• Acta Optica Sinica
  • Vol. 40, Issue 17, 1710001 (2020)
Wenxiu Zhang1、2、3、*, Zhencai Zhu1、2、3, Yonghe Zhang1、2、3, Xinyu Wang1、2, and Guopeng Ding1、2
Author Affiliations
  • 1Innovation Academy for Microsatellite, Chinese Academy of Sciences, Shanghai 201203, China
  • 2Key Laboratory of Microsatellites, Chinese Academy of Sciences, Shanghai 201203, China
  • 3University of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    DOI: 10.3788/AOS202040.1710001 Cite this Article Set citation alerts
    Wenxiu Zhang, Zhencai Zhu, Yonghe Zhang, Xinyu Wang, Guopeng Ding. Cell Image Segmentation Method Based on Residual Block and Attention Mechanism[J]. Acta Optica Sinica, 2020, 40(17): 1710001 Copy Citation Text show less
    Overall framework of proposed model
    Fig. 1. Overall framework of proposed model
    Structure of residual block
    Fig. 2. Structure of residual block
    Structure of attention mechanism
    Fig. 3. Structure of attention mechanism
    Display of dataset. (a) Training set image; (b) corresponding ground truth
    Fig. 4. Display of dataset. (a) Training set image; (b) corresponding ground truth
    Local image blocks and corresponding ground truth
    Fig. 5. Local image blocks and corresponding ground truth
    Cell segmentation results of group 1. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Fig. 6. Cell segmentation results of group 1. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Cell segmentation results of group 2. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Fig. 7. Cell segmentation results of group 2. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Cell segmentation results of group 3. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Fig. 8. Cell segmentation results of group 3. (a) Origin image; (b) ground truth; (c) Otsu model; (d) FCN-8s model; (e) U-Net model; (f) SegNet model; (g) R2U-Net model; (h) proposed model
    Detail comparison of segmentation results. (a) Ground truth; (b) FCN-8s model; (c) U-Net model; (d) SegNet model; (e) R2U-Net model; (f) proposed model
    Fig. 9. Detail comparison of segmentation results. (a) Ground truth; (b) FCN-8s model; (c) U-Net model; (d) SegNet model; (e) R2U-Net model; (f) proposed model
    ModelPixel AccuracyIoUDice Score
    Otsu0.79060.23470.3778
    FCN-8s0.90080.66720.7964
    U-Net0.93490.74050.8504
    SegNet0.93360.73590.8516
    R2U-Net0.94350.76360.8691
    Proposed model0.94470.78220.8775
    Table 1. Quantitative analysis of different segmentation models
    U-NetResidual blockAttention mechanismPixel AccuracyIoUDice Score
    +0.93470.74820.8592
    ++0.93950.75050.8603
    ++0.93880.75650.8652
    +++0.94630.77570.8776
    Table 2. Influence of each module on whole model
    Wenxiu Zhang, Zhencai Zhu, Yonghe Zhang, Xinyu Wang, Guopeng Ding. Cell Image Segmentation Method Based on Residual Block and Attention Mechanism[J]. Acta Optica Sinica, 2020, 40(17): 1710001
    Download Citation