Mayo Clinic REDMOD Detects Pancreatic Cancer Up to Three Years Before Clinical Diagnosis

--- headline: "Mayo Clinic REDMOD Detects Pancreatic Cancer Up to Three Years Before Clinical Diagnosis" slug: mayo-clinic-redmod-pancreatic-cancer-detection category: llms-genai story_number: "08" date: 2026-05-04 ---

An artificial intelligence model developed by Mayo Clinic can identify pancreatic cancer on routine CT scans up to three years before doctors would normally catch it -- a breakthrough that could fundamentally alter survival odds for one of the deadliest cancers in medicine.

The system, called REDMOD (Radiomics-based Early Detection Model), detected 73% of prediagnostic pancreatic cancers at a median of 475 days before clinical diagnosis, according to a landmark validation study published in the journal Gut in April 2026. By comparison, specialist radiologists reviewing the same scans caught just 39% of cases -- making the AI nearly twice as sensitive at identifying invisible early-stage malignant changes.

The gap widens further for the earliest detections. In scans taken more than two years before diagnosis, REDMOD identified 68% of cases compared to just 23% by radiologists -- nearly three times the detection rate for cancers that would otherwise go entirely unnoticed.

The Curability Gap

Pancreatic cancer remains one of oncology's cruelest diagnoses. More than 85% of patients receive their diagnosis after the disease has already spread, and five-year survival rates remain below 15%, according to the National Cancer Institute. Projections show it will become the second-leading cause of cancer-related death in the United States by 2030. The reason is brutally simple: the disease rarely produces detectable symptoms in its early stages, when surgery could still offer a cure.

"The greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable," says Dr. Ajit Goenka, the study's senior author and a Mayo Clinic radiologist and nuclear medicine specialist. "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."

That temporal window matters enormously. The researchers note that modeling studies indicate increasing the proportion of localized pancreatic cancers from 10% to 50% would more than double survival rates, underscoring that the timing of diagnosis is the single most critical determinant of outcomes.

How REDMOD Works

Unlike AI systems that look for visible tumors, REDMOD operates at a level beneath human perception. The model performs automated segmentation of the pancreas and extracts hundreds of quantitative radiomic features -- mathematical descriptors of tissue texture, density, and structure that capture faint biological changes as cancer begins to develop at what researchers call "stage 0."

The system is designed to analyze CT scans that patients have already received for other reasons, such as monitoring kidney stones or evaluating abdominal pain. It runs automatically without time-intensive manual preparation, making it practical for integration into existing clinical workflows. For high-risk patients -- particularly those with new-onset diabetes, a known early indicator of pancreatic cancer -- REDMOD could serve as an automated screening layer applied to imaging they are already receiving.

Validation Across Institutions

The study analyzed CT scans from 219 patients across multiple hospitals who showed no evidence of disease on initial radiologist review but were subsequently diagnosed with pancreatic cancer. Of these, 40% were diagnosed within 3 to 12 months of the scan, 35% between 12 and 24 months, and 25% more than 24 months later. Their scans were compared against 1,243 control patients matched by age, sex, and scan date who remained cancer-free for at least three years.

Critically, REDMOD demonstrated consistent performance across CT scans from multiple institutions, imaging systems, and protocols -- addressing a common weakness of AI medical models that perform well on training data but falter in real-world diversity. The model also showed temporal stability: in patients who underwent multiple scans, REDMOD produced consistent risk assessments 90-92% of the time when the same patient was scanned months apart.

The model correctly identified just over 81% of scans in an independent multi-hospital validation group of 539 patients as cancer-free, and 87.5% in the public NIH-PCT dataset of 80 patients.

From Laboratory to Clinic

Mayo Clinic researchers are now advancing this work through a prospective clinical study called AI-PACED (Artificial Intelligence for Pancreatic Cancer Early Detection), which will evaluate how clinicians can integrate AI-guided detection into care for patients at elevated risk. The study combines AI analysis of routine imaging with longitudinal follow-up to assess real-world performance, including early detection rates, false positive rates, and clinical outcomes.

"While prospective validation is paramount to confirm clinical utility, the REDMOD framework represents a significant advance towards shifting the paradigm for sporadic pancreatic ductal adenocarcinoma from a late-stage symptomatic diagnosis to proactive pre-clinical interception, offering tangible hope for improving outcomes in this challenging disease," the researchers concluded.

The Broader Picture

REDMOD arrives amid a surge of AI-driven diagnostic tools making their way from research papers to clinical practice. What distinguishes this work is its focus on detecting cancer before it becomes cancer in any clinically visible sense -- analyzing tissue that looks entirely normal to trained human eyes. The research is part of Mayo Clinic's Precure initiative, which aims to predict and prevent disease by identifying the earliest biological changes before symptoms begin.

The study was supported by the National Institutes of Health, the Hoveida Family Foundation, the Mayo Clinic Comprehensive Cancer Center, and the Champions for Hope Pancreas Cancer Research Program of the Funk-Zitiello Foundation. The study was conducted in collaboration with MD Anderson Cancer Center.

If the AI-PACED trial validates REDMOD's laboratory performance in a prospective clinical setting, the implications extend well beyond pancreatic cancer. The principle that AI can detect pre-clinical disease signatures invisible to human experts in routine imaging could be applied to other cancers where early detection remains the primary barrier to survival. For now, though, the 475-day head start REDMOD provides against one of medicine's most lethal cancers is significant enough on its own.