Early and Prediagnostic Detection of Pancreatic Cancer from Computed Tomography

arXiv ID: 2601.22134v1

Published: 2026-01-29

Authors: Wenxuan Li, Pedro R. A. S. Bassi, Lizhou Wu, Xinze Zhou, Yuxuan Zhao, Qi Chen, Szymon Plotka, Tianyu Lin, Zheren Zhu, Marisa Martin, Justin Caskey, Shanshan Jiang, Xiaoxi Chen, Jaroslaw B. Ćwikla, Artur Sankowski, Yaping Wu, Sergio Decherchi, Andrea Cavalli, Chandana Lall, Cristian Tomasetti, Yaxing Guo, Xuan Yu, Yuqing Cai, Hualin Qiao, Jie Bao, Chenhan Hu, Ximing Wang, Arkadiusz Sitek, Kai Ding, Heng Li, Meiyun Wang, Dexin Yu, Guang Zhang, Yang Yang, Kang Wang, Alan L. Yuille, Zongwei Zhou

Categories: cs.CV

Relevance Score: 1.00 / 1.00

View on arXiv Download PDF

Summary

This paper introduces ePAI, an AI system for early and prediagnostic detection of pancreatic ductal adenocarcinoma (PDAC) from computed tomography (CT) scans. ePAI demonstrated high performance in both internal (AUC 0.939-0.999, sensitivity 95.3%, specificity 98.7%) and external validation cohorts, accurately localizing PDACs as small as 2mm. Crucially, it detected previously overlooked PDACs 3-36 months before clinical diagnosis, significantly outperforming board-certified radiologists in sensitivity.

Medical Relevance

Pancreatic ductal adenocarcinoma is frequently diagnosed at an advanced, inoperable stage, leading to extremely poor patient outcomes. This AI-driven approach offers the potential for significantly earlier detection, enabling timely surgical intervention and improving the chances of curative treatment and prolonged survival.

AI Health Application

The AI system, ePAI, is designed to act as an assistive tool for radiologists to improve the early and prediagnostic detection of pancreatic ductal adenocarcinoma (PDAC) from CT scans. It aims to overcome the limitation of human oversight by identifying subtle lesions indicative of PDAC months to years before clinical diagnosis, thereby potentially enabling earlier treatment and improving patient survival rates.

Key Points

Developed ePAI, an AI-powered system designed for the early and prediagnostic detection of pancreatic ductal adenocarcinoma (PDAC) from routine CT scans.
Trained on data from 1,598 patients, ePAI achieved strong internal validation results (1,009 patients) with an AUC of 0.939-0.999, sensitivity of 95.3%, and specificity of 98.7% for detecting PDACs under 2 cm, precisely localizing lesions as small as 2mm.
Validated externally across 6 centers involving 7,158 patients, ePAI maintained robust performance with an AUC of 0.918-0.945, sensitivity of 91.5%, and specificity of 88.0%, capable of localizing PDACs as small as 5mm.
Demonstrated significant prediagnostic detection capability, identifying PDACs in 75 of 159 patients on CT scans taken 3 to 36 months before clinical diagnosis, with a median lead time of 347 days.
In a multi-reader study, ePAI significantly outperformed 30 board-certified radiologists by 50.3% (P < 0.05) in sensitivity for early and prediagnostic PDAC detection, while maintaining a comparable specificity of 95.4%.
The system's ability to precisely localize very small, often overlooked, PDAC lesions on historical scans addresses a major challenge in pancreatic cancer early detection.

Methodology

The study involved developing an automated AI system, ePAI, trained on CT data from 1,598 patients. Its performance was assessed through multi-phase validation: internal testing (1,009 patients), external testing across six different medical centers (7,158 patients), and a comparative multi-reader study where ePAI's performance was pitted against 30 board-certified radiologists. Evaluation metrics included Area Under the Receiver Operating Characteristic Curve (AUC), sensitivity, specificity, and the capability for precise lesion localization and prediagnostic detection lead time.

Key Findings

ePAI achieved an internal validation AUC of 0.939-0.999, 95.3% sensitivity, and 98.7% specificity, localizing PDACs as small as 2mm. In external validation, it showed an AUC of 0.918-0.945, 91.5% sensitivity, and 88.0% specificity, localizing PDACs as small as 5mm. It detected PDACs 3-36 months pre-diagnosis in 75 of 159 patients, with a median lead time of 347 days. Furthermore, ePAI outperformed 30 radiologists by 50.3% (P < 0.05) in sensitivity while maintaining comparable specificity (95.4%).

Clinical Impact

ePAI has the strong potential to serve as an invaluable assistive tool for radiologists, significantly enhancing the early and prediagnostic detection of pancreatic cancer. By identifying subtle, often overlooked lesions months to years before clinical diagnosis, this technology could dramatically increase the proportion of patients eligible for curative surgical resection, thereby improving treatment effectiveness and patient survival rates.

Limitations

The abstract does not explicitly state any specific limitations of the study or the ePAI system.

Future Directions

The abstract does not explicitly suggest future research directions, though prospective clinical trials and integration into clinical workflows would be implied next steps for an assistive diagnostic tool.

Medical Domains

Oncology Radiology Gastroenterology Surgical Oncology

Keywords

Pancreatic Cancer PDAC Early Detection Artificial Intelligence Computed Tomography CT Scan Prediagnostic Machine Learning

Abstract

Pancreatic ductal adenocarcinoma (PDAC), one of the deadliest solid malignancies, is often detected at a late and inoperable stage. Retrospective reviews of prediagnostic CT scans, when conducted by expert radiologists aware that the patient later developed PDAC, frequently reveal lesions that were previously overlooked. To help detecting these lesions earlier, we developed an automated system named ePAI (early Pancreatic cancer detection with Artificial Intelligence). It was trained on data from 1,598 patients from a single medical center. In the internal test involving 1,009 patients, ePAI achieved an area under the receiver operating characteristic curve (AUC) of 0.939-0.999, a sensitivity of 95.3%, and a specificity of 98.7% for detecting small PDAC less than 2 cm in diameter, precisely localizing PDAC as small as 2 mm. In an external test involving 7,158 patients across 6 centers, ePAI achieved an AUC of 0.918-0.945, a sensitivity of 91.5%, and a specificity of 88.0%, precisely localizing PDAC as small as 5 mm. Importantly, ePAI detected PDACs on prediagnostic CT scans obtained 3 to 36 months before clinical diagnosis that had originally been overlooked by radiologists. It successfully detected and localized PDACs in 75 of 159 patients, with a median lead time of 347 days before clinical diagnosis. Our multi-reader study showed that ePAI significantly outperformed 30 board-certified radiologists by 50.3% (P < 0.05) in sensitivity while maintaining a comparable specificity of 95.4% in detecting PDACs early and prediagnostic. These findings suggest its potential of ePAI as an assistive tool to improve early detection of pancreatic cancer.