Mayo Clinic AI detects pancreatic cancer years before diagnosis

Overview

Updated May 31

Pancreatic cancer kills nearly nine in ten patients within five years, mostly because doctors catch it too late to operate. On April 28, 2026, Mayo Clinic researchers published a validation study showing their artificial intelligence model can flag the disease on routine CT scans up to three years before a clinical diagnosis.

The model, called REDMOD (Radiomics-based Early Detection Model), found 73% of prediagnostic cancers in scans previously read as normal. Radiologists working without the AI caught 39%. The system now moves into a clinical trial that will test whether earlier flags translate into more curable surgeries.

Why it matters

Caught early, pancreatic cancer is survivable for most patients. Caught late, it kills nine in ten. This tool shifts that timing window.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Play on this story Voices Debate Predict Connections

Key Indicators

3 years

How early REDMOD spots cancer

The AI flagged signs of disease on CT scans taken up to three years before patients were diagnosed.

73%

AI sensitivity

Share of prediagnostic cancers REDMOD identified in scans previously interpreted as normal.

39%

Radiologist sensitivity

Share found by specialists reviewing the same scans without AI assistance.

88%

AI specificity

Share of healthy patients the model correctly cleared, limiting false positives.

13%

Current five-year survival

The U.S. pancreatic cancer survival rate has stalled at 13% for three straight years.

80%+

Stage IA five-year survival

When pancreatic cancer is found at the earliest stage, more than 80% of patients survive five years.

1,755

CT scans analyzed

Total scans across the development and test sets in the validation study.

67,530

U.S. cases projected in 2026

American Cancer Society estimate of new pancreatic cancer diagnoses this year.

Voices

Curated perspectives — historical figures and your fellow readers.

Ever wondered what historical figures would say about today's headlines?

Play

Exploring all sides of a story is often best achieved with Play.

Most of these play right now — no account needed. Sign up to save scores, keep a streak, and unlock Debate and Predict. Log in Sign Up

Predict 4 ways this could play out. Back the one you believe — contrarian picks score more when a scenario has a resolution date. Log in to play

People Involved

Ajit Harishkumar Goenka, MD

Senior author of the REDMOD validation study

Leading the AI-PACED clinical trial at Mayo Clinic

Organizations Involved

Mayo Clinic

Academic medical center

Developer of REDMOD; running the AI-PACED clinical trial

A nonprofit academic medical system based in Rochester, Minnesota.

Pancreatic Cancer Action Network

Patient advocacy nonprofit

Funds early-detection research; tracks survival statistics

A U.S. nonprofit pushing for earlier detection and better treatment of pancreatic cancer.

Timeline

January 2020 May 2026

5 events Latest: May 8th, 2026 · 4 months ago

Tap a bar to jump to that date

May 2026
AI-PACED clinical trial enrollment underway
Latest Clinical trial

Mayo advances REDMOD into a prospective study testing AI-guided screening in high-risk patients, measuring outcomes and false positives.
May 8th, 2026
April 2026
Bloomberg and major outlets cover the breakthrough
Coverage

Mainstream press picks up the study, framing it as the most credible early-detection advance for pancreatic cancer in a decade.
April 29th, 2026
Mayo Clinic publishes REDMOD validation study
Research

Researchers report in Gut that the AI flagged 73% of prediagnostic cancers across 1,755 scans, beating radiologists nearly two to one.
April 28th, 2026
January 2024
Pancreatic cancer survival stalls at 13%
Context

PanCAN reports the third straight year with no movement in five-year survival, even as other cancers improve.
January 1st, 2024
January 2020
DeepMind AI matches radiologists on breast cancer screening
Milestone

Nature publishes results showing a Google AI reduces false negatives in mammography. The first credible proof that AI imaging can rival specialists.
January 1st, 2020

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

1950–1970

Pap smear adoption (1950s–1960s)

George Papanicolaou's cervical cell test, developed in the 1940s, became the first widespread cancer screening tool after the American Cancer Society endorsed it in 1957. By the 1960s, U.S. women were getting routine Pap smears as part of gynecologic care.

Then

Cervical cancer detection shifted from late-stage diagnosis to precancerous lesions that could be removed in clinic.

Now

Cervical cancer deaths in the U.S. fell roughly 70% between 1955 and 1992. The Pap smear became the model for what successful cancer screening looks like.

Why this matters now

Pancreatic cancer today sits where cervical cancer sat before the Pap smear: no general screening test, late diagnoses, low survival. REDMOD will not become a Pap smear overnight, but it is the first credible candidate for changing the same dynamic.

August 2011

National Lung Screening Trial (2011)

The National Cancer Institute reported that low-dose CT screening cut lung cancer deaths by 20% among heavy smokers compared to chest X-rays. The trial enrolled 53,454 high-risk patients across 33 U.S. centers.

Then

The U.S. Preventive Services Task Force recommended annual low-dose CT screening for high-risk smokers in 2013. Medicare began covering it in 2015.

Now

Lung cancer mortality has fallen steadily since, with screening cited as a meaningful contributor. Uptake remains uneven; only about 16% of eligible patients get screened.

Why this matters now

The lung trial is the closest precedent for what AI-PACED is trying to prove: that earlier imaging-based detection of a deadly cancer translates into lives saved. It also shows the gap between proving a screening tool works and getting eligible patients to actually use it.

January 2020

DeepMind breast cancer screening AI (2020)

Google's DeepMind unit published a Nature study showing its AI reduced false negatives by 9.4% and false positives by 5.7% in U.S. mammograms. In a head-to-head test, the model outperformed all six human radiologists, even with less patient history to work from.

Then

The result triggered the first wave of FDA-cleared AI tools for mammography reading. Hospitals began piloting AI-assisted double reads.

Now

By 2025, AI-assisted mammography became a standard offering at large U.S. radiology groups, though insurance coverage and workflow integration lagged the technology.

Why this matters now

REDMOD follows the same playbook: train on a hard imaging problem, beat specialists in a controlled study, then face a long road to real-world adoption. The 2020 milestone also showed that the gap between published study and routine clinical use can stretch five years or more.