Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Tomographic Foundation Model -- FORCE: Flow-Oriented Reconstruction Conditioning Engine

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read One Poisson-flow prior reconstructs low-dose, sparse-view, and metal CT.

desk verdict A credible unsupervised CT reconstruction framework with modest but real gains, undercut by an under-specified starting-time heuristic and missing quantitative support. read the letter →

arxiv 2506.02149 v1 pith:6MEWX6UF submitted 2025-06-02 eess.IV cs.LGeess.SP

Wenjun Xia , Chuang Niu , Ge Wang This is my paper · ORCID
classification eess.IVcs.LGeess.SP
keywords CTreconstructionPoissonflowgenerativemodelPFGM++unsupervisedlearninglow-dosesparse-viewmetalartifactreductiondatafidelityconditioning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

FORCE is a CT reconstruction framework built on the Poisson flow generative model PFGM++. The paper's central claim is that a single learned prior, combined with task-specific data-fidelity conditioning, can reconstruct clinically useful CT images from low-dose, sparse-view, and metal-corrupted measurements without paired training data. On benchmark tests, FORCE reaches 37.32 dB PSNR on 25% dose CT and 39.23 dB PSNR with 0.947 SSIM on 96-view sparse-view CT, beating all compared unsupervised methods. If correct, this would give radiology a protocol-agnostic reconstruction engine that avoids the paired-data bottleneck and hallucination risk of supervised deep learning.

What carries the argument

The central object is the PFGM++ ordinary differential equation in an augmented data space, where data samples move as charged particles along Poisson-field force lines toward a spherical prior at infinity; reverse integration returns to the data manifold. Training uses an EDM-style denoising objective with hyperspherical noise directions. The key mechanism is the conditioning wrapper in Algorithm 2: at each flow step, a task-specific operator (RED, OS-SART, or metal-trace sinogram substitution) enforces the projection constraint $Hx=p$, and a momentum-corrected, TV-regularized denoised iterate supplies the flow direction. The Phase II starting-time heuristic — choosing a mid-noise point from the FID-alignment curve — is what distinguishes FORCE from prior diffusion-based reconstruction and is the key to masking artifacts without losing anatomy.

What would settle it

Measure the Fig. 3 FID curve for paired sparse-view and metal-corrupted images and run FORCE with the low-dose-derived starting time; if no alignment phase exists for those tasks, or if the optimal starting time is far from Phase II, the reported gains will not reproduce.

Watch

Extended reading notes

Core claim

FORCE treats CT reconstruction as conditional sampling from a learned prior rather than as a supervised image-to-image mapping. The prior is a PFGM++ model trained on clean CT slices; at inference, the reverse Poisson-flow ODE is interleaved with a data-fidelity conditioning step — RED for low-dose, OS-SART for sparse-view, and metal-trace sinogram substitution for metal artifacts — together with total-variation regularization and momentum applied to the denoised intermediate. The paper's key discovery is the starting-time heuristic: initializing the reverse flow at a mid-noise level taken from the alignment phase of an FID curve between paired low-dose and normal-dose images lets the prior suppress artifact patterns while retaining enough anatomical structure to guide reconstruction. The claim is that this recipe transfers across three artifact types and yields the best unsupervised results among the compared methods on the low-dose, sparse-view, and CT-MAR benchmarks.

Load-bearing premise

The method assumes that a mid-noise starting point selected from the FID-alignment curve of paired low-dose and normal-dose images will also mask artifacts and preserve anatomy for sparse-view and metal-corrupted data, where the artifact structure is different.

Editorial extensions

If this is right

  • On the 25% dose benchmark, FORCE reconstructs low-dose CT at 37.32 dB PSNR and 48.58 FID, improving over CycleGAN, Noise2Sim, and BM3D.
  • On the 96-view sparse-view benchmark, FORCE reconstructs CT at 39.23 dB PSNR, 0.947 SSIM, and 48.09 FID, outperforming FBP, BM3D, CycleGAN, and the sparsified sparse-PFGM baseline.
  • Because the prior is trained on clean images only, the same PFGM++ model can serve multiple tasks simultaneously by swapping only the conditioning operator.
  • The Phase II starting-time result implies that any learned generative prior can be steered toward reconstruction by selecting an appropriate mid-noise initialization rather than re-training the model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the Phase II heuristic holds beyond the three tested tasks, a practical protocol would be to measure the FID-alignment curve once per artifact class and use its alignment phase to set the starting noise for any new scanner or protocol.
  • The same conditioning framework could extend to other inverse problems in imaging where paired clean and corrupted data are scarce, such as MRI, PET, or deblurring, whenever a clean-image prior and a projection operator are available.
  • A testable follow-up would be to compare FORCE against supervised and self-supervised networks on identical paired training sets, to quantify the accuracy-versus-hallucination trade-off that motivates the unsupervised design.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes FORCE, a CT reconstruction framework that combines a pretrained PFGM++ generative prior with task-specific data-fidelity conditioning during reverse-ODE sampling. Three conditioning strategies are used: a quadratic proximity term labeled RED for low-dose CT, OS-SART for sparse-view CT, and sinogram substitution for metal artifact reduction (MAR). TV regularization and momentum are incorporated into the sampling updates, and the starting noise level is chosen using an empirical FID curve between paired low-dose and normal-dose images (Fig. 3). Experiments on the AAPM-Mayo and CT-MAR datasets report PSNR/SSIM/FID values for low-dose and sparse-view CT, plus a qualitative MAR example, claiming superiority over unsupervised baselines.

