• Acta Photonica Sinica
  • Vol. 52, Issue 12, 1210001 (2023)
Lihao DING, Zhishan GAO, Dan ZHU*, Qun YUAN, and Zhenyan GUO
Author Affiliations
  • School of Electronic Engineering and Optoelectronic Technology,Nanjing University of Science and Technology,Nanjing 210094,China
  • show less
    DOI: 10.3788/gzxb20235212.1210001 Cite this Article
    Lihao DING, Zhishan GAO, Dan ZHU, Qun YUAN, Zhenyan GUO. Classification Method of Breast Tissue OCT Images Based on a Double Filtering Residual Network[J]. Acta Photonica Sinica, 2023, 52(12): 1210001 Copy Citation Text show less
    DF-ResNet model workflow
    Fig. 1. DF-ResNet model workflow
    Three types of raw data images
    Fig. 2. Three types of raw data images
    Effect of α parameters on model performance
    Fig. 3. Effect of α parameters on model performance
    Training curve for model performance
    Fig. 4. Training curve for model performance
    ROC curve of DF-ResNet model
    Fig. 5. ROC curve of DF-ResNet model
    The thermodynamic diagram of the two models
    Fig. 6. The thermodynamic diagram of the two models
    Bar chart of model parameter quantity and classification accuracy
    Fig. 7. Bar chart of model parameter quantity and classification accuracy
    NameIndicator
    SystemWindows 11
    Memory32 GB
    Central processing unitAMD Ryzen 7 5800H(3.20 GHz)
    Graphics processing unitNVIDIA RTX3070 GPU
    Table 1. Detailed system parameters
    MethodImagesSenSpePPVNNPAccuracy
    ResNet-28FB0.987 50.995 30.990 60.993 80.938 5
    S0.878 10.971 90.939 80.941 0
    T0.950 00.940 60.888 90.974 1
    Oct_ResNet-28FB0.987 50.987 50.975 30.993 70.954 2
    S0.915 60.976 60.951 30.958 6
    T0.959 40.967 20.936 00.979 4
    DF-ResNetFB0.990 60.992 20.984 50.995 30.968 8
    S0.946 90.981 30.961 90.973 6
    T0.968 80.979 70.959 80.984 3
    Table 2. Comparison of performance among three models
    MethodImagesSenSpePPVNNPAccuracy
    DF-ResNetFB1.0001.0001.0001.0000.941 2
    S0.882 00.971 00.938 00.943 0
    T0.941 00.941 00.889 00.970 0
    Table 3. The recognition and classification performance of DF-ResNet on a new dataset
    MethodAccuracyParameter quantity/MiB
    VGG-190.793 9143.7
    EfficientNet0.916 75.290
    ResNet-280.938 519.14
    DenseNet-1690.957 214.15
    Oct_ResNet-280.954 215.99
    DF-ResNet0.968 816.52
    Table 4. model parameter quantity and classification accuracy
    Lihao DING, Zhishan GAO, Dan ZHU, Qun YUAN, Zhenyan GUO. Classification Method of Breast Tissue OCT Images Based on a Double Filtering Residual Network[J]. Acta Photonica Sinica, 2023, 52(12): 1210001
    Download Citation