A sparse-view cone-beam CT reconstruction algorithm based on bidirectional flow field- guided projection completion

doi:10.12122/j.issn.1673-4254.2025.02.21

Abstract

Abstract:

Objective We propose a sparse-view cone-beam CT reconstruction algorithm based on bidirectional flow field guided projection completion (BBC-Recon) to solve the ill-posed inverse problem in sparse-view cone-beam CT imaging. Methods The BBC-Recon method consists of two main modules: the projection completion module and the image restoration module. Based on flow field estimation, the projection completion module, through the designed bidirectional and multi-scale correlators, fully calculates the correlation information and redundant information among projections to precisely guide the generation of bidirectional flow fields and missing frames, thus achieving high-precision completion of missing projections and obtaining pseudo complete projections. The image restoration module reconstructs the obtained pseudo complete projections and then refines the image to remove the residual artifacts and further improve the image quality. Results The experimental results on the public datasets of Mayo Clinic and Guilin Medical University showed that in the case of a 4-fold sparse angle, compared with the suboptimal method, the BBC-Recon method increased the PSNR index by 1.80% and the SSIM index by 0.29%, and reduced the RMSE index by 4.12%; In the case of an 8-fold sparse angle, the BBC-Recon method increased the PSNR index by 1.43% and the SSIM index by 1.49%, and reduced the RMSE index by 0.77%. Conclusion The BBC-Recon algorithm fully exploits the correlation information between projections to allow effective removal of streak artifacts while preserving image structure information, and demonstrates significant advantages in maintaining inter-slice consistency.

Key words: sparse-view cone-beam CT, bidirectional multi-scale correlation, bidirectional flow field estimation, dual-domain training

Wenwei LI, Zerui MAO, Yongbo WANG, Zhaoying BIAN, Jing HUANG. A sparse-view cone-beam CT reconstruction algorithm based on bidirectional flow field- guided projection completion[J]. Journal of Southern Medical University, 2025, 45(2): 395-408.

Figures/Tables 19

References 36

1	Scarfe WC, Farman AG, Sukovic P. Clinical applications of cone-beam computed tomography in dental practice[J]. J Can Dent Assoc, 2006, 72(1): 75-80.
2	Brenner DJ, Hall EJ. Computed tomography: an increasing source of radiation exposure[J]. N Engl J Med, 2007, 357(22): 2277-84.
3	Berrington de González A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007[J]. Arch Intern Med, 2009, 169(22): 2071-7.
4	Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithm[J]. J Opt Soc Am A, 1984, 1(6): 612.
5	Andersen AH, Kak AC. Simultaneous algebraic reconstruction technique (SART): a superior implementation of the art algorithm[J]. Ultrason Imaging, 1984, 6(1): 81-94.
6	Sidky EY, Pan XC. Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization[J]. Phys Med Biol, 2008, 53(17): 4777-807.
7	Li S, Cao Q, Chen Y, et al. Dictionary learning based sinogram inpainting for CT sparse reconstruction[J]. Optik, 2014, 125(12): 2862-7.
8	Bertram M, Rose G, Schafer D, et al. Directional interpolation of sparsely sampled cone-beam CT sinogram data[C]//2004 2nd IEEE International Symposium on Biomedical Imaging: Nano to Macro. Arlington, VA, USA. IEEE, 2004: 928-31.
9	Zhang H, Sonke JJ. Directional sinogram interpolation for sparse angular acquisition in cone-beam computed tomography[J]. J Xray Sci Technol, 2013, 21(4): 481-96.
10	Muhlich M, Aach T. Analysis of multiple orientations[J]. IEEE Trans Image Process, 2009, 18(7): 1424-37.
11	Li ZB, Yu LF, Trzasko JD, et al. Adaptive nonlocal means filtering based on local noise level for CT denoising[J]. Med Phys, 2014, 41(1): 011908.
12	Litjens G, Kooi T, Bejnordi BE, et al. A survey on deep learning in medical image analysis[J]. Med Image Anal, 2017, 42: 60-88.
13	Razzak MI, Naz S, Zaib A. Deep learning for medical image processing: overview, challenges and future[EB/OL]. 2017: arXiv: 1704.06825.
14	JinKyong Hwan, McCann MT, Froustey E, et al. Deep convolutional neural network for inverse problems in imaging[J]. IEEE Trans Image Process, 2017, 26(9): 4509-22.
15	Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C]//Medical image computing and computer-assisted intervention-MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer International Publishing, 2015: 234-41.
16	Chen H, Zhang Y, Kalra MK, et al. Low-dose CT with a residual encoder-decoder convolutional neural network[J]. IEEE Trans Med Imaging, 2017, 36(12): 2524-35.
17	Zhong WH, Li TL, Hou SK, et al. Unsupervised disentanglement strategy for mitigating artifact in photoacoustic tomography under extremely sparse view[J]. Photoacoustics, 2024, 38: 100613.
18	Wu WW, Hu DL, Niu C, et al. DRONE: dual-domain residual-based optimization NEtwork for sparse-view CT reconstruction[J]. IEEE Trans Med Imaging, 2021, 40(11): 3002-14.
19	Wu WW, Guo XD, Chen Y, et al. Deep embedding-attention-refinement for sparse-view CT reconstruction[J]. IEEE Trans Instrum Meas, 2023, 72: 4501111.
20	Lahiri A, Maliakal G, Klasky ML, et al. Sparse-view cone beam CT reconstruction using data-consistent supervised and adversarial learning from scarce training data[J]. IEEE Trans Comput Imag, 2023, 9: 13-28.
21	Chao LY, Wang ZW, Zhang HB, et al. Sparse-view cone beam CT reconstruction using dual CNNs in projection domain and image domain[J]. Neurocomputing, 2022, 493: 536-47.
22	Kim S, Kim B, Lee J, et al. Sparsier2Sparse: self-supervised convolutional neural network-based streak artifacts reduction in sparse-view CT images[J]. Med Phys, 2023, 50(12): 7731-47.
23	Pan JY, Zhang HY, Wu WF, et al. Multi-domain integrative Swin transformer network for sparse-view tomographic reconstruction[J]. Patterns, 2022, 3(6): 100498.
24	Genzel M, Gühring I, MacDonald J, et al. Near-exact recovery for tomographic inverse problems via deep learning[J]. Inter Confer Machine Learn, 2022，2022: 7368-81.
25	Mildenhall B, Srinivasan PP, Tancik M, et al. NeRF: representing scenes as neural radiance fields for view synthesis[C]//European Conference on Computer Vision. Cham: Springer, 2020: 405-421.
26	Chung H, Ryu D, McCann MT, et al. Solving 3D inverse problems using pre-trained 2D diffusion models[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC, Canada. IEEE, 2023: 22542-51.
27	Li ZR, Wang YY, Zhang JJ, et al. Two-and-a-half order score-based model for solving 3D ill-posed inverse problems[J]. Comput Biol Med, 2024, 168: 107819.
28	Liu ZT, Fang Y, Li CJ, et al. Geometry-aware attenuation learning for sparse-view CBCT reconstruction[EB/OL]. 2023: arXiv: 2303. 14739.
29	Cai YH, Wang JH, Yuille A, et al. Structure-aware sparse-view X-ray 3D reconstruction[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA. IEEE, 2024: 11174-83.
30	Liu ZT, Zhao HX, Qin WH, et al. 3D vessel reconstruction from sparse-view dynamic DSA images via vessel probability guided attenuation learning[EB/OL]. 2024: arXiv: 2405.10705.
31	Teed Z, Deng J. RAFT: recurrent all-pairs field transforms for optical flow[C]//European Conference on Computer Vision. Cham: Springer, 2020: 402-419.
32	Kong LT, Jiang BY, Luo DH, et al. IFRNet: intermediate feature refine network for efficient frame interpolation[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA. IEEE, 2022: 1959-68.
33	Li Z, Zhu ZL, Han LH, et al. AMT: all-pairs multi-field transforms for efficient frame interpolation[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC, Canada. IEEE, 2023: 9801-10.
34	Charbonnier P, Blanc-Feraud L, Aubert G, et al. Two deterministic half-quadratic regularization algorithms for computed imaging[C]//Proceedings of 1st International Conference on Image Processing. Austin, TX, USA. IEEE, 1994: 168-72.
35	Moen TR, Chen BY, Holmes DR 3rd, et al. Low-dose CT image and projection dataset[J]. Med Phys, 2021, 48(2): 902-11.
36	Huang YY, Liu WJ, Yao CQ, et al. A multimodal dental dataset facilitating machine learning research and clinic services[J]. Sci Data, 2024, 11(1): 1291.

