Deep learning CT model for stratified diagnosis of pancreatic cystic neoplasms: multicenter development, validation, and real-world clinical impact

1

The study developed an AI-powered CT model, PCN-AI, to enhance the diagnosis of pancreatic cystic neoplasms (PCN) using data from 1835 patients.
2

PCN-AI significantly improved diagnostic accuracy for radiologists, increasing AUC from 0.786 to 0.845 and reducing interpretation time by 23.7%.
3

In a real-world cohort, PCN-AI identified missed malignant PCN cases in 45.45% of patients, enabling timely interventions and reducing clinical workload.
4

The model achieved robust performance across classification tasks, with AUCs ranging from 0.845 to 0.988, demonstrating its potential in clinical practice.
5

PCN-AI aligns with WHO diagnostic criteria, offering hierarchical classification of PCN subtypes and addressing gaps in current diagnostic methods.

Npj Digital Medicine

by Xiaohan Yuan, Chengwei Chen, Zhang Shi, Wenbin Liu, Xinyue Zhang, Ming Yang, Mengmeng Zhu, Jieyu Yu, Fang Liu, Jing Li, Yunshuo Zhang, Hui Jiang, Bozhu Chen, Jianping Lu, Chengwei Shao, Yun Bian
October 13, 2025

1