Opto-Electronic Engineering, Volume. 51, Issue 4, 240011-1(2024)

Graph neural network-based WSI cancer survival prediction method

Shijie Ye and Yongxiong Wang*
Author Affiliations
  • Institute of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
  • show less
    Figures & Tables(10)
    Architecture of the BC GraphSurv model mainly including modules such as WSI preprocessing, (a) Feature extraction and graph structure generation, (b) WA-GAT branch, (c) MP-GCN branches, and (d) Feature fusion
    Schematic diagram of HF-Net
    Schematic diagram of WA-GAT
    Comparison of KM curves and P-values of several commonly used methods
    Comparison of patchs with different IG values
    Visualization of WSI pathological environment
    • Table 1. Comparison of experimental results of feature extraction network

      View table
      View in Article

      Table 1. Comparison of experimental results of feature extraction network

      C-indexSDParameters/MB
      ResNet500.72250.01197.49
      EfficientNet-b50.73060.022115.93
      HF-Net0.79500.013113.72
    • Table 2. Comparison of ablation experiment results

      View table
      View in Article

      Table 2. Comparison of ablation experiment results

      WA-GATMP-GCNGATGCNGFC-indexSD
      0.75060.021
      0.72170.019
      0.78420.007
      0.78120.008
      0.79100.003
      0.79500.013
    • Table 3. First comparison experiments of different methods

      View table
      View in Article

      Table 3. First comparison experiments of different methods

      MethodBRCAKIRC
      C-indexSDC-indexSD
      MLP0.61760.0270.58620.019
      Attention-MIL[26] (2018)0.70910.0520.65900.044
      WSISA[8] (2017)0.68020.0830.61510.057
      DeepGraphSurv[12] (2018)0.74020.0120.69450.045
      Patch GCN[13] (2021)0.75140.0360.67750.067
      H2-MIL[27] (2022)0.73380.0550.69150.024
      Tea-Graph[28] (2022)0.75410.0210.71090.023
      HEAT[3] (2023)0.75290.0090.70110.018
      BC-GraphSurv0.79500.0130.74580.020
    • Table 4. Second comparison experiments of different methods

      View table
      View in Article

      Table 4. Second comparison experiments of different methods

      MethodBRCAKIRC
      C-indexSDC-indexSD
      MLP0.61760.0270.58620.019
      MIL-Transformer[29] (2021)0.69050.0460.66420.029
      SurvTRACE[30] (2022)0.73820.0190.69740.027
      BC-GraphSurv0.79500.0130.74580.020
    Tools

    Get Citation

    Copy Citation Text

    Shijie Ye, Yongxiong Wang. Graph neural network-based WSI cancer survival prediction method[J]. Opto-Electronic Engineering, 2024, 51(4): 240011-1

    Download Citation

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

    Category: Article

    Received: Jan. 9, 2024

    Accepted: Mar. 12, 2024

    Published Online: Jul. 8, 2024

    The Author Email: Wang Yongxiong (王永雄)

    DOI:10.12086/oee.2024.240011

    Topics