Acta Optica Sinica, Volume. 44, Issue 8, 0834001(2024)
Limited-Angle CT Image Reconstruction Based on Swin-Transformer Iterative Unfolding for PTCT Imaging
Figures & Tables(15)
Fig. 1. Schematic diagram of PTCT geometric model. (a) PTCT three-dimensional geometric model; (b) PTCT two-dimensional geometric model; (c) PTCT projection data distribution
Fig. 7. RCT reconstruction results of 2DeteCT dataset obtained by different methods. (a)-(g) Reconstruction results of case 1; (h)-(n) reconstruction results of case 2
Fig. 8. PTCT reconstruction results of 2DeteCT dataset obtained by different methods. (a)-(g) Reconstruction results of case 1; (h)-(n) reconstruction results of case 2
Fig. 9. Reconstruction results of ablation experiment of different ablation networks on 2DeteCT dataset
Fig. 10. Box diagrams of numerical distribution of sample indexes of ablation network for test set. Numerical distributions of (a) RMSE, (b) SSIM, and (c) PSNR of test set samples in different ablation networks
Fig. 11. PTCT reconstruction results of ACCC dataset obtained by different methods. (a)-(g) Reconstruction results of case 1; (h)-(n) reconstruction results of case 2
Table 1. Experimental parameter settings
View tableTable 1. Experimental parameter settings
Item 2DeteCT ACCC Scanning method RCT PTCT PTCT Scan angle /（°） 90 90（β） 90（β） Tube voltage /kV — — 100 Tube current /μA — — 83 SO /mm 100 100 80 SD /mm 240 240 259 Detector pixel size /mm 0.085 0.085 0.085 Array length /pixel 256 256 1536 Reconstruction matrix size /（pixel×pixel） 256×256 256×256 256×256 Number of sampling points 150 150 150 Sampling model — Equl-spatial Equl-spatial
Table 2. Reconstruction image indexes of different methods on 2DeteCT dataset
View tableTable 2. Reconstruction image indexes of different methods on 2DeteCT dataset
Method PSNR of case 1 /dB SSIM of case 1 RMSE of case 1 PSNR of case 2 /dB SSIM of case 2 RMSE of case 2 FBP 22.0108 0.3628 0.0793 20.4884 0.3350 0.0945 SIRT 26.0630 0.6650 0.0499 25.2571 0.6447 0.0546 SwinIR 33.1356 0.9354 0.0220 32.9674 0.9355 0.0225 FISTA-Net 32.4324 0.8980 0.0239 32.4536 0.8982 0.0238 LEARN 34.1125 0.9223 0.0197 34.4578 0.9240 0.0189 STICA-Net 36.2793 0.9403 0.0147 37.0529 0.9426 0.0140
Table 3. Reconstruction image indexes of different methods on ACCC dataset
View tableTable 3. Reconstruction image indexes of different methods on ACCC dataset
Network PSNR of case 1 /dB SSIM of case 1 RMSE of case 1 PSNR of case 2 /dB SSIM of case 2 RMSE of case 2 FBP 12.8890 0.4429 0.2268 13.7951 0.4646 0.2043 SIRT 20.7439 0.7661 0.0918 20.3331 0.7498 0.0962 SwinIR 31.9921 0.9292 0.0251 33.6882 0.9381 0.0207 FISTA-Net 29.2278 0.8625 0.0346 30.9193 0.8929 0.0285 LEARN 26.4515 0.8396 0.0476 24.4574 0.8379 0.0599 STICA-Net 33.5012 0.9560 0.0211 35.5502 0.9530 0.0167
Table 4. Number of parameters, number of floating-point operations, and reasoning time of different models
View tableTable 4. Number of parameters, number of floating-point operations, and reasoning time of different models
Model SwinIR LEARN FISTA-Net STICA-Net Number of parameters /106 1.22 3.01 0.52 5.38 Number of floating-point operations /109 81.22 196.61 51.00 352.90 Time /ms 34.00 235.26 54.81 314.28
Tools
Get Citation
Copy Citation Text
Wei Yuan, Yarui Xi, Chuandong Tan, Chuanjiang Liu, Guorong Zhu, Fenglin Liu. Limited-Angle CT Image Reconstruction Based on Swin-Transformer Iterative Unfolding for PTCT Imaging[J]. Acta Optica Sinica, 2024, 44(8): 0834001
Paper Information
Category: X-Ray Optics
Received: Nov. 22, 2023
Accepted: Jan. 16, 2024
Published Online: Apr. 11, 2024
The Author Email: Liu Fenglin (liufl@cqu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20