Multicenter deep learning model for pancreatic cancer detection using endoscopic ultrasound

doi:10.4251/wjgo.119889

Advanced Search

BPG is committed to discovery and dissemination of knowledge

Home / Archive / Volume 18, Issue 9

This Article

(0)

Peer-Review Report of This Article

CrossCheck and Google Search of This Article

Academic Rules and Norms of This Article

Citation of this article

Corresponding Author of This Article

Research Domain of This Article

Article-Type of This Article

Open-Access Policy of This Article

Times Cited Counts in Google of This Article

Journal Information of This Article

Publication Name

World Journal of Gastrointestinal Oncology

ISSN

1948-5204

Publisher of This Article

Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA

- Full Article (XML)

Retrospective Study

Copyright: ©Author(s) 2026. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial (CC BY-NC 4.0) license. No commercial re-use. See permissions. Published by Baishideng Publishing Group Inc.

World J Gastrointest Oncol. Sep 15, 2026; 18(9): 119889
Published online Sep 15, 2026. doi: 10.4251/wjgo.119889

Multicenter deep learning model for pancreatic cancer detection using endoscopic ultrasound

Xin-Ying Yu, Jun-Qiang Ye, Zhen He, Qiang He

Xin-Ying Yu, Qiang He, Department of Gastroenterology, Beijing Tiantan Hospital, Capital Medical University, Beijing 100071, China

Jun-Qiang Ye, Electronic Information and Communication, Huazhong University of Science and Technology, Wuhan 430074, Hubei Province, China

Zhen He, Department of Gastroenterology, Beijing Friendship Hospital, Capital Medical University, Beijing 100050, China

Co-corresponding authors: Zhen He and Qiang He.

Author contributions: Yu XY and He Z designed the research study and revised the manuscript; Ye JQ developed the machine learning algorithms and wrote the initial manuscript draft; He Q collected the data, performed the research, and contributed to the drafting and revision of the manuscript; all authors have read and approve the final manuscript. Our study was a multicenter collaborative effort involving two major centers. To appropriately reflect the contributions and responsibilities of each participating site, we designated one corresponding author per center. Specifically, He Q from the Department of Gastroenterology, Beijing Tiantan Hospital, Capital Medical University, serves as the corresponding author for the first center. The other co-corresponding author (He Z) is from the Department of Gastroenterology, Beijing Friendship Hospital, Capital Medical University, representing the second center. This arrangement ensures that each center has a dedicated point of contact for scientific inquiries, data verification, and administrative matters. It also acknowledges the equal intellectual and logistical input from both sites, which is a common and transparent practice in multicenter studies. We believe that co-corresponding authorship accurately represents the collaborative nature of our work and facilitates efficient communication with the research community.

AI contribution statement: The initial draft of the manuscript (including all scientific content such as abstract, introduction, materials and methods, results, discussion, and conclusion) was entirely written by the author without the use of any artificial intelligence text generation tools (such as ChatGPT). The core scientific content, research design, data interpretation, or conclusions are not generated by artificial intelligence. During the revision and submission preparation process, we only use AI assisted language polishing tools to improve grammar, spelling, and readability, similar to using professional editing services. This is a common practice to ensure language clarity. The manuscript did not use artificial intelligence tools to generate any numbers, images, or other visual elements. There are no artificial intelligence tools involved in research design, data analysis, or result interpretation.

Institutional review board statement: The study was reviewed and approved by the IRB of Beijing Tiantan Hospital, Capital Medical University (Approval No. KY 2020-089-02).

Informed consent statement: All study participants provided informed written consent prior to study enrollment.

Conflict-of-interest statement: The authors declare that they have no conflict of interest.

Data sharing statement: No additional data are available.

Corresponding author: Qiang He, Department of Gastroenterology, Beijing Tiantan Hospital, Capital Medical University, No. 119 South Fourth Ring Road West, Fengtai District, Beijing 100071, China. 229476289@qq.com

Received: February 10, 2026
Revised: March 29, 2026
Accepted: May 12, 2026
Published online: September 15, 2026
Processing time: 197 Days and 7.5 Hours

Abstract

BACKGROUND

The early diagnosis of pancreatic cancer via endoscopic ultrasound (EUS) is challenging. Existing deep learning models mostly rely on single-center data and often experience significant generalization decline across different centers due to variations in imaging equipment, settings, and populations.

AIM

To improve pancreatic cancer detection by developing a robust deep learning model capable of handling data heterogeneity across multiple EUS centers.

METHODS

We constructed a multicenter dataset (383 patients, 2362 images) from two hospitals. After performing frequency-domain analysis of cross-center heterogeneity to quantify imaging variations and distribution shifts between the centers, we developed a multicenter data-oriented deep learning classification network named MCEUS-C2Net, integrating local and global features with channel attention. The model was trained and validated on one center's data (182 training, 60 validation, 60 test cases) and independently evaluated on the second (81 cases). Patient-level partitioning ensured independent cross-center evaluation.

RESULTS

MCEUS-C2Net achieved accuracy, sensitivity, and F1 score of 94.51%, 97.09%, and 95.24% in the internal test, outperforming ResNet-50, Swin transformer, and MedViTV2. In external validation, it maintained robust performance with 90.59% accuracy, 90.50% sensitivity, and an area under the curve of 0.9704, showing the smallest performance decline among all models. Confusion matrix analysis confirmed fewer false positives and negatives. Frequency-domain analysis identified significant cross-center distribution differences. Attention heatmaps demonstrated that MCEUS-C2Net stably and autonomously focused on pancreatic lesion sites with high clinical consistency, whereas comparison models exhibited dispersed attention. These results confirm the model’s superior cross-center robustness and interpretability for pancreatic cancer identification.

CONCLUSION

Our approach achieves high-performance pancreatic cancer classification across multicenter data. It effectively addresses imaging heterogeneity, providing a robust, clinically viable solution for standardized EUS diagnosis in real-world settings.

Keywords: Pancreatic cancer lesions; Endoscopic ultrasound; Deep learning; Multicenter study; Computer-aided diagnosis; Attention mechanism

Core Tip: This study introduces the first multicenter endoscopic ultrasound dataset for pancreatic lesions. We also propose a novel deep learning model, MCEUS-C2Net. It integrates local and global features with channel attention mechanisms. This design effectively overcomes cross-center imaging heterogeneity. The model accurately differentiates cancerous from noncancerous pancreatic lesions. During external validation, it demonstrated exceptional generalization. Ultimately, this research provides a valuable benchmark dataset and a robust deep learning network. This network holds the potential to enhance the accuracy of clinical decision-making in real-world settings.