Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Optics and Precision Engineering, Volume. 33, Issue 9, 1456(2025)

Hierarchical feature refinement with frequency-enhanced learning method for nuclei segmentation in spatial omics

Xiuqi LI1,2, Jinze LI1,2, Qi YANG2, Yingxue LI2, Cairong ZHAO3, Lianqun ZHOU1,2、*, and Jia YAO1,2、*
Author Affiliations
  • 1Division of Life Sciences and Medicine， School of Biomedical Engineering （Suzhou）， University of Science and Technology of China， Hefei230026， China
  • 2Key Laboratory of Biomedical Detection Technology， Suzhou Institute of Biomedical Engineering and Technology， Chinese Academy of Sciences， Suzhou15163， China
  • 3College of Electronics and Information Engineering， Tongji University， Shanghai201804， China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (11)
    Equations (18)
    References (39)
    Tools
    Paper Information
    Topics
    Figures & Tables(11)
    Network architecture of FFVM-UKAN

    Fig. 1. Network architecture of FFVM-UKAN

    Download full size
    View in Article
    Network architecture and operational details of shallow feature extraction module

    Fig. 2. Network architecture and operational details of shallow feature extraction module

    Download full size
    View in Article
    Network architecture and operational details of deep feature refinement module

    Fig. 3. Network architecture and operational details of deep feature refinement module

    Download full size
    View in Article
    Network architecture of PFDLM

    Fig. 4. Network architecture of PFDLM

    Download full size
    View in Article
    Comparison of different segmentation results after ablation of each module

    Fig. 5. Comparison of different segmentation results after ablation of each module

    Download full size
    View in Article
    Class activation heatmaps with different methods in ablation experiments

    Fig. 6. Class activation heatmaps with different methods in ablation experiments

    Download full size
    View in Article
    Comparison of different methods on in-house and public MoNuSeg datasets

    Fig. 7. Comparison of different methods on in-house and public MoNuSeg datasets

    Download full size
    View in Article
    Comparison of different methods on human liver sample from 10X Genomics

    Fig. 8. Comparison of different methods on human liver sample from 10X Genomics

    Download full size
    View in Article

    • Table 1. Ablation experiments on in-house and public MoNuSeg datasets

      View table
      View in Article

      Table 1. Ablation experiments on in-house and public MoNuSeg datasets

      ModelPFDLMTok-KANIn-house datasetMoNuSeg dataset
      mIoUDiceAccSpeSenmIoUDiceAccSpeSen
      VM-UNet0.845 40.916 20.961 10.973 00.921 30.665 40.799 10.916 80.936 70.836 1
      FFVM-UNet√0.855 30.922 00.964 10.977 50.919 20.685 20.813 20.922 80.941 20.848 4
      VM-UKAN√0.850 60.919 20.963 40.981 60.902 60.678 70.808 60.920 80.939 50.845 1
      FFVM-UKAN√√0.859 50.924 50.965 60.981 90.911 30.690 90.817 20.925 00.944 20.847 3

    • Table 2. Comparison of different methods on in-house and public MoNuSeg datasets

      View table
      View in Article

      Table 2. Comparison of different methods on in-house and public MoNuSeg datasets

      ModelIn-house datasetMoNuSeg dataset
      mIoUDiceAccSpeSenmIoUDiceAccSpeSen
      UNet［18］0.818 40.900 20.953 70.968 60.903 80.604 20.746 50.887 00.898 70.850 8
      UNet++［34］0.823 00.902 90.954 80.968 10.910 20.600 90.745 30.892 20.912 40.818 9
      Swin-UNet［22］0.811 30.895 80.950 60.991 10.832 20.627 20.770 80.914 60.961 30.725 5
      VM-UNet［25］0.845 40.916 20.961 10.973 00.921 30.665 40.799 10.916 80.936 70.836 1
      FFVM-UKAN0.859 50.924 50.965 60.981 90.911 30.690 90.817 20.925 00.944 20.847 3

    • Table 3. Comparison experiments on human liver sample from 10X Genomics with nuclei segmentation and gene mapping

      View table
      View in Article

      Table 3. Comparison experiments on human liver sample from 10X Genomics with nuclei segmentation and gene mapping

      ModelmIoUDiceAccSpeSenAverage time/smIoUgeneDicegeneAccgeneSpegeneSengene
      UNet0.687 80.815 00.931 70.947 90.855 90.048 70.690 70.817 00.870 40.844 00.928 8
      UNet++0.677 30.807 60.931 40.955 60.819 00.076 70.696 80.821 30.879 00.872 90.892 6
      Swin-UNet0.320 90.469 60.877 40.998 00.325 10.020 30.331 10.497 50.789 70.995 90.334 1
      VM-UNet0.681 50.809 10.939 60.986 20.733 70.018 90.699 10.822 90.900 60.972 70.741 2
      FFVM-UKAN0.700 50.822 40.942 90.980 00.779 90.048 10.723 10.839 30.906 30.961 10.785 3
    Tools

    Get Citation

    Copy Citation Text

    Xiuqi LI, Jinze LI, Qi YANG, Yingxue LI, Cairong ZHAO, Lianqun ZHOU, Jia YAO. Hierarchical feature refinement with frequency-enhanced learning method for nuclei segmentation in spatial omics[J]. Optics and Precision Engineering, 2025, 33(9): 1456

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category:

    Received: Jan. 9, 2025

    Accepted: --

    Published Online: Jul. 22, 2025

    The Author Email: Lianqun ZHOU (zhoulq@sibet.ac.cn)

    DOI:10.37188/OPE.20253309.1456

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology