Organizations: University of Central Florida · The Hong Kong University of Science and Technology (Guangzhou) · Tampere University · Ellis Institute Finland
Sparse-view cone-beam computed tomography (CBCT) is severely under-constrained by limited X-ray projections. Learned reconstruction can provide complementary anatomical information, but using it as a prior in patient-specific Gaussian reconstruction introduces two challenges. First, the discrepancy between the learned prior and current reconstruction is spatially non-uniform: some local differences contain useful missing 3D information, whereas others reflect prior bias. Second, even a volumetrically beneficial correction can conflict with the observed X-ray measurements. Since X-ray measurements are line integrals, spatially separated corrections may affect overlapping rays, with their signed projection responses reinforcing or partially canceling in the measurement residual, so their feasibility cannot always be assessed independently. In this paper, we introduce RePrior, a multi-stage progressive prior-correction framework that addresses correction actionability, measurement feasibility, and representation expressivity. Selective Correction Proposal (SCP) first identifies actionable signed local corrections from prior--reconstruction discrepancies. Measurement-Constrained Reconciliation (MCR) then jointly enforces their consistency with patient-specific X-ray measurements, while Hybrid Residual Enhancement (HRE) recovers additional correction capacity through complementary voxel- and Gaussian-domain degrees of freedom around the reconciled anchor. RePrior requires neither test-time ground truth nor case-specific network fine-tuning. RePrior gains approximately 2.6~dB in 3D PSNR over R2-Gaussian across LUNA16, PANORAMA, and PENGWIN. Beyond prior-assisted reconstruction, it delivers additional gains in volumetric quality and observed-view projection fidelity, with consistent gains in held-out-view fidelity in the backbone-matched evaluation.
Figures & tables
Figure 1: We introduce RePrior , a multi-stage progressive prior-correction framework which transfers learned priors into patient-specific sparse-view CBCT reconstruction through selective correction proposal, measurement-constrained reconciliation, and hybrid residual enhancement.
Figure 2: Challenges in prior-guided correction. Left: local prior differences contain both useful 3D corrections and prior bias. Right: some volumetrically beneficial corrections still conflict with X-ray measurements.
Figure 3: Overview of RePrior. (a) Learned prior VP and prior-assisted reconstruction VR . (b) SCP predicts a signed correction DA . (c) MCR jointly reconciles structured correction modes using their X-ray responses H , producing DB and the feasible anchor VB . (d) HRE combines voxel- and Gaussian-domain residual freedom under measurement guards to obtain V∗ .
Figure 4: Ray-coupled correction responses on a representative case, illustrating measurement conflict and signed compensation.
LUNA16 (90)
PANORAMA (90)
PENGWIN (50)
Method
PSNR ↑
SSIM ↑
PSNR ↑
SSIM ↑
PSNR ↑
SSIM ↑
SAX-NeRF (CVPR’24)
29.41
0.8311
27.57
0.8358
30.86
0.8726
X-Gaussian (ECCV’24)
19.55
0.3030
19.24
0.3546
20.41
0.3468
R 2 -GS (NeurIPS’24)
29.12
0.8357
28.42
0.8534
30.29
0.8817
Exact-GS (CVPR’26)
26.71
0.7598
26.51
0.7711
28.46
0.8134
DeepSparse-10 (TMI’26)
30.74
0.8660
26.35
0.8386
27.12
0.8461
Table 1: Main comparison on sparse-view CBCT reconstruction. Best and second-best results are shown in bold and underlined, respectively.
Figure 5: Qualitative comparison on representative LUNA16, PANORAMA, and PENGWIN cases. Red and cyan boxes mark the highlighted local regions.
Backbone
Cases
Method
3D PSNR ↑
3D SSIM ↑
2D PSNR ↑
2D SSIM ↑
R 2 -GS
90
R 2 -GS
29.12
0.8357
43.17
0.9864
90
Prior-R 2 -GS
32.09
0.8826
44.09
0.9906
90
RePrior
32.58
0.8881
44.24
0.9910
Exact-GS
30
Exact-GS
27.25
0.7749
40.60
0.9754
30
Prior-Exact-GS
31.27
0.8688
44.22
0.9893
30
RePrior
31.87
0.8754
44.35
0.9898
Table 2: Generalization across Gaussian reconstruction backbones on LUNA16.
Configuration
3D PSNR ↑
3D SSIM ↑
2D PSNR ↑
2D SSIM ↑
Prior-R 2 -GS
32.03
0.8966
46.46
0.9956
+ SCP
32.58
0.9033
46.30
0.9952
+ MCR
32.24
0.8989
46.63
0.9957
+ HRE (Full)
32.43
0.9012
47.10
0.9956
Table 3: Stage-wise ablation of RePrior.
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
Feature family
Dim.
Main quantities
Multi-scale frequency
18
Five band-energy ratios, spectral centroid, high/low ratio, mid/total ratio, and total energy for VP and VR
Differential morphology
38
Gradient magnitude, surface normal, Hessian eigenstructure, structure-tensor descriptors, local variance, contrast, Laplacian, and boundary distance
Local attenuation
2
Local attenuation values of VP and VR
Direct disagreement
5
Absolute and relative attenuation discrepancy, gradient difference, normal disagreement, and local NCC
Discrepancy spectrum
7
Five discrepancy-band ratios, spectral centroid, and total discrepancy energy
Structural-anchor relation
5
Structural-feature distance and frozen AGREE/DISAGREE/P1-ONLY/P2-ONLY indicators
Appendix
Table 4: Composition of the 75 -dimensional SCP discrepancy descriptor. All features are available at test time; ground truth is used only to build the training targets in Eqs. ( 21 ) and ( 22 ).
Quantity
Constraint
Position
∥Δxi∥2≤0.2 normalized scene units
Attenuation / density
log(ρi/ρiref)≤0.1
Per-axis scale
log(si,a/si,aref)≤0.1
Physical scale
0.001≤si,a≤1
Rotation
geodesic change ≤10∘
Topology
fixed; no densification or pruning
Appendix
Table 5: Gaussian trust bounds used by HRE.
Figure S1: Additional qualitative results on LUNA16 (Part I). Each row shows an additional representative slice, while columns compare SAX-NeRF, X-Gaussian, R 2 -GS, Exact-GS, DeepSparse-10, DeepSparse-25, RePrior, and GT.
Figure S2: Additional qualitative results on LUNA16 (Part II). Additional slices and viewing planes using the same method ordering as in Fig. S1 .
Figure S3: Additional qualitative results on PANORAMA (Part I). Each row shows an additional representative slice, while columns compare SAX-NeRF, X-Gaussian, R 2 -GS, Exact-GS, DeepSparse-10, DeepSparse-25, RePrior, and GT.
Figure S4: Additional qualitative results on PANORAMA (Part II). Additional slices and viewing planes using the same method ordering as in Fig. S3 .
Figure S5: Additional qualitative results on PENGWIN (Part I). Each row shows an additional representative slice, while columns compare SAX-NeRF, X-Gaussian, R 2 -GS, Exact-GS, DeepSparse-10, DeepSparse-25, RePrior, and GT.
Figure S6: Additional qualitative results on PENGWIN (Part II). Additional slices and viewing planes using the same method ordering as in Fig. S5 .