Significance. If the reported results are reproducible, the paper makes a valuable contribution: a single flow-based generative prior, trained without paired supervision, can serve multiple CT reconstruction tasks through task-specific conditioning. The use of PFGM++ is well motivated by its sampling efficiency and stability, and the idea of selecting the start of the reverse trajectory from an FID alignment phase is interesting. However, the current evidence is insufficient to support the central claim: key hyperparameters are missing, the starting-time heuristic is under-specified, the MAR evaluation is purely qualitative, and no statistical analysis is provided. The manuscript would be strengthened substantially by adding reproducibility details, a sensitivity analysis, and quantitative MAR results.

major comments (4)
  1. [Sec. II-C-c and Fig. 3] The starting sampling noise is chosen from the 'Phase II' region of an empirical FID curve between paired low-dose and normal-dose CT images, but the manuscript does not specify how the phase boundaries are determined from the FID curve, does not state whether the paired images used to construct Fig. 3 are disjoint from the 455 test slices used in Tables I and II, and does not report whether the same sigma_0 is used for all three tasks. If the FID curve was built on test pairs, the reported PSNR and FID numbers are selection results rather than independent evaluations. Since the entire reverse trajectory and the final reconstruction depend on this choice, the paper should provide the exact selection rule, the dataset split, and a sensitivity analysis over sigma_0.
  2. [Sec. III, Tables I and II] The experiments do not report essential hyperparameters for FORCE: the TV weight lambda, the RED conditioning weight eta, the PFGM++ augmented dimension D, the number of sampling steps T, the time schedule {t_i}, and the starting time t_0 (or sigma_0). Without these values, the results in Tables I and II cannot be reproduced. Additionally, no error bars or statistical significance tests (e.g., over test slices) are provided, so the reported improvements over CycleGAN and Noise2Sim cannot be assessed for significance.
  3. [Sec. III-C and Fig. 6] The metal artifact reduction study is presented with a single qualitative image and no quantitative results. No MAR table, no numerical comparison against LI or other baselines, and no description of the number of test cases or the metric definitions are given. Consequently, the abstract claim of 'superior performance' in MAR is not supported by the evidence in the manuscript.
  4. [Sec. II-C-a, Eq. (10)] The regularizer labeled RED in Eq. (10) is not the RED (Regularization by Denoising) regularizer. RED is typically defined as (1/2) x^T (x - f(x)) where f is a denoising operator, whereas Eq. (10) uses x_i, the current iterate, without a denoiser. This is not a merely terminological issue: the convergence properties and the interpretation as RED would not apply to the stated objective. The authors should either correct the formula to use a proper denoiser or rename the method and justify the proximity form used.
minor comments (5)
  1. [Algorithm 2] The momentum coefficient update uses xi_i = (1 + sqrt(1 + 4 xi_{i-1}^2))/2, but the loop begins with i=0 and xi_0 is initialized to 1; xi_{-1} is undefined. Please clarify the indexing, likely by initializing xi_{-1} or by shifting indices.
  2. [Section I] There is a typo in the first paragraph: 'Flow-Oriented Reconstruction Conditioning Engin' should be 'Engine'.
  3. [Section III, Fig. 6] In the caption of Fig. 6, 'The display windows is [-160 240] HU' should be 'The display window is [-160, 240] HU.'
  4. [Section II-C-b] The word 'signficantly' is misspelled; it should be 'significantly.'
  5. [Eq. (3)] The notation S^{N+D-1}(1) is unclear; it should be defined explicitly as the surface area of the unit sphere (or hemisphere) in the augmented space.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: FORCE is an independent PFGM++ prior plus standard data-fidelity conditioning; the FID-based start-time heuristic is a calibration/disclosure risk, not a by-construction reduction.

full rationale

FORCE's reconstruction chain is not circular. The learned prior f_theta is trained with the standard PFGM++ denoising objective (Eq. 7) on normal-dose CT images only; no paired data enter the training loss. The data-fidelity conditioning uses external methods (RED, OS-SART, sinogram substitution) that are standard and are not defined in terms of the target reconstruction. The one empirically calibrated element—the starting noise level sigma_0 from the FID curve in Fig. 3 (Sec. II-C-c)—is a hyperparameter heuristic; it does not define the PSNR/FID metrics used for evaluation nor does it mathematically force the output, so it is a validation/leakage concern rather than a circular derivation. Self-citations to the authors' prior diffusion work [76]-[78] provide context and inherit the conditioning idea, but the load-bearing generative model is externally cited PFGM++ [79,80] and the conditioning algorithms are external. Under the rule that circularity requires a demonstrated reduction by construction, no such reduction is present.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The method is a combination of established generative modeling and conditioning components; no new physical entities are introduced. The main unstated inputs are hyperparameters (t0, lambda, eta, D, T) and the empirical FID heuristic. The axioms are mostly standard math and domain assumptions about the validity of conditioning and regularization during sampling.

