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

Laser & Optoelectronics Progress, Volume. 58, Issue 8, 0810020(2021)

A Method for Brain Tumor Segmentation Using Cascaded Modified U-Net

Jinghui Chu, Kailong Huang, and Wei Lü*
Author Affiliations
  • School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (9)
    Equations (0)
    References (29)
    Cited By (1)
    Tools
    Paper Information
    Topics
    Figures & Tables(9)
    Three stages cascaded architecture

    Fig. 1. Three stages cascaded architecture

    Download full size
    Multi-levels feature maps fusing

    Fig. 2. Multi-levels feature maps fusing

    Download full size
    Dilated convolution

    Fig. 3. Dilated convolution

    Download full size
    Condition random fields as recurrent neural network

    Fig. 4. Condition random fields as recurrent neural network

    Download full size
    Modified U-Net structure

    Fig. 5. Modified U-Net structure

    Download full size
    Different modalities of MRI and label images

    Fig. 6. Different modalities of MRI and label images

    Download full size
    Results of segmentation

    Fig. 7. Results of segmentation

    Download full size

    • Table 1. Dice scores and Hausdorff distances on BraTS 2018 validation set (66 cases) with different methods

      View table

      Table 1. Dice scores and Hausdorff distances on BraTS 2018 validation set (66 cases) with different methods

      ModelDiceWTDiceTCDiceETHDWTHDTCHDET
      Base0.891500.801730.7779814.900348.638815.17406
      Base+DC0.902980.819790.775628.077728.349645.23973
      Base+DC+MFF0.905870.820190.786395.728048.425804.39199
      Base+DC+MFF(Dice+CE)0.907490.820530.779835.519818.678493.06801
      Base+DC+MFF+CRF(Dice+CE)0.909320.825360.785493.818777.848734.32642

    • Table 2. Performance comparison of segmentation methods on BraTS 2018 validation set (66 cases)

      View table

      Table 2. Performance comparison of segmentation methods on BraTS 2018 validation set (66 cases)

      ModelDice WTDice TCDice ETHD WTHD TCHD ET
      Ours0.909320.825360.785493.818777.848734.32642
      Myronenko[12] (single model)0.90420.85960.81454.48348.27773.8048
      Mehta et al[28] (ensembles)0.90920.85300.81374.45197.15352.7092
      Mehta et al[28]0.9090.8250.7884.9238.3163.520
      Wang et al[27]0.9080.8690.8075.866.093.01
      Hua et al[29]0.90480.83640.77685.17596.27803.5123
      McKinley et al[26]0.9030.8470.7964.174.933.55
    Tools

    Get Citation

    Copy Citation Text

    Jinghui Chu, Kailong Huang, Wei Lü. A Method for Brain Tumor Segmentation Using Cascaded Modified U-Net[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810020

    Download Citation

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

    Category: Image Processing

    Received: Aug. 24, 2020

    Accepted: Sep. 20, 2020

    Published Online: Apr. 12, 2021

    The Author Email: Wei Lü (luwei@tju.edu.cn)

    DOI:10.3788/LOP202158.0810020

    Topics

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