Acta Optica Sinica, Volume. 40, Issue 3, 0310002(2020)

Computed Tomography Image Classification Algorithm Based on Improved Deep Residual Network

Sheng Huang, Feifei Li**, and Qiu Chen*
Author Affiliations
  • School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
  • show less
    Figures & Tables(14)
    Residual block
    Architecture of SK-ResNet
    Structure of discriminator
    Process of unsupervised pre-training
    Example of labeled areas of lung (areas 1, 2 denote pathology area)
    Image patch examples. (a) Source domain; (b) target domain
    Trend graph of classification results for training sets with different scales
    Misclassified examples
    • Table 1. Different dataset separation schemes

      View table

      Table 1. Different dataset separation schemes

      SchemeSize of training setSize of test setSplit manner
      A413210315-fold
      B25602479Case select
      C11289750Random select
    • Table 2. Comparison of classification results of SK-ResNet with different depths

      View table

      Table 2. Comparison of classification results of SK-ResNet with different depths

      NetworkStructure ofResNet blockfavg
      ABC
      SK-ResNet 10[1,1,1,1]0.95860.94530.9651
      SK-ResNet 14[2,2,1,1]0.96090.94760.9640
      SK-ResNet 18[2,2,2,2]0.95330.93540.9553
      SK-ResNet 34[3,4,6,3]0.94970.93290.9549
    • Table 3. Comparison of results on SK-ResNet before and after using transfer learning

      View table

      Table 3. Comparison of results on SK-ResNet before and after using transfer learning

      MethodABC
      Without transfer0.96090.94760.9640
      DIM0.97990.96540.9707
      DIM+PM0.98180.96770.9756
    • Table 4. Classification confusion matrix of SK-ResNet

      View table

      Table 4. Classification confusion matrix of SK-ResNet

      Ground truthPredicted result
      EMFBGGNMMN
      EM0.91000.080.01
      FB00.960.0400
      GG00.020.9800
      NM000.030.940.03
      MN0000.060.94
    • Table 5. Classification confusion matrix of SK-ResNet using transfer learning

      View table

      Table 5. Classification confusion matrix of SK-ResNet using transfer learning

      Ground truthPredicted result
      EMFBGGNMMN
      EM0.97000.030
      FB00.960.0400
      GG00.020.9800
      NM000.030.960.01
      MN0000.030.97
    • Table 6. Comparison of classification performances of SK-ResNet and other methods

      View table

      Table 6. Comparison of classification performances of SK-ResNet and other methods

      MethodABC
      Ref.[14]0.77250.76440.7987
      Ref. [16]0.94650.92150.9392
      Ref. [23]0.94830.92690.9554
      AlexNet[11]0.89620.88210.9226
      Pre-trained AlexNet[11]0.94710.93370.9609
      Pre-trained VGG-16[24]0.96030.94350.9664
      ResNet-18[17]0.93710.93060.9432
      Pre-trained ResNet-18 [17]0.95810.94430.9651
      Pre-trained DenseNet-121[25]0.97410.95790.9715
      SK-ResNet0.96090.94760.9640
      SK-ResNet (transfer)0.98180.96770.9756
    Tools

    Get Citation

    Copy Citation Text

    Sheng Huang, Feifei Li, Qiu Chen. Computed Tomography Image Classification Algorithm Based on Improved Deep Residual Network[J]. Acta Optica Sinica, 2020, 40(3): 0310002

    Download Citation

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

    Category: Image Processing

    Received: Sep. 3, 2019

    Accepted: Oct. 21, 2019

    Published Online: Feb. 10, 2020

    The Author Email: Feifei Li (feifeilee@ieee.org), Qiu Chen (q.chen@ieee.org)

    DOI:10.3788/AOS202040.0310002

    Topics