Early Pancreatic Cancer Detection: MRI in New Diabetes and AI Analysis of Earlier CT Scans

New Hope Against a Silent Threat

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most silent and lethal cancers. At diagnosis, 75% are inoperable, and 5-year survival is below 12%. Lack of routine screening contributes to late detection. Two studies published in July 2025 offer potential change: AI-assisted CT analysis and metabolic-risk-based MRI screening.

1. PANCANAI: Can AI Read Tumor Shadows on CT?

Publication: Investigative Radiology, 24 June 2025 | Lead: Laura Degand, University of Copenhagen

This retrospective study examined 1220 CT scans from 1083 biopsy-confirmed pancreatic cancer patients in 2006-2016. A pretrained AI model, PANCANAI, analyzed tumor lesions and main pancreatic duct (MPD) width.

Key Findings:

• CT Sensitivity at Diagnosis: 91.8%

• CT Taken ≥1 Year Before Diagnosis: 53.9%

• Stage I Sensitivity: 82.9%

• Portal-Phase Sensitivity: 92.1%; arterial phase: 90.9%; delayed phase: 83.5%

Interpretation:

AI can identify cancer traces in approximately half of CT scans taken a year before diagnosis, detecting a radiomic fingerprint before symptoms emerge.

2. PANDOME: Can MRI Screen Higher-Risk Diabetes Patients?

Publication: The Journal of Clinical Endocrinology & Metabolism, July 2025 | Lead: Dr. Richard Frank, US

This prospective study used MRI screening in 109 people over 50 with new-onset diabetes (NOD) or deteriorating diabetes (DD).

Definitions:

• New-Onset Diabetes (NOD): HbA1c elevation within 12 months

• Deteriorating Diabetes (DD): Diabetes for at least 2 years, with >2% HbA1c increase over 6 months not explained by weight gain or nonadherence

Key Findings:

• Cystic lesions averaging 6 mm were found in 50% of participants

• 7 underwent EUS; 4 required biopsy

• One biopsy in the DD group identified Stage IB PDAC .

• The DD Group Had More:

  • Elevated HbA1c (P = .02)

  • Weight loss (P = .0038)

  • Insulin requirement (P < .0001)

• ENDPAC Risk Score:

  • 75% of DD patients were high-risk (NOD: 35.6%)

Interpretation: Worsening diabetes may signal pancreatic cancer. This prospective study demonstrated for the first time that MRI may enable early diagnosis in higher-risk people.

Meaningful Progress in Early Pancreatic Cancer Detection

Silent progression and ineffective screening have long made pancreatic cancer difficult to diagnose early. PANCANAI and PANDOME suggest a paradigm shift. Retrospective AI analysis of archived images and prospective screening of high-risk individuals identified through subtle metabolic changes both advance detection before clinical presentation. They validate tools and reshape who and how to screen. Together, they mark an important step toward earlier diagnosis and potential survival gains in one of medicine's deadliest diseases.