free parameters (5)
  • Starting sampling noise sigma_0 (Phase II time t_0) = not reported
    Selected from the FID curve between paired low-dose and normal-dose images (Sec. II-C-c, Fig. 3); task-dependent and not disclosed numerically.
  • TV regularization weight lambda = not reported
    Used in the TV-regularized denoising step in Algorithm 2; no value or selection procedure given.
  • RED conditioning weight eta = not reported
    Used in Eq. (10) for low-dose conditioning; no value or selection procedure given.
  • PFGM++ augmented dimension D = not reported
    Dimension of the augmented space in PFGM++ training and sampling; not specified in experiments.
  • Number of sampling steps T and time schedule {t_i} = not reported
    Sampling schedule in Algorithm 2 is not defined with concrete values.
assumptions (6)
  • standard math PFGM++ forward ODE and Poisson field equations (Eqs. 2-3) define a valid generative process.
    Taken from Xu et al. [79,80], used without re-derivation.
  • standard math EDM-style denoising objective (Eq. 5) and its equivalence to PFGM++ training with r = sigma sqrt(D) (Eq. 7).
    Adopted from Karras et al. [82] and Xu et al. [80]; no derivation in this paper.
  • standard math The MAP/Bayesian formulation (Eqs. 8-9) links data fidelity and regularizer to the posterior.
    Standard MBIR background, cited to Bayesian CT literature.
  • domain assumption Conditioning with RED, OS-SART, or sinogram substitution during PFGM++ sampling yields samples from the posterior p(x|p).
    Assumed in Sec. II-C-a; the paper provides no theoretical guarantee that these conditioning steps preserve the target posterior.
  • ad hoc to paper FID alignment phases (Fig. 3) imply Phase II is the optimal starting region for all three tasks.
    Empirical heuristic introduced in Sec. II-C-c; no rigorous criterion for choosing the phase boundary.
  • domain assumption TV regularization and Nesterov momentum on the denoised estimate improve reconstruction without biasing the final result.
    Proposed in Sec. II-C-b; no ablation or theoretical support shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tomographic Foundation Model -- FORCE: Flow-Oriented Reconstruction Conditioning Engine." pith.science (2026). https://pith.science/paper/6MEWX6UF

@misc{pith2026250602149,
  author       = {Pith},
  title        = {Pith review of: Tomographic Foundation Model -- FORCE: Flow-Oriented Reconstruction Conditioning Engine},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6MEWX6UF}},
  note         = {Machine review of arXiv:2506.02149}
}
read the original abstract

Computed tomography (CT) is a major medical imaging modality. Clinical CT scenarios, such as low-dose screening, sparse-view scanning, and metal implants, often lead to severe noise and artifacts in reconstructed images, requiring improved reconstruction techniques. The introduction of deep learning has significantly advanced CT image reconstruction. However, obtaining paired training data remains rather challenging due to patient motion and other constraints. Although deep learning methods can still perform well with approximately paired data, they inherently carry the risk of hallucination due to data inconsistencies and model instability. In this paper, we integrate the data fidelity with the state-of-the-art generative AI model, referred to as the Poisson flow generative model (PFGM) with a generalized version PFGM++, and propose a novel CT framework: Flow-Oriented Reconstruction Conditioning Engine (FORCE). In our experiments, the proposed method shows superior performance in various CT imaging tasks, outperforming existing unsupervised reconstruction approaches.

Figures

Figures reproduced from arXiv: 2506.02149 by the authors.

Figure 1
Figure 1. Conceptual comparison between diffusion and Poisson flow models. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Conditioned sampling trajectories. The red lines illustrate the reverse [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FID between paired low-dose and normal-dose CT images under [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Pelvis reconstructions with different methods from 25% dose data. The display window is [-160, 240] HU. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Abdominal reconstructions with different methods from 96 projection views. The display window is [-160, 240] HU. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Dental reconstruction with different methods from metal corrupted data. The display windows is [-160 240] HU. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. FORCE-Interior: A Poisson Flow Generative Prior for Interior Tomography Reconstruction

    eess.IV 2026-07 conditional novelty 5.0 of 10

    A Poisson-flow generative prior, initialized with a full-FOV OS-SART reconstruction and re-conditioned on truncated projections each step, improves interior-tomography ROI reconstruction quality at small ROI radii.

Reference graph

Works this paper leans on

93 extracted references · 61 canonical work pages · cited by 1 Pith paper

  1. [1]

    Projected lifetime cancer risks from current computed tomography imaging,

    R. Smith-Bindman, P. W. Chu, H. A. Firdaus, C. Stewart, M. Malekhe- dayat, S. Alber, W. E. Bolch, M. Mahendra, A. B. de Gonz ´alez, and D. L. Miglioretti, “Projected lifetime cancer risks from current computed tomography imaging,”JAMA Internal Medicine, 2025

    2025

  2. [2]

    A. C. Kak and M. Slaney,Principles of computerized tomographic imaging. SIAM, 2001

    2001

  3. [3]

    Algebraic reconstruction techniques (ART) for three-dimensional electron microscopy and X-ray photography,

    R. Gordon, R. Bender, and G. T. Herman, “Algebraic reconstruction techniques (ART) for three-dimensional electron microscopy and X-ray photography,”Journal of Theoretical Biology, vol. 29, no. 3, pp. 471– 481, 1970. 8

    1970

  4. [4]

