, researchers developed the Radiomics-based Early Detection Model (REDMOD), which analyzes subtle changes in pancreatic tissue that cannot be distinguished through conventional visual assessment.

Early detection of pancreatic ductal adenocarcinoma remains challenging because tumors often become visible only after the disease has progressed beyond the stage at which surgery is possible.

The researchers evaluated 1,462 CT scans, including 219 from patients who subsequently developed pancreatic cancer and 1,243 from individuals without the disease. Scans were collected across multiple institutions and divided into training and independent testing datasets.

REDMOD automatically identifies the pancreas and examines 40 imaging features, including subtle differences in tissue texture and structure. These measurements are combined using machine learning to distinguish potentially cancerous changes from normal tissue.

In the independent test dataset, REDMOD identified 73 percent of patients who subsequently developed pancreatic cancer, compared with 39 percent detected by radiologists. The median interval between imaging and diagnosis was 475 days.

Importantly, the system maintained 68 percent sensitivity in scans obtained more than two years before diagnosis, suggesting that cancer-associated changes may be detectable considerably earlier than previously possible.

The model also demonstrated consistent results across repeated scans and different imaging institutions. External validation showed specificity ranging from 81 to 88 percent.

The findings suggest that automated analysis of existing CT scans could help identify patients requiring closer investigation, particularly those with risk factors such as newly diagnosed diabetes and unexplained weight loss.

However, the model incorrectly flagged approximately one in five cancer-free individuals, highlighting the need for confirmatory testing and careful clinical interpretation.

The retrospective study also did not establish whether earlier AI detection improves patient survival. Prospective validation in high-risk populations will be essential before the technology can be incorporated into routine diagnostic pathways.

Newsletters

Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.

Newsletter Signup Image

Explore More in Pathology

Dive deeper into the world of pathology. Explore the latest articles, case studies, expert insights, and groundbreaking research.

False

Advertisement

Recommended

False

Affiliations:

Specialties:

Areas of Expertise:

Contributions:

False

The Pathologist
Subscribe

About

  • About Us
  • Work at Conexiant Europe
  • Terms and Conditions
  • Privacy Policy
  • Your Privacy Choices
  • Advertise With Us
  • Contact Us

Copyright © 2026 Texere Publishing Limited (trading as Conexiant), with registered number 08113419 whose registered office is at Booths No. 1, Booths Park, Chelford Road, Knutsford, England, WA16 8GS.