DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation

IntroductionHepatocellular carcinoma (HCC), a predominant subtype of liver cancer, remains Q7 a major contributor to global cancer mortality. Accurate delineation of liver tumors in CT and MRI scans is critical for treatment planning and clinical decision-making. However, manual segmentation is time...

Full description

Saved in:
Bibliographic Details
Main Authors: Siming Zheng, A. S. M. Sharifuzzaman Sagar, Yu Chen, Zehao Yu, Shi Ying, Yongyi Zeng
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-06-01
Series:Frontiers in Oncology
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fonc.2025.1569083/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849423177499803648
author Siming Zheng
Siming Zheng
A. S. M. Sharifuzzaman Sagar
Yu Chen
Zehao Yu
Shi Ying
Yongyi Zeng
Yongyi Zeng
author_facet Siming Zheng
Siming Zheng
A. S. M. Sharifuzzaman Sagar
Yu Chen
Zehao Yu
Shi Ying
Yongyi Zeng
Yongyi Zeng
author_sort Siming Zheng
collection DOAJ
description IntroductionHepatocellular carcinoma (HCC), a predominant subtype of liver cancer, remains Q7 a major contributor to global cancer mortality. Accurate delineation of liver tumors in CT and MRI scans is critical for treatment planning and clinical decision-making. However, manual segmentation is time-consuming, errorprone, and inconsistent, necessitating reliable automated approaches.MethodsThis study presents a novel U-shaped segmentation framework inspired by U-Net, designed to enhance accuracy and robustness. The encoder incorporates Dynamic Multi-Head Self-Attention (D-MSA) to capture both global and local spatial dependencies, while the decoder uses skip connections to preserve spatial detail. Additionally, a Feature Mix Module (FM-M) blends multiscale features, and a Residual Module (RM) refines feature representations and stabilizes training. The proposed framework addresses key challenges such as boundary precision, complex structural relationships, and dataset imbalance.ResultsExperimental results demonstrate superior segmentation performance, achieving a mean Dice score of 86.12 on the ATLAS dataset and 93.12 on the LiTS dataset.DiscussionThe proposed method offers a robust, efficient tool for liver tumor segmentation and holds strong potential to streamline diagnostic workflows and improve automated medical image analysis in clinical practice.
format Article
id doaj-art-7872f4f1aef14671a7a10f3a7d8ce4db
institution Kabale University
issn 2234-943X
language English
publishDate 2025-06-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Oncology
spelling doaj-art-7872f4f1aef14671a7a10f3a7d8ce4db2025-08-20T03:30:44ZengFrontiers Media S.A.Frontiers in Oncology2234-943X2025-06-011510.3389/fonc.2025.15690831569083DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentationSiming Zheng0Siming Zheng1A. S. M. Sharifuzzaman Sagar2Yu Chen3Zehao Yu4Shi Ying5Yongyi Zeng6Yongyi Zeng7The First Affiliated Hospital of Fujian Medical University, Fuzhou, ChinaDepartment of Hepatopancreatobiliary Surgery, First Hospital of Ningbo University, Ningbo, ChinaDepartment of Artificial and Robotics, Sejong University, Seoul, Republic of KoreaAI Lab, MetaSyntec Co., LTD, George Town, Cayman IslandsDepartment of Hepatopancreatobiliary Surgery, First Hospital of Ningbo University, Ningbo, ChinaDepartment of Engineering Technology, Ningbo Wedge Medical Technology Co., LTD, Ningbo, ChinaThe First Affiliated Hospital of Fujian Medical University, Fuzhou, ChinaDepartment of Hepatopancreatobiliary Surgery, Mengchao Hepatobiliary Hospital of Fujian Medical University, Fuzhou, ChinaIntroductionHepatocellular carcinoma (HCC), a predominant subtype of liver cancer, remains Q7 a major contributor to global cancer mortality. Accurate delineation of liver tumors in CT and MRI scans is critical for treatment planning and clinical decision-making. However, manual segmentation is time-consuming, errorprone, and inconsistent, necessitating reliable automated approaches.MethodsThis study presents a novel U-shaped segmentation framework inspired by U-Net, designed to enhance accuracy and robustness. The encoder incorporates Dynamic Multi-Head Self-Attention (D-MSA) to capture both global and local spatial dependencies, while the decoder uses skip connections to preserve spatial detail. Additionally, a Feature Mix Module (FM-M) blends multiscale features, and a Residual Module (RM) refines feature representations and stabilizes training. The proposed framework addresses key challenges such as boundary precision, complex structural relationships, and dataset imbalance.ResultsExperimental results demonstrate superior segmentation performance, achieving a mean Dice score of 86.12 on the ATLAS dataset and 93.12 on the LiTS dataset.DiscussionThe proposed method offers a robust, efficient tool for liver tumor segmentation and holds strong potential to streamline diagnostic workflows and improve automated medical image analysis in clinical practice.https://www.frontiersin.org/articles/10.3389/fonc.2025.1569083/fullliver cancersegmentationdeep learninghepatocellular carcinomaliver tumor
spellingShingle Siming Zheng
Siming Zheng
A. S. M. Sharifuzzaman Sagar
Yu Chen
Zehao Yu
Shi Ying
Yongyi Zeng
Yongyi Zeng
DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
Frontiers in Oncology
liver cancer
segmentation
deep learning
hepatocellular carcinoma
liver tumor
title DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
title_full DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
title_fullStr DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
title_full_unstemmed DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
title_short DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
title_sort dyntransnet dynamic transformer network with multi scale attention for liver cancer segmentation
topic liver cancer
segmentation
deep learning
hepatocellular carcinoma
liver tumor
url https://www.frontiersin.org/articles/10.3389/fonc.2025.1569083/full
work_keys_str_mv AT simingzheng dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT simingzheng dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT asmsharifuzzamansagar dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT yuchen dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT zehaoyu dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT shiying dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT yongyizeng dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation
AT yongyizeng dyntransnetdynamictransformernetworkwithmultiscaleattentionforlivercancersegmentation