    Simultaneous iterative reconstruction technique: Physical interpretation based on the generalized least squares solution,

    J. Trampert and J.-J. Leveque, “Simultaneous iterative reconstruction technique: Physical interpretation based on the generalized least squares solution,”Journal of Geophysical Research: Solid Earth, vol. 95, no. B8, pp. 12 553–12 559, 1990

    1990

  5. [5]

    Simultaneous algebraic reconstruction technique (SART): a superior implementation of the ART algorithm,

    A. H. Andersen and A. C. Kak, “Simultaneous algebraic reconstruction technique (SART): a superior implementation of the ART algorithm,” Ultrasonic Imaging, vol. 6, no. 1, pp. 81–94, 1984

    1984

  6. [6]

    Maximum likelihood from incomplete data via the EM algorithm,

    A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via the EM algorithm,”Journal of the royal statistical society: series B (methodological), vol. 39, no. 1, pp. 1–22, 1977

    1977

  7. [7]

    Ray contribution masks for structure adaptive sinogram filtering,

    M. Balda, J. Hornegger, and B. Heismann, “Ray contribution masks for structure adaptive sinogram filtering,”IEEE Transactions on Medical Imaging, vol. 31, no. 6, pp. 1228–1239, 2012

    2012

  8. [8]

    Projection space denoising with bilateral filtering and CT noise modeling for dose reduction in CT,

    A. Manduca, L. Yu, J. D. Trzasko, N. Khaylova, J. M. Kofler, C. M. McCollough, and J. G. Fletcher, “Projection space denoising with bilateral filtering and CT noise modeling for dose reduction in CT,” Medical Physics, vol. 36, no. 11, pp. 4911–4919, 2009

    2009

Show all 93 references
  1. [9]

    Image reconstruction from projections: Iii. projection completion methods (theory)

    R. Lewitt and B. RHT, “Image reconstruction from projections: Iii. projection completion methods (theory).” 1978

    1978

  2. [10]

    Reduction of ct artifacts caused by metallic implants

    W. A. Kalender, R. Hebel, and J. Ebersberger, “Reduction of ct artifacts caused by metallic implants.”Radiology, vol. 164, no. 2, pp. 576–577, 1987

    1987

  3. [11]

    Normalized metal artifact reduction (nmar) in computed tomography,

    E. Meyer, R. Raupach, M. Lell, B. Schmidt, and M. Kachelrieß, “Normalized metal artifact reduction (nmar) in computed tomography,” Medical physics, vol. 37, no. 10, pp. 5482–5493, 2010

    2010

  4. [12]

    Penalized weighted least-squares approach to sinogram noise reduction and image reconstruction for low-dose x-ray computed tomography,

    J. Wang, T. Li, H. Lu, and Z. Liang, “Penalized weighted least-squares approach to sinogram noise reduction and image reconstruction for low-dose x-ray computed tomography,”IEEE transactions on medical imaging, vol. 25, no. 10, pp. 1272–1283, 2006

    2006

  5. [13]

    A sinogram denoising algorithm for low-dose computed tomography,

    D. Karimi, P. Deman, R. Ward, and N. Ford, “A sinogram denoising algorithm for low-dose computed tomography,”BMC medical imaging, vol. 16, pp. 1–14, 2016

    2016

  6. [14]

    Metal artifact reduction in ct images by sinogram tv inpainting,

    X. Duan, L. Zhang, Y . Xiao, J. Cheng, Z. Chen, and Y . Xing, “Metal artifact reduction in ct images by sinogram tv inpainting,” in2008 IEEE Nuclear Science Symposium Conference Record. IEEE, 2008, pp. 4175– 4177

    2008

  7. [15]

    Metal artifact reduction in dual energy ct by sinogram segmentation based on active contour model and tv inpainting,

    H. Xue, L. Zhang, Y . Xiao, Z. Chen, and Y . Xing, “Metal artifact reduction in dual energy ct by sinogram segmentation based on active contour model and tv inpainting,” in2009 IEEE Nuclear Science Symposium Conference Record (NSS/MIC). IEEE, 2009, pp. 904–908

    2009

  8. [16]

    A new ct metal artifacts reduction algorithm based on fractional-order sinogram inpainting,

    Y . Zhang, Y .-F. Pu, J.-R. Hu, Y . Liu, and J.-L. Zhou, “A new ct metal artifacts reduction algorithm based on fractional-order sinogram inpainting,”Journal of X-ray science and technology, vol. 19, no. 3, pp. 373–384, 2011

    2011

  9. [17]

    Total variation based iterative image reconstruction,

    G. Yu, L. Li, J. Gu, and L. Zhang, “Total variation based iterative image reconstruction,” inInternational Workshop on Computer Vision for Biomedical Image Applications. Springer, 2005, pp. 526–534

    2005

  10. [18]

    Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization,

    E. Y . Sidky and X. Pan, “Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization,” Physics in Medicine & Biology, vol. 53, no. 17, p. 4777, 2008

    2008

  11. [19]

    Few-view image reconstruction combining total variation and a high- order norm,

    Y . Zhang, W.-H. Zhang, H. Chen, M.-L. Yang, T.-Y . Li, and J.-L. Zhou, “Few-view image reconstruction combining total variation and a high- order norm,”International Journal of Imaging Systems and Technology, vol. 23, no. 3, pp. 249–255, 2013

