• Acta Photonica Sinica
  • Vol. 52, Issue 8, 0817002 (2023)
Muqing ZHANG1,2, Yutong HAN1,2, Bonian CHEN1,2, and Jianxin ZHANG1,2,*
Author Affiliations
  • 1School of Computer Science and Engineering,Dalian Minzu University,Dalian 116600,China
  • 2Institute of Machine Intelligence and Biocomputing,Dalian Minzu University,Dalian 116600,China
  • show less
    DOI: 10.3788/gzxb20235208.0817002 Cite this Article
    Muqing ZHANG, Yutong HAN, Bonian CHEN, Jianxin ZHANG. Magnetic Resonance Imaging Brain Tumor Segmentation Using Multiscale Ghost Generative Adversarial Network[J]. Acta Photonica Sinica, 2023, 52(8): 0817002 Copy Citation Text show less
    Overall architecture of multiscale ghost generate adversarial network for MRI brain tumor segmentation
    Fig. 1. Overall architecture of multiscale ghost generate adversarial network for MRI brain tumor segmentation
    Structural of the ghost module
    Fig. 2. Structural of the ghost module
    Structure of the multiscale generator
    Fig. 3. Structure of the multiscale generator
    Structure of the discriminator
    Fig. 4. Structure of the discriminator
    MRI images in different modes and ground truth
    Fig. 5. MRI images in different modes and ground truth
    Visualization segmentation results on the BraTS 2020 training dataset
    Fig. 6. Visualization segmentation results on the BraTS 2020 training dataset
    MethodsDSC
    ETWTTCAverage
    U-Net(baseline)0.7840.9000.7970.827
    GU-Net0.8260.9150.8270.856
    GU-Net+Ghost0.8210.9210.8350.859
    GU-Net+Mul0.8240.9200.8460.863
    MG2AN(ours)0.8250.9220.8540.867
    Table 1. Ablation experiments on BraTS2020 training dataset
    MethodsDSC
    ETWTTCAverage
    U-Net(baseline)0.7690.8940.7990.821
    GU-Net0.7710.8990.8030.824
    GU-Net+Ghost0.7740.9010.8090.828
    GU-Net+Mul0.7740.9010.8210.832
    MG2AN(ours)0.7820.9030.8260.837
    Table 2. Ablation experiments on BraTS2020 validation dataset
    MethodsDSCHausdorff95
    ETWTTCAverageETWTTCAverage
    Ref.[110.7070.8780.7790.788----
    Ref.[140.7090.8730.8140.79912.3015.4512.4713.40
    Ref.[150.7890.9000.8190.8363.735.646.055.14
    Ref.[170.7670.8970.7900.8184.616.928.406.64
    Ref.[230.7520.8990.8150.82212.567.398.069.34
    MG2AN0.7700.9020.8360.8363.747.065.555.45
    Table 3. Compared results with representative methods on BraTS2019 validation set
    MethodsDSCHausdorff95
    ETWTTCAverageETWTTCAverage
    Ref.[120.7600.9000.8000.82026.85.2512.414.82
    Ref.[130.7630.8990.8160.82633.265.287.7415.43
    Ref.[150.7870.9010.8170.83517.954.969.7710.89
    Ref.[180.7870.8720.8110.82324.366.4418.9516.58
    Ref.[230.7640.8990.8100.82432.567.3912.0617.33
    MG2AN0.7820.9030.8260.83729.414.548.9113.43
    Table 4. Compared results with representative methods on BraTS2020 validation set
    Muqing ZHANG, Yutong HAN, Bonian CHEN, Jianxin ZHANG. Magnetic Resonance Imaging Brain Tumor Segmentation Using Multiscale Ghost Generative Adversarial Network[J]. Acta Photonica Sinica, 2023, 52(8): 0817002
    Download Citation