Sparse Scanning Dual-Domain Image Reconstruction Model Based on Deep Learning

Zhengheng Li; Chenyin Ni; Chunmin Zhang

doi:10.3788/LOP242283

Laser & Optoelectronics Progress, Volume. 62, Issue 14, 1434001(2025)

Sparse Scanning Dual-Domain Image Reconstruction Model Based on Deep Learning

Zhengheng Li^1、**, Chenyin Ni^2、*, and Chunmin Zhang³

¹School of Electronic and Optical Engineering, Nanjing University of Science & Technology, Nanjing 210094, Jiangsu , China

²School of Physics, Nanjing University of Science & Technology, Nanjing 210094, Jiangsu , China

³CT Business Division, Unicomp Technology Co., Ltd., Wuxi 214111, Jiangsu , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(20)

Fig. 1. Schematic diagrams of CBCT geometric model. (a) Industrial CBCT scanning mode; (b) CBCT sparse-view scanning imaging mode

Download full size

Fig. 2. Reconstruction process of wavelet transform and inverse transform

Download full size

Fig. 3. Forward and reverse processes of DDPM for sparse scan CT reconstruction

Download full size

Fig. 4. CT reconstruction images with different time steps

Download full size

Fig. 5. Comparison of grayscale values of CT reconstructed images with different time steps

Download full size

Fig. 6. Linear reconstruction process of sparse view CT

Download full size

Fig. 7. Low frequency diffusion module and high frequency refinement module

Download full size

Fig. 8. Generator and discriminator network framework

Download full size

Fig. 9. Domain transformation of WDRG

Download full size

Fig. 10. Micro focus industrial CBCT equipment

Download full size

Fig. 11. Reconstruction results of slice 1 by different algorithms. (a) Reference image; (b) FBP; (c) SART; (d) WGAN; (e)DDPM; (f) DDDM; (g) WDRG

Download full size

Fig. 12. Reconstruction results of slice 2 by different algorithms. (a) Reference image; (b) FBP; (c) SART; (d) WGAN; (e) DDPM; (f) DDDM; (g) WDRG

Download full size

Fig. 13. Reconstruction results of different algorithms in BGA dataset. (a) Reference image; (b) FBP; (c) SART; (d) WGAN; (e) DDPM; (f) DDDM; (g) WDRG

Download full size

Fig. 14. Reconstruction results of BGA dataset when $N_{S} = 90$ . (a) Reference image; (b) FBP; (c) SART; (d) WGAN; (e) DDPM; (f) DDDM; (g) WDRG

Download full size

Table 1. Quantitative results of CT reconstruction
View table
Table 1. Quantitative results of CT reconstruction
t PSNR /dB SSIM RMSE
20 31.18 0.7324 0.0276
50 33.97 0.9182 0.0200
100 33.10 0.9253 0.0221
200 34.82 0.9116 0.0182
500 31.58 0.9216 0.0264
1000 31.57 0.8193 0.0264

Table 2. Dataset experimental parameter setting

View table

Table 2. Dataset experimental parameter setting

Parameter	2DeteCT	BGA
Scanning methodt	CBCT	CBCT
Rotate angle /（°）	30	30
Tube voltage /kV		105
Current /μA		120
Focus to object distance /mm	45	36.87
Focus to detector distance /mm	520	313.39
Detector pixel size×array length /（mm×pixel）	0.085×512	0.085×1536
Reconstruction matrix /（pixel×pixel）	512×512	512×512

Table 3. Quantitative reconstruction results of slice 1 by different algorithms
View table
Table 3. Quantitative reconstruction results of slice 1 by different algorithms
Algorithm PSNR /dB SSIM RMSE
FBP 20.34 0.4661 0.0962
SART 23.78 0.5352 0.0647
WGAN 29.13 0.8769 0.0350
DDPM 33.74 0.9149 0.0205
DDDM 39.04 0.9639 0.0112
WDRG 41.79 0.9654 0.0081

Table 4. Quantitative reconstruction results of slice 2 by different algorithms
View table
Table 4. Quantitative reconstruction results of slice 2 by different algorithms
Methods PSNR /dB SSIM RMSE
FBP 18.18 0.3923 0.1232
SART 23.63 0.5176 0.0659
WGAN 27.17 0.8502 0.0438
DDPM 32.65 0.9319 0.0233
DDDM 38.59 0.9560 0.0118
WDRG 39.35 0.9574 0.0108

Table 5. Quantitative reconstruction results of different algorithms on BGA dataset

View table

Table 5. Quantitative reconstruction results of different algorithms on BGA dataset

Algorithm	slice 1			slice 2
Algorithm	PSNR /dB	SSIM	RMSE	PSNR /dB	SSIM	RMSE
FBP	15.71	0.1577	0.1640	13.33	0.2841	0.2154
SART	19.71	0.1643	0.1034	17.87	0.3604	0.1278
WGAN	24.52	0.9479	0.0594	20.88	0.8250	0.0903
DDPM	37.27	0.9905	0.0137	25.87	0.9779	0.0509
DDDM	37.96	0.9937	0.0126	28.55	0.9582	0.0374
WDRG	39.33	0.9947	0.0108	33.39	0.9905	0.0214

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when $N_{S} = 90$

View table

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when $N_{S} = 90$

Algorithm	slice 1			slice 2
Algorithm	PSNR /dB	SSIM	RMSE	PSNR /dB	SSIM	RMSE
FBP	10.79	0.2106	0.2888	13.91	0.1481	0.2016
SART	16.54	0.2837	0.1489	21.02	0.1797	0.0889
WGAN	23.11	0.6424	0.0699	27.58	0.4679	0.0418
DDPM	30.52	0.9593	0.0298	31.12	0.9758	0.0278
DDDM	33.13	0.9815	0.0220	34.50	0.9912	0.0188
WDRG	36.17	0.9734	0.0155	36.29	0.9904	0.0153

Tools

Get Citation

Copy Citation Text

Zhengheng Li, Chenyin Ni, Chunmin Zhang. Sparse Scanning Dual-Domain Image Reconstruction Model Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2025, 62(14): 1434001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: X-Ray Optics

Received: Nov. 19, 2024

Accepted: Feb. 7, 2025

Published Online: Jul. 17, 2025

The Author Email: Zhengheng Li (li_zhengheng@njust.edu.cn), Chenyin Ni (chenyin.ni@njust.edu.cn)

DOI:10.3788/LOP242283

CSTR:32186.14.LOP242283

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Quantitative results of CT reconstruction

Table 1. Quantitative results of CT reconstruction

Table 2. Dataset experimental parameter setting

Table 2. Dataset experimental parameter setting

Table 3. Quantitative reconstruction results of slice 1 by different algorithms

Table 3. Quantitative reconstruction results of slice 1 by different algorithms

Table 4. Quantitative reconstruction results of slice 2 by different algorithms

Table 4. Quantitative reconstruction results of slice 2 by different algorithms

Table 5. Quantitative reconstruction results of different algorithms on BGA dataset

Table 5. Quantitative reconstruction results of different algorithms on BGA dataset

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when NS=90

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when NS=90

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when $N_{S} = 90$

Table 6. Quantitative reconstruction results of BGA dataset by different algorithms when $N_{S} = 90$