Laser & Optoelectronics Progress, Volume. 61, Issue 12, 1217001(2024)

Sparse-View Projection Spectral CT Reconstruction via HAMEN

Junyu Qi1, Zaifeng Shi1,2、*, Fanning Kong1, Tianhao Ge1, and Lili Zhang1
Author Affiliations
  • 1School of Microelectronics, Tianjin University, Tianjin 300072, China
  • 2Tianjin Key Laboratory of Imaging and Sensing Microelectronic Technology, Tianjin 300072, China
  • show less
    Figures & Tables(10)
    Workflow of the reconstruction method
    HAMEN structure
    Module structures of channel attention and spatial attention in HAM
    Effect comparison of different reconstruction methods
    Effect comparison of different reconstruction methods using 30 projection views
    • Table 1. Parameter settings for sparse-view projection spectral CT

      View table

      Table 1. Parameter settings for sparse-view projection spectral CT

      ParameterContent
      Projection geometryCone beam
      PCD modelPcTK27,Johns Hopkins University
      Energy bin20‒35,35‒50,50‒65,65‒80,80‒120 keV
      Image size256×256
      Sparse-view strategy60 scans,every 6°
    • Table 2. Quantitative metrics of brain slice test results of different reconstruction methods

      View table

      Table 2. Quantitative metrics of brain slice test results of different reconstruction methods

      Energy binParameterFDKTVRED-CNN17MultiChU-Net19MWCNN20HAMEN
      20‒35 keVPSNR/dB25.8933.4034.4235.0937.1637.64
      SSIM0.42150.89630.98020.98370.99090.9935
      FSIM0.93030.96220.98230.98150.99250.9946
      35‒50 keVPSNR/dB27.0531.9235.6835.9037.2337.95
      SSIM0.66340.94600.98250.98590.99430.9974
      FSIM0.92750.96310.98750.98370.99400.9948
      50‒65 keVPSNR/dB27.1531.2037.1037.2938.4538.69
      SSIM0.80510.96100.98100.98650.99490.9968
      FSIM0.92500.96340.98860.98670.99420.9948
      65‒80 keVPSNR/dB27.8532.5337.9038.2138.8339.09
      SSIM0.85170.96970.98720.98770.99520.9970
      FSIM0.92380.96360.98860.98780.99420.9948
      80‒120 keVPSNR/dB28.0634.8438.8038.9339.8540.03
      SSIM0.88390.98200.98980.99050.99570.9972
      FSIM0.92300.96230.98870.98760.99430.9947
    • Table 3. Quantitative metrics of brain slice test results of HAMEN with different structures

      View table

      Table 3. Quantitative metrics of brain slice test results of HAMEN with different structures

      HAMMFFEEDF20‒35 keV50‒65 keV80‒120 keV
      PSNR /dBSSIMFSIMPSNR /dBSSIMFSIMPSNR /dBSSIMFSIM
      34.340.98350.983536.770.98720.983937.190.98850.9863
      36.410.98620.987737.990.98910.988138.670.99240.9889
      36.730.98920.989137.980.99230.989638.790.99370.9894
      34.830.98590.987937.080.98790.988437.460.98970.9881
      37.620.99310.992138.680.99650.992439.980.99690.9925
      37.120.98920.991338.120.99260.991539.650.99340.9911
      37.640.99350.994638.690.99680.994840.030.99720.9947
    • Table 4. Quantitative metrics of brain slice test results using 30 projection views

      View table

      Table 4. Quantitative metrics of brain slice test results using 30 projection views

      Energy binParameterFDKTVRED-CNN17MultiChU-Net19MWCNN20HAMEN
      20‒35 keVPSNR /dB21.1527.4131.0234.7335.1435.71
      SSIM0.23360.64230.70560.83480.84500.8546
      FSIM0.91180.92060.97810.97740.97780.9816
      35‒50 keVPSNR /dB22.2429.1733.5935.2035.8236.17
      SSIM0.41580.89860.90910.92090.92550.9348
      FSIM0.91410.95900.98380.97910.98160.9830
      50‒65 keVPSNR /dB22.2329.0535.9436.8237.4137.76
      SSIM0.57830.93360.94420.95490.95870.9632
      FSIM0.91420.95990.98520.98180.98330.9844
      65‒80 keVPSNR /dB22.8730.2637.0237.1738.2038.83
      SSIM0.64650.93720.95210.96560.96960.9719
      FSIM0.91400.95310.98560.98290.98380.9844
      80‒120 keVPSNR /dB23.0230.7038.1238.5839.2139.60
      SSIM0.70010.95270.96110.97330.97390.9775
      FSIM0.91410.95700.98580.98150.98200.9841
    • Table 5. Quantitative metrics of reconstructed images of energy bin 1 with different noise levels

      View table

      Table 5. Quantitative metrics of reconstructed images of energy bin 1 with different noise levels

      Photon densityParameterFDKTVRED-CNN17MultiChU-Net19MWCNN20HAMEN
      5×106PSNR /dB25.8933.4034.4235.0937.1637.64
      SSIM0.42150.89630.98020.98370.99090.9935
      FSIM0.93030.96220.98230.98150.99250.9946
      5×105PSNR /dB23.8531.4232.1734.3736.3136.90
      SSIM0.41090.88290.97360.97690.98030.9863
      FSIM0.92390.92180.96570.97980.98770.9884
      5×104PSNR /dB21.1728.5929.6233.2335.3635.89
      SSIM0.39370.86830.95740.96570.97060.9742
      FSIM0.91850.91640.95480.97320.98120.9827
    Tools

    Get Citation

    Copy Citation Text

    Junyu Qi, Zaifeng Shi, Fanning Kong, Tianhao Ge, Lili Zhang. Sparse-View Projection Spectral CT Reconstruction via HAMEN[J]. Laser & Optoelectronics Progress, 2024, 61(12): 1217001

    Download Citation

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

    Category: Medical Optics and Biotechnology

    Received: Jul. 12, 2023

    Accepted: Sep. 18, 2023

    Published Online: Jun. 5, 2024

    The Author Email: Zaifeng Shi (shizaifeng@tju.edu.cn)

    DOI:10.3788/LOP231696

    CSTR:32186.14.LOP231696

    Topics