EUVIMEDEUVIMED entdecken
Crossref

LW-CTrans: A lightweight hybrid network of CNN and Transformer for 3D medical image segmentation

Hulin Kuang, Yahui Wang, Xianzhen Tan, Jialin Yang, Jiarui Sun, Jin Liu, Wu Qiu, Jingyang Zhang, Jiulou Zhang, Chunfeng Yang, Jianxin Wang, Yang Chen

Medical Image Analysis · 2025-05 · 102 · 103545

DOI: 10.1016/j.media.2025.103545 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Medical Image Analysis, 2025-05. Verfasst von Hulin Kuang und 11 weiteren Autor:innen. Beteiligte Einrichtung: Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha 410000, China. Hauptthemen laut MEDLINE: Imaging, Three-Dimensional, Neural Networks, Computer. Die Arbeit wurde bisher 72-mal zitiert (Crossref) und nennt 76 Literaturangaben. Gefördert von: Science and Technology Program of Hunan Province, National Natural Science Foundation of China, Southeast University.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recent models based on convolutional neural network (CNN) and Transformer have achieved the promising performance for 3D medical image segmentation. However, these methods cannot segment small targets well even when equipping large parameters. Therefore, We design a novel lightweight hybrid network that combines the strengths of CNN and Transformers (LW-CTrans) and can boost the global and local representation capability at different stages. Specifically, we first design a dynamic stem that can accommodate images of various resolutions. In the first stage of the hybrid encoder, to capture local features with fewer parameters, we propose a multi-path convolution (MPConv) block. In the middle stages of the hybrid encoder, to learn global and local features meantime, we propose a multi-view pooling based Transformer (MVPFormer) which projects the 3D feature map onto three 2D subspaces to deal with small objects, and use the MPConv block for enhancing local representation learning. In the final stage, to mostly capture global features, only the proposed MVPFormer is used. Finally, to reduce the parameters of the decoder, we propose a multi-stage feature fusion module. Extensive experiments on 3 public datasets for three tasks: stroke lesion segmentation, pancreas cancer segmentation and brain tumor segmentation, show that the proposed LW-CTrans achieves Dices of 62.35±19.51%, 64.69±20.58% and 83.75±15.77% on the 3 datasets, respectively, outperforming 16 state-of-the-art methods, and the numbers of parameters (2.08M, 2.14M and 2.21M on 3 datasets, respectively) are smaller than the non-lightweight 3D methods and close to the lightweight methods. Besides, LW-CTrans also achieves the best performance for small lesion segmentation.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Imaging, Three-DimensionalNeural Networks, ComputerHumansAlgorithms

Stichwörter der Autor:innen

3D medical image segmentationConvolutional neural network and transformerLightweight networkSmall lesion segmentation

Autor:innen und Institutionen

  • Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha 410000, China.
  • School of Computer Science and Engineering, Southeast University, Nanjing 210096, China; Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University),…
  • School of Life Science and Technology, Huazhong University of Science and Technology, Wuhan 430000, China.
  • Department of Radiology, The First Affiliated Hospital of Nanjing Medical University, Nanjing, 210096, China; Lab for Artificial Intelligence in Medical Imaging (LAIMI), School of Medical Imaging, Nanjing Medical Univer…
  • Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha 410000, China; Xinjiang Engineering Research Center of Big Data and Intelligent Software, School…

Förderung

  • Science and Technology Program of Hunan Province
  • National Natural Science Foundation of China
  • Southeast University
  • Central South University

Zitieren

APA

Kuang, H., Wang, Y., Tan, X., Yang, J., Sun, J., Liu, J., Qiu, W., Zhang, J., Zhang, J., Yang, C., Wang, J., & Chen, Y. (2025). LW-CTrans: A lightweight hybrid network of CNN and Transformer for 3D medical image segmentation. Medical Image Analysis, 102, 103545. https://doi.org/10.1016/j.media.2025.103545

Vancouver

Kuang H, Wang Y, Tan X, Yang J, Sun J, Liu J, et al. LW-CTrans: A lightweight hybrid network of CNN and Transformer for 3D medical image segmentation. Medical Image Analysis. 2025;102:103545. doi:10.1016/j.media.2025.103545

BibTeX
@article{Kuang2025,
  author = {Kuang, Hulin and Wang, Yahui and Tan, Xianzhen and Yang, Jialin and Sun, Jiarui and Liu, Jin and Qiu, Wu and Zhang, Jingyang and Zhang, Jiulou and Yang, Chunfeng and Wang, Jianxin and Chen, Yang},
  title = {LW-CTrans: A lightweight hybrid network of CNN and Transformer for 3D medical image segmentation},
  journal = {Medical Image Analysis},
  year = {2025},
  volume = {102},
  pages = {103545},
  doi = {10.1016/j.media.2025.103545}
}
Kostenlos starten →