Method	Axial			Coronal			Sagittal
Method	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE
FDK	30.9635	0.9677	84.0423	31.2190	0.7775	82.9474	31.1780	0.9634	83.1501
ASD-POCS	37.5809	0.9928	39.3558	38.2010	0.8781	38.1144	37.7018	0.9902	39.0479
FBPConvNet	38.7494	0.9946	35.8485	39.5930	0.9156	34.5152	39.0173	0.9936	35.4066
SAX-NeRF	36.2327	0.9902	45.9355	36.8490	0.8478	44.5609	36.4561	0.9880	45.4288
Dual-CNN	35.5758	0.9885	50.2866	36.4565	0.8518	48.2645	35.8967	0.9866	49.4623
TOSM	37.8617	0.9933	38.1402	38.4367	0.8666	36.9161	38.0829	0.9919	37.6555
BBC-Recon	39.4176	0.9953	33.3770	40.2626	0.9252	32.1349	39.7163	0.9945	32.9198

Method	Axial			Coronal			Sagittal
Method	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE
FDK	30.9635	0.9677	84.0423	31.2190	0.7775	82.9474	31.1780	0.9634	83.1501
ASD-POCS	37.5809	0.9928	39.3558	38.2010	0.8781	38.1144	37.7018	0.9902	39.0479
FBPConvNet	38.7494	0.9946	35.8485	39.5930	0.9156	34.5152	39.0173	0.9936	35.4066
SAX-NeRF	36.2327	0.9902	45.9355	36.8490	0.8478	44.5609	36.4561	0.9880	45.4288
Dual-CNN	35.5758	0.9885	50.2866	36.4565	0.8518	48.2645	35.8967	0.9866	49.4623
TOSM	37.8617	0.9933	38.1402	38.4367	0.8666	36.9161	38.0829	0.9919	37.6555
BBC-Recon	39.4176	0.9953	33.3770	40.2626	0.9252	32.1349	39.7163	0.9945	32.9198

Method	Axial			Coronal			Sagittal
Method	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE
FDK	26.0442	0.9056	148.0213	26.3768	0.6830	145.5730	26.3245	0.8951	145.9907
ASD-POCS	34.9296	0.9866	53.3403	35.9176	0.8432	50.7807	35.2250	0.9839	52.4890
FBPConvNet	36.2275	0.9903	46.9397	37.3632	0.8807	44.4852	36.5693	0.9885	46.1690
SAX-NeRF	34.7707	0.9862	54.3179	35.5614	0.8335	52.2195	35.0749	0.9833	53.5090
Dual-CNN	32.9485	0.9807	67.4798	34.3106	0.8104	64.2221	33.4249	0.9775	66.2729
TOSM	37.0262	0.9918	41.9589	37.7133	0.8602	40.3936	37.2900	0.9903	41.3459
BBC-Recon	37.2343	0.9923	42.0751	38.3114	0.8956	40.0071	37.6296	0.9910	41.2971

Method	Axial			Coronal			Sagittal
Method	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE
FDK	26.0442	0.9056	148.0213	26.3768	0.6830	145.5730	26.3245	0.8951	145.9907
ASD-POCS	34.9296	0.9866	53.3403	35.9176	0.8432	50.7807	35.2250	0.9839	52.4890
FBPConvNet	36.2275	0.9903	46.9397	37.3632	0.8807	44.4852	36.5693	0.9885	46.1690
SAX-NeRF	34.7707	0.9862	54.3179	35.5614	0.8335	52.2195	35.0749	0.9833	53.5090
Dual-CNN	32.9485	0.9807	67.4798	34.3106	0.8104	64.2221	33.4249	0.9775	66.2729
TOSM	37.0262	0.9918	41.9589	37.7133	0.8602	40.3936	37.2900	0.9903	41.3459
BBC-Recon	37.2343	0.9923	42.0751	38.3114	0.8956	40.0071	37.6296	0.9910	41.2971

Method	Axial			Coronal			Sagittal
Method	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE	PSNR	SSIM	RMSE
FDK	31.6740	0.9933	40.2281	32.2543	0.7547	38.8053	31.9684	0.9102	39.6143
ASD-POCS	39.2508	0.9988	17.2538	39.9507	0.8566	16.5868	39.6122	0.9492	16.8998
FBPConvNet	40.2464	0.9992	17.6568	42.7248	0.9533	16.1393	41.2941	0.9796	16.9073
SAX-NeRF	35.9638	0.9975	24.7276	36.9414	0.8466	23.5581	36.3126	0.9361	24.2745
Dual-CNN	36.3912	0.9978	24.0718	37.2046	0.8321	23.4742	36.8023	0.9328	23.7924
TOSM	38.8606	0.9987	17.9193	39.7214	0.8671	17.1003	39.2508	0.9471	17.5033
BBC-Recon	41.0348	0.9993	16.4155	43.9115	0.9694	14.7886	42.3120	0.9883	15.6204