    2013

  12. [20]

    Sparse-view X-ray CT reconstruction via total generalized variation regularization,

    S. Niu, Y . Gao, Z. Bian, J. Huang, W. Chen, G. Yu, Z. Liang, and J. Ma, “Sparse-view X-ray CT reconstruction via total generalized variation regularization,”Physics in Medicine & Biology, vol. 59, no. 12, pp. 2997–3017, 2014

    2014

  13. [21]

    Few-view image reconstruc- tion with fractional-order total variation,

    Y . Zhang, W. Zhang, Y . Lei, and J. Zhou, “Few-view image reconstruc- tion with fractional-order total variation,”Journal of the Optical Society of America A, vol. 31, no. 5, pp. 981–995, 2014

    2014

  14. [22]

    Low-dose spectral CT reconstruction using image gradientℓ 0–norm and tensor dictionary,

    W. Wu, Y . Zhang, Q. Wang, F. Liu, P. Chen, and H. Yu, “Low-dose spectral CT reconstruction using image gradientℓ 0–norm and tensor dictionary,”Applied Mathematical Modelling, vol. 63, pp. 538–557, 2018

    2018

  15. [23]

    A high-quality photon- counting CT technique based on weight adaptive total-variation and image-spectral tensor factorization for small animals imaging,

    W. Wu, D. Hu, K. An, S. Wang, and F. Luo, “A high-quality photon- counting CT technique based on weight adaptive total-variation and image-spectral tensor factorization for small animals imaging,”IEEE Transactions on Instrumentation and Measurement, vol. 70, pp. 1–14, 2020

    2020

  16. [24]

    Low-dose X-ray CT reconstruction via dictionary learning,

    Q. Xu, H. Yu, X. Mou, L. Zhang, J. Hsieh, and G. Wang, “Low-dose X-ray CT reconstruction via dictionary learning,”IEEE Transactions on Medical Imaging, vol. 31, no. 9, pp. 1682–1697, 2012

    2012

  17. [25]

    Sparse-view spectral ct reconstruction using spectral patch- based low-rank penalty,

    K. Kim, J. C. Ye, W. Worstell, J. Ouyang, Y . Rakvongthai, G. El Fakhri, and Q. Li, “Sparse-view spectral ct reconstruction using spectral patch- based low-rank penalty,”IEEE transactions on medical imaging, vol. 34, no. 3, pp. 748–760, 2014

    2014

  18. [26]

    Convolutional sparse coding for compressed sensing CT reconstruction,

    P. Bao, W. Xia, K. Yang, W. Chen, M. Chen, Y . Xi, S. Niu, J. Zhou, H. Zhang, H. Sunet al., “Convolutional sparse coding for compressed sensing CT reconstruction,”IEEE Transactions on Medical Imaging, vol. 38, no. 11, pp. 2607–2619, 2019

    2019

  19. [27]

    Spectral CT reconstruction with image sparsity and spectral mean,

    Y . Zhang, Y . Xi, Q. Yang, W. Cong, J. Zhou, and G. Wang, “Spectral CT reconstruction with image sparsity and spectral mean,”IEEE Trans- actions on Computational Imaging, vol. 2, no. 4, pp. 510–523, 2016

    2016

  20. [28]

    Sparse-view X- ray CT reconstruction usingℓ 1 regularization with learned sparsifying transform,

    I. Y . Chun, X. Zheng, Y . Long, and J. A. Fessler, “Sparse-view X- ray CT reconstruction usingℓ 1 regularization with learned sparsifying transform,” inInternational Meeting on Fully 3D Image Reconstruction in Radiology and Nuclear Medicine, 2017, pp. 115–9

    2017

  21. [29]

    Spectral ct reconstruction—assist: Aided by self-similarity in image- spectral tensors,

    W. Xia, W. Wu, S. Niu, F. Liu, J. Zhou, H. Yu, G. Wang, and Y . Zhang, “Spectral ct reconstruction—assist: Aided by self-similarity in image- spectral tensors,”IEEE Transactions on Computational Imaging, vol. 5, no. 3, pp. 420–436, 2019

    2019

  22. [30]

    Font-sir: Fourth-order nonlocal tensor decomposition model for spectral ct image reconstruction,

    X. Chen, W. Xia, Y . Liu, H. Chen, J. Zhou, Z. Zha, B. Wen, and Y . Zhang, “Font-sir: Fourth-order nonlocal tensor decomposition model for spectral ct image reconstruction,”IEEE Transactions on Medical Imaging, vol. 41, no. 8, pp. 2144–2156, 2022

    2022

  23. [31]

    Bayesian statistical reconstruction for low-dose X-ray computed tomography using an adaptive-weighting nonlocal prior,

    Y . Chen, D. Gao, C. Nie, L. Luo, W. Chen, X. Yin, and Y . Lin, “Bayesian statistical reconstruction for low-dose X-ray computed tomography using an adaptive-weighting nonlocal prior,”Computerized Medical Imaging and Graphics, vol. 33, no. 7, pp. 495–500, 2009

    2009

  24. [32]

    Iterative deblurring for ct metal artifact reduction,

    G. Wang, D. L. Snyder, J. A. O’Sullivan, and M. W. Vannier, “Iterative deblurring for ct metal artifact reduction,”IEEE transactions on medical imaging, vol. 15, no. 5, pp. 657–664, 1996

