Skip to main content

ORIGINAL RESEARCH article

Front. Oncol.
Sec. Cancer Imaging and Image-directed Interventions
Volume 15 - 2025 | doi: 10.3389/fonc.2025.1450379
This article is part of the Research Topic Precision Medical Imaging for Cancer Diagnosis and Treatment - Vol. II View all 33 articles

SiameseNet based on multiple instance learning for accurate identification of the histological grade of ICC tumors

Provisionally accepted
Zhizhan Fu Zhizhan Fu 1Fazhi Feng Fazhi Feng 2Xingguang He Xingguang He 2*Tongtong Li Tongtong Li 3*Xiansong Li Xiansong Li 3*Ziluo Jituome Ziluo Jituome 3Zixing Huang Zixing Huang 3*Jinlin Ye Jinlin Ye 1*
  • 1 The Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou People's Hospital, Quzhou, China
  • 2 Yangtze Delta Region Institute (Quzhou), University of Electronic Science and Technology of China., Quzhou, China
  • 3 Department of Radiology, West China Hospital, Sichuan University, Chengdu, 610041, China., Chengdu, China

The final, formatted version of the article will be published soon.

    After hepatocellular carcinoma (HCC), Intrahepatic cholangiocarcinoma (ICC) is the most common primary liver cancer. Timely and accurate identification of the histological grade of ICC tumors is crucial for formulating appropriate diagnosis and treatment plans. To study the role of a deep-learning model based on multiple-instance learning and cross-attention mechanisms in predicting ICC histological grade, we included the data of 424 ICC patients, with 381 in the training cohort and 43 in the testing cohort. We introduced a dual-branch deep neural network called SiameseNet that mitigates performance degradation resulting from tumor heterogeneity by employing multiple-instance learning. The proposed approach integrates image information from two distinct modalities through a cross-attention mechanism, ultimately achieving a highperformance predictive network. The area under the curve and receiver operating characteristic curve were used to evaluate the model performance. The proposed network achieved an accuracy of 86.0%, area under the curve of 86.2%, sensitivity of 84.6%, and specificity of 86.7% in the testing cohort. This model can be used to assist doctors in the timely assessment of the histological grade of tumor in patients and to develop personalized diagnosis and treatment plans.

    Keywords: intrahepatic cholangiocarcinoma, Histological grade, Multiple Instance Learning, Cross-attention mechanism, CT-based diagnostics

    Received: 17 Jun 2024; Accepted: 15 Jan 2025.

    Copyright: © 2025 Fu, Feng, He, Li, Li, Jituome, Huang and Ye. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    * Correspondence:
    Xingguang He, Yangtze Delta Region Institute (Quzhou), University of Electronic Science and Technology of China., Quzhou, China
    Tongtong Li, Department of Radiology, West China Hospital, Sichuan University, Chengdu, 610041, China., Chengdu, China
    Xiansong Li, Department of Radiology, West China Hospital, Sichuan University, Chengdu, 610041, China., Chengdu, China
    Zixing Huang, Department of Radiology, West China Hospital, Sichuan University, Chengdu, 610041, China., Chengdu, China
    Jinlin Ye, The Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou People's Hospital, Quzhou, China

    Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.