• Acta Optica Sinica
  • Vol. 39, Issue 4, 0410002 (2019)
Yuexiang Shi1 and Cai Chen2、*
Author Affiliations
  • 1 College of Information Engineering, Xiangtan University, Xiangtan, Hunan 411105, China
  • 2 LED Lighting Drive and Control Application Engineering Technology Research Center of Guizhou, Tongren, Guizhou 554300, China
  • show less
    DOI: 10.3788/AOS201939.0410002 Cite this Article Set citation alerts
    Yuexiang Shi, Cai Chen. Non-Rigid Registration Segmentation Algorithm Based on Optimal Atlas Multi-Model Image[J]. Acta Optica Sinica, 2019, 39(4): 0410002 Copy Citation Text show less
    Framework of registration and segmentation
    Fig. 1. Framework of registration and segmentation
    Regional fusion segmentation of lung tissues. (a)-(c) Images to be fused; (d)-(f) fusion and segmentation images; (g)-(j) binary images of mask
    Fig. 2. Regional fusion segmentation of lung tissues. (a)-(c) Images to be fused; (d)-(f) fusion and segmentation images; (g)-(j) binary images of mask
    Atlas image matching based on optimal similarity
    Fig. 3. Atlas image matching based on optimal similarity
    Adaptive weighted meshing of locally B-spline. (a) 2D meshing; (b) adaptive mobile grid-model
    Fig. 4. Adaptive weighted meshing of locally B-spline. (a) 2D meshing; (b) adaptive mobile grid-model
    Optimization process of the iterative computation with L-BFGS
    Fig. 5. Optimization process of the iterative computation with L-BFGS
    Image registration-segmentation process and results based on different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    Fig. 6. Image registration-segmentation process and results based on different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    MSE trend of image registration under different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    Fig. 7. MSE trend of image registration under different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    Mask image and original mask effect under different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    Fig. 8. Mask image and original mask effect under different Atlas similarities. (a) 0.93; (b) 0.90; (c) 0.88; (d) 0.85
    Detection of location of lobar fissures based on proposed registration-segmentation method. (a) Middle layer of lung; (b) upper layer of lung; (c) lower layer of lung
    Fig. 9. Detection of location of lobar fissures based on proposed registration-segmentation method. (a) Middle layer of lung; (b) upper layer of lung; (c) lower layer of lung
    Iterative optimization of registration evaluation criteria. (a) NMI and CC; (b) NMI and SSD; (c) SSD and CC; (d) ROI accuracy
    Fig. 10. Iterative optimization of registration evaluation criteria. (a) NMI and CC; (b) NMI and SSD; (c) SSD and CC; (d) ROI accuracy
    Optimizing results of average evaluation indicators. (a) Iterative optimization of NMI; (b) iterative optimization of SSD; (c) iterative optimization of CC; (d) iterative optimization of ROI accuracy
    Fig. 11. Optimizing results of average evaluation indicators. (a) Iterative optimization of NMI; (b) iterative optimization of SSD; (c) iterative optimization of CC; (d) iterative optimization of ROI accuracy
    Three-dimensional reconstruction model for segmentation of lung registration results. (a) Front view and (b) side view for the similarity of 0.90; (c) front view and (d) side view for the similarity of 0.93
    Fig. 12. Three-dimensional reconstruction model for segmentation of lung registration results. (a) Front view and (b) side view for the similarity of 0.90; (c) front view and (d) side view for the similarity of 0.93
    Serial numberMap image numberS(t)
    1Atlas_1718702_0066T0.906
    2Atlas_1718702_0068T0.895
    3Atlas_1718702_0071T0.892
    4Atlas_1718702_0080T0.875
    5Atlas_1718705_0075T0.870
    6Atlas_1714609_0084T0.867
    7Atlas_1713559_0058T0.854
    8Atlas_1713419_0103T0.801
    Table 1. Search results with optimal Atlas image strategy
    NMISSDCCROI accuracyCalculating time /s
    Before registration1.10360.06240.80320.8903-
    Affine registration1.24830.04450.90640.918595.20
    Global MI registration1.42160.03510.92560.923865.42
    Nifty_Reg registration1.56820.03140.94420.935646.33
    Elastix registration1.54280.03290.94170.930742.85
    Proposed method1.84320.01120.98650.956028.44
    Table 2. Comparison of mean values of image registration data from different algorithms
    Yuexiang Shi, Cai Chen. Non-Rigid Registration Segmentation Algorithm Based on Optimal Atlas Multi-Model Image[J]. Acta Optica Sinica, 2019, 39(4): 0410002
    Download Citation