EUVIMEDEUVIMED entdecken
Crossref

MLF ‐ MUnet : A Novel Multi‐Level Fusion Unet Framework With Multi‐Attention for Medical Image Segmentation

Juan Qin, Xulong Fu, Yongqi Sun, Haiping Ren, Jinhua Zhu, Kun Li

International Journal of Imaging Systems and Technology · 2026-11 · 36 (6)

DOI: 10.1002/ima.70485 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Imaging Systems and Technology, online erschienen am 09.11.2026. Verfasst von Juan Qin und 5 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In medical image analysis, deep learning techniques, especially the algorithmic models based on convolutional neural networks (CNNs), have received extensive attention and in‐depth research in academia and clinical medicine in recent years. Among various CNNs architectures, with its unique encoder‐decoder structure, Unet can effectively fuse local details with global semantic features. It has become a mainstream method for medical image segmentation. However, traditional Unet is limited by its restricted receptive field and inability to model long‐range dependencies, making it difficult to accurately capture fine edges and contextual information in complex scenes. In response to these limitations, this paper proposes a novel multilevel feature fusion Unet framework (MLF‐MUnet) incorporating multi‐attention mechanisms. The model replaces the standard convolutional layer of Unet with the squeeze‐and‐excitation residual convolution module (SE‐ResCov) and introduces the convolutional block attention module (CBAM). This design can effectively suppress background noise, enhance the identification of critical anatomical boundaries, and synergistically optimize gradient propagation. To enhance long‐range dependency modeling, MLF‐MUnet incorporates a multi‐level fusion (MLF) module. This enables it to efficiently integrate global context information in the decoding phase, further enhancing the feature extraction capability of the network. Experimental results on three public medical imaging datasets ( SpineSagT2W, ISIC, and FML ) show that MLF‐MUnet performs well on multiple metrics. Its robust segmentation capability and excellent generalizability make it suitable for diverse medical image analysis tasks.

Zitieren

APA

Qin, J., Fu, X., Sun, Y., Ren, H., Zhu, J., & Li, K. (2026). MLF ‐ MUnet : A Novel Multi‐Level Fusion Unet Framework With Multi‐Attention for Medical Image Segmentation. International Journal of Imaging Systems and Technology, 36(6). https://doi.org/10.1002/ima.70485

Vancouver

Qin J, Fu X, Sun Y, Ren H, Zhu J, Li K. MLF ‐ MUnet : A Novel Multi‐Level Fusion Unet Framework With Multi‐Attention for Medical Image Segmentation. International Journal of Imaging Systems and Technology. 2026;36(6). doi:10.1002/ima.70485

BibTeX
@article{Qin2026,
  author = {Qin, Juan and Fu, Xulong and Sun, Yongqi and Ren, Haiping and Zhu, Jinhua and Li, Kun},
  title = {MLF ‐ MUnet : A Novel Multi‐Level Fusion Unet Framework With Multi‐Attention for Medical Image Segmentation},
  journal = {International Journal of Imaging Systems and Technology},
  year = {2026},
  volume = {36},
  number = {6},
  doi = {10.1002/ima.70485}
}
Kostenlos starten →