AI Model Detects Pancreatic Cancer Years Before Symptoms Appear, Study Shows

An artificial intelligence system developed by Mayo Clinic researchers, known as REDMOD (Radiomics-based Early Detection Model), has demonstrated the ability to detect pancreatic cancer on routine abdominal CT scans up to approximately three years before clinical diagnosis, according to a study published in the journal Gut on April 28, 2026 ^[1][2].

The AI model identified subtle tissue changes in CT scans that were initially interpreted as normal, detecting cancer on average about 475 days, or approximately 16 months, before diagnosis ^[2]. In head-to-head comparisons, REDMOD detected 73% of prediagnostic pancreatic cancer cases, compared to about 39% detected by radiologists reviewing the same scans ^[2]. For scans taken more than two years before diagnosis, REDMOD detected 68% of cases versus 23% by radiologists ^[2].

REDMOD also demonstrated high accuracy in classifying scans from individuals who did not develop pancreatic cancer, correctly identifying over 80% of such cases. This included 81% accuracy in an independent cohort and 87.5% in the NIH-PCT dataset ^[3]. The model’s predictions remained consistent over time, producing the same result in 90–92% of cases when patients had multiple scans months apart ^[3].

Pancreatic cancer is often diagnosed at an advanced stage, with more than 85% of cases detected too late for curative treatment. The five-year survival rates remain low, around 10–15% ^[2]. Modeling studies suggest that increasing the proportion of localized pancreatic ductal carcinomas from 10% to 50% could more than double survival rates ^[2].

What Is Known

REDMOD has shown significant promise in early detection of pancreatic cancer, outperforming traditional radiological methods in identifying cases years before clinical symptoms appear ^[1][2][3]. The study highlights the potential of AI in improving early cancer detection, which could lead to earlier interventions and improved survival rates ^[2].

What Remains Unclear

While the AI model’s ability to detect cancer earlier than current methods is clear, the claim that REDMOD could shift diagnosis to a ‘stage 0’ or pre-clinical stage is not confirmed by multiple independent sources ^[3]. Additionally, the assertion that REDMOD was tested on nearly 2,000 CT scans lacks corroboration across multiple outlets ^[1].