    1996

  25. [33]

    Iterative image reconstruction for cerebral perfusion ct using a pre- contrast scan induced edge-preserving prior,

    J. Ma, H. Zhang, Y . Gao, J. Huang, Z. Liang, Q. Feng, and W. Chen, “Iterative image reconstruction for cerebral perfusion ct using a pre- contrast scan induced edge-preserving prior,”Physics in Medicine & Biology, vol. 57, no. 22, p. 7519, 2012

    2012

  26. [34]

    A perspective on deep imaging,

    G. Wang, “A perspective on deep imaging,”IEEE Access, vol. 4, pp. 8914–8924, 2016

    2016

  27. [35]

    G. Wang, Y . Zhang, X. Ye, and X. Mou,Machine learning for tomo- graphic imaging. IOP Publishing, 2019

    2019

  28. [36]

    Low-dose ct denoising via sinogram inner-structure transformer,

    L. Yang, Z. Li, R. Ge, J. Zhao, H. Si, and D. Zhang, “Low-dose ct denoising via sinogram inner-structure transformer,”IEEE transactions on medical imaging, vol. 42, no. 4, pp. 910–921, 2022

    2022

  29. [37]

    Sinogram denoising via attention residual dense convolutional neural network for low-dose computed tomography,

    Y .-J. Ma, Y . Ren, P. Feng, P. He, X.-D. Guo, and B. Wei, “Sinogram denoising via attention residual dense convolutional neural network for low-dose computed tomography,”Nuclear Science and Techniques, vol. 32, pp. 1–14, 2021

    2021

  30. [38]

    Ct sinogram-consistency learning for metal-induced beam hardening correction,

    H. S. Park, S. M. Lee, H. P. Kim, J. K. Seo, and Y . E. Chung, “Ct sinogram-consistency learning for metal-induced beam hardening correction,”Medical physics, vol. 45, no. 12, pp. 5376–5384, 2018

    2018

  31. [39]

    Deep learning based sinogram correction for metal artifact reduction,

    M. U. Ghani and W. C. Karl, “Deep learning based sinogram correction for metal artifact reduction,”Electronic Imaging, vol. 30, p. art00005, 2018

    2018

  32. [40]

    Fast enhanced ct metal artifact reduction using data domain deep learning,

    ——, “Fast enhanced ct metal artifact reduction using data domain deep learning,”IEEE Transactions on Computational Imaging, vol. 6, pp. 181–193, 2019

    2019

  33. [41]

    Low-dose CT via convolutional neural network,

    H. Chen, Y . Zhang, W. Zhang, P. Liao, K. Li, J. Zhou, and G. Wang, “Low-dose CT via convolutional neural network,”Biomedical Optics Express, vol. 8, no. 2, pp. 679–694, 2017

    2017

  34. [42]

    Low-dose CT with a residual encoder-decoder convolutional neural network,

    H. Chen, Y . Zhang, M. K. Kalra, F. Lin, Y . Chen, P. Liao, J. Zhou, and G. Wang, “Low-dose CT with a residual encoder-decoder convolutional neural network,”IEEE Transactions on Medical Imaging, vol. 36, no. 12, pp. 2524–2535, 2017

    2017

  35. [43]

    Deep con- volutional neural network for inverse problems in imaging,

    K. H. Jin, M. T. McCann, E. Froustey, and M. Unser, “Deep con- volutional neural network for inverse problems in imaging,”IEEE Transactions on Image Processing, vol. 26, no. 9, pp. 4509–4522, 2017

    2017

  36. [44]

    A deep convolutional neural network us- ing directional wavelets for low-dose X-ray CT reconstruction,

    E. Kang, J. Min, and J. C. Ye, “A deep convolutional neural network us- ing directional wavelets for low-dose X-ray CT reconstruction,”Medical Physics, vol. 44, no. 10, pp. e360–e375, 2017

    2017

  37. [45]

    Low-dose CT image denoising using a generative adversarial network with wasserstein distance and perceptual loss,

    Q. Yang, P. Yan, Y . Zhanget al., “Low-dose CT image denoising using a generative adversarial network with wasserstein distance and perceptual loss,”IEEE Transactions on Medical Imaging, vol. 37, no. 6, pp. 1348– 1357, 2018

    2018

  38. [46]

    Noise suppression with similarity-based self-supervised deep learning,

    C. Niu, M. Li, F. Fan, W. Wu, X. Guo, Q. Lyu, and G. Wang, “Noise suppression with similarity-based self-supervised deep learning,”IEEE transactions on medical imaging, vol. 42, no. 6, pp. 1590–1602, 2022

    2022

  39. [47]

    Adn: artifact disen- tanglement network for unsupervised metal artifact reduction,

    H. Liao, W.-A. Lin, S. K. Zhou, and J. Luo, “Adn: artifact disen- tanglement network for unsupervised metal artifact reduction,”IEEE Transactions on Medical Imaging, vol. 39, no. 3, pp. 634–643, 2019. 9

    2019

  40. [48]

    Dicdnet: deep interpretable convolutional dictionary network for metal artifact reduction in ct images,

    H. Wang, Y . Li, N. He, K. Ma, D. Meng, and Y . Zheng, “Dicdnet: deep interpretable convolutional dictionary network for metal artifact reduction in ct images,”IEEE Transactions on Medical Imaging, vol. 41, no. 4, pp. 869–880, 2021

    2021

  41. [49]

    Low-dimensional manifold-constrained disentanglement network for metal artifact reduction,

    C. Niu, W. Cong, F.-L. Fan, H. Shan, M. Li, J. Liang, and G. Wang, “Low-dimensional manifold-constrained disentanglement network for metal artifact reduction,”IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 6, no. 6, pp. 656–666, 2021

    2021

  42. [50]

    Metal artifact reduction in 2d ct images with self-supervised cross-domain learning,

    L. Yu, Z. Zhang, X. Li, H. Ren, W. Zhao, and L. Xing, “Metal artifact reduction in 2d ct images with self-supervised cross-domain learning,” Physics in Medicine & Biology, vol. 66, no. 17, p. 175003, 2021

    2021

  43. [51]

    LEARN: Learned experts’ assessment- based reconstruction network for sparse-data CT,

    H. Chen, Y . Zhang, Y . Chen, J. Zhang, W. Zhang, H. Sun, Y . Lv, P. Liao, J. Zhou, and G. Wang, “LEARN: Learned experts’ assessment- based reconstruction network for sparse-data CT,”IEEE Transactions on Medical Imaging, vol. 37, no. 6, pp. 1333–1347, 2018

    2018

  44. [52]

    CNN-based projected gradient descent for consistent CT image reconstruction,

    H. Gupta, K. H. Jin, H. Q. Nguyenet al., “CNN-based projected gradient descent for consistent CT image reconstruction,”IEEE Transactions on Medical Imaging, vol. 37, no. 6, pp. 1440–1453, 2018

    2018

  45. [53]

    Optimizing a parameterized plug-and-play ADMM for iterative low-dose CT reconstruction,

    J. He, Y . Yang, Y . Wang, D. Zeng, Z. Bian, H. Zhang, J. Sun, Z. Xu, and J. Ma, “Optimizing a parameterized plug-and-play ADMM for iterative low-dose CT reconstruction,”IEEE Transactions on Medical Imaging, vol. 38, no. 2, pp. 371–382, 2018

    2018

  46. [54]

    Adaptive convolutional dictionary network for ct metal artifact reduction,

    H. Wang, Y . Li, D. Meng, and Y . Zheng, “Adaptive convolutional dictionary network for ct metal artifact reduction,”arXiv preprint arXiv:2205.07471, 2022

    2022 arXiv

  47. [55]

    FISTA-net: Learning a fast iterative shrinkage thresholding network for inverse problems in imaging,

    J. Xiang, Y . Dong, and Y . Yang, “FISTA-net: Learning a fast iterative shrinkage thresholding network for inverse problems in imaging,”IEEE Transactions on Medical Imaging, vol. 40, no. 5, pp. 1329–1339, 2021

    2021

  48. [56]

    MAGIC: Manifold and graph integrative convolutional network for low-dose CT reconstruction,

    W. Xia, Z. Lu, Y . Huang, Z. Shi, Y . Liu, H. Chen, Y . Chen, J. Zhou, and Y . Zhang, “MAGIC: Manifold and graph integrative convolutional network for low-dose CT reconstruction,”IEEE Transactions on Medical Imaging, vol. 40, no. 12, pp. 3459–3472, 2021

    2021

  49. [57]

    Regformer: A local– nonlocal regularization-based model for sparse-view ct reconstruction,

    W. Xia, Z. Yang, Z. Lu, Z. Wang, and Y . Zhang, “Regformer: A local– nonlocal regularization-based model for sparse-view ct reconstruction,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 8, no. 2, pp. 184–194, 2023

    2023

  50. [58]

    Convolutional neural network based metal artifact reduction in x-ray computed tomography,

    Y . Zhang and H. Yu, “Convolutional neural network based metal artifact reduction in x-ray computed tomography,”IEEE transactions on medical imaging, vol. 37, no. 6, pp. 1370–1381, 2018

    2018

  51. [59]

    A dual-stream deep convolutional network for reducing metal streak artifacts in ct images,

    L. Gjesteby, H. Shan, Q. Yang, Y . Xi, Y . Jin, D. Giantsoudi, H. Paganetti, B. De Man, and G. Wang, “A dual-stream deep convolutional network for reducing metal streak artifacts in ct images,”Physics in Medicine & Biology, vol. 64, no. 23, p. 235003, 2019

    2019

  52. [60]

    Deep sinogram completion with image prior for metal artifact reduction in ct images,

    L. Yu, Z. Zhang, X. Li, and L. Xing, “Deep sinogram completion with image prior for metal artifact reduction in ct images,”IEEE transactions on medical imaging, vol. 40, no. 1, pp. 228–238, 2020

    2020

  53. [61]

    Image reconstruction by domain-transform manifold learning,

    B. Zhu, J. Z. Liu, S. F. Cauley, B. R. Rosen, and M. S. Rosen, “Image reconstruction by domain-transform manifold learning,”Nature, vol. 555, no. 7697, pp. 487–492, 2018

    2018

  54. [62]

    Radon inversion via deep learning,

    J. He, Y . Wang, and J. Ma, “Radon inversion via deep learning,”IEEE Transactions on Medical Imaging, vol. 39, no. 6, pp. 2076–2087, 2020

    2020

  55. [63]

    Downsampled imaging geometric modeling for accurate CT reconstruction via deep learning,

    J. He, S. Chen, H. Zhang, X. Tao, W. Lin, S. Zhang, D. Zeng, and J. Ma, “Downsampled imaging geometric modeling for accurate CT reconstruction via deep learning,”IEEE Transactions on Medical Imaging, vol. 40, no. 11, pp. 2976–2985, 2021

    2021

  56. [64]

    Hybrid-domain neural network processing for sparse-view CT reconstruction,

    D. Hu, J. Liu, T. Lv, Q. Zhao, Y . Zhang, G. Quan, J. Feng, Y . Chen, and L. Luo, “Hybrid-domain neural network processing for sparse-view CT reconstruction,”IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 5, no. 1, pp. 88–98, 2020

    2020

  57. [65]

    CLEAR: comprehensive learning enabled adversarial reconstruction for subtle structure enhanced low- dose CT imaging,

    Y . Zhang, D. Hu, Q. Zhao, G. Quan, J. Liu, Q. Liu, Y . Zhang, G. Coatrieux, Y . Chen, and H. Yu, “CLEAR: comprehensive learning enabled adversarial reconstruction for subtle structure enhanced low- dose CT imaging,”IEEE Transactions on Medical Imaging, vol. 40, no. 11, pp. 30...

    2021

  58. [66]

    Learning to reconstruct CT images from the vvbp-tensor,

    X. Tao, Y . Wang, L. Lin, Z. Hong, and J. Ma, “Learning to reconstruct CT images from the vvbp-tensor,”IEEE Transactions on Medical Imaging, vol. 40, no. 11, pp. 3030–3041, 2021

    2021

  59. [67]

    Learned primal-dual reconstruction,

    J. Adler and O. ¨Oktem, “Learned primal-dual reconstruction,”IEEE Transactions on Medical Imaging, vol. 37, no. 6, pp. 1322–1332, 2018

    2018

  60. [68]

    Dudonet: Dual domain network for ct metal artifact reduction,

    W.-A. Lin, H. Liao, C. Peng, X. Sun, J. Zhang, J. Luo, R. Chellappa, and S. K. Zhou, “Dudonet: Dual domain network for ct metal artifact reduction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 10 512–10 521

    2019

  61. [69]

    Idol-net: An interactive dual-domain parallel network for ct metal artifact reduction,

    T. Wang, Z. Lu, Z. Yang, W. Xia, M. Hou, H. Sun, Y . Liu, H. Chen, J. Zhou, and Y . Zhang, “Idol-net: An interactive dual-domain parallel network for ct metal artifact reduction,”IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 6, no. 8, pp. 874–885, 2022

    2022

  62. [70]

    Dan-net: Dual-domain adaptive-scaling non-local network for ct metal artifact reduction,

    T. Wang, W. Xia, Y . Huang, H. Sun, Y . Liu, H. Chen, J. Zhou, and Y . Zhang, “Dan-net: Dual-domain adaptive-scaling non-local network for ct metal artifact reduction,”Physics in Medicine & Biology, vol. 66, no. 15, p. 155009, 2021

    2021

  63. [71]

    Cyclegan denoising of ex- treme low-dose cardiac ct using wavelet-assisted noise disentanglement,

    J. Gu, T. S. Yang, J. C. Ye, and D. H. Yang, “Cyclegan denoising of ex- treme low-dose cardiac ct using wavelet-assisted noise disentanglement,” Medical image analysis, vol. 74, p. 102209, 2021

    2021

  64. [72]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in Neural Information Processing Systems, vol. 33, pp. 6840– 6851, 2020

    2020

  65. [73]

    Image super-resolution via iterative refinement,

    C. Saharia, J. Ho, W. Chan, T. Salimans, D. J. Fleet, and M. Norouzi, “Image super-resolution via iterative refinement,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022

    2022

  66. [74]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,”Advances in Neural Information Processing Systems, vol. 34, pp. 8780–8794, 2021

    2021

  67. [75]

    A denoising diffusion probabilistic model for metal artifact reduction in ct,

    G. M. Karageorgos, J. Zhang, N. Peters, W. Xia, C. Niu, H. Paganetti, G. Wang, and B. De Man, “A denoising diffusion probabilistic model for metal artifact reduction in ct,”IEEE Transactions on Medical Imaging, 2024

    2024

  68. [76]

    Low-dose ct using denoising diffusion probabilistic model for 20x speedup,

    W. Xia, Q. Lyu, and G. Wang, “Low-dose ct using denoising diffusion probabilistic model for 20x speedup,”arXiv preprint arXiv:2209.15136, 2022

    2022 arXiv

  69. [77]

    Patch-based denoising diffusion probabilistic model for sparse-view CT reconstruction,

    W. Xia, W. Cong, and G. Wang, “Patch-based denoising diffusion probabilistic model for sparse-view CT reconstruction,”arXiv preprint arXiv:2211.10388, 2022

    2022 arXiv

  70. [78]

    Diffusion prior regularized iterative reconstruction for low-dose ct,

    W. Xia, Y . Shi, C. Niu, W. Cