Early detection of pancreatic cancer with an AI tool may save lives

An artificial intelligence (AI)-powered imaging analysis system performed as well or better than clinical readers in detecting (early) pancreatic cancer. Early identification of pre-malignant and malignant lesions could significantly improve survival outcomes for these patients.

“There is no doubt that early detection is essential to improve the prognosis of pancreatic cancer,” stated Prof. Atsuhiro Masuda (Kobe University, Japan) [1]. The authors developed an AI-based imaging analysis tool to detect early pancreatic cancer on both contrast- and non-contrast CT scans. They analysed 3,916 CT images from 2,061 patients. “The system evaluated features like pancreatic masses, atrophy, main pancreatic duct dilatation, and main pancreatic duct stenosis,” explained Prof. Masuda. The performance of the AI tool was compared with that of clinical readers.

The AI-based tool demonstrated comparable diagnostic performance to human readers across all diagnostic features, achieving accuracies of ≥0.90 for both contrast CT and non-contrast CT images. “In contrast CT scans, the sensitivity for detecting pancreatic cancer was higher for AI (98.9% vs 88.9%), while the specificity was slightly higher for the human readers (99.3% vs 93.0%),” mentioned Prof. Masuda. In non-contrast CT, the sensitivity for diagnosing pancreatic cancer was substantially higher in the AI group (83.9% vs 53.4%), whereas the specificity remained slightly higher among the human readers (99.5% vs 92.0%). “Notably, for tumours smaller than 20 mm, the AI-based analysis system outperformed the human readers,” added Prof. Masuda.

“Our AI model for detecting pancreatic cancer performed equally well or better than human readers,” concluded Prof. Masuda. “This AI application could support earlier detection of pancreatic cancer and ultimately improve the prognoses of patients.”

1. Masuda A, et al. Development and future perspectives of an AI-based imaging support system for early detection of pancreatic cancer. OP121, Prediction and prognosis of pre-malignant and malignant pancreatic lesions, UEG Week, 4–7 October 2025, Berlin, Germany.

Copyright ©2025 Medicom Medical Publishers



Posted on 1 December 20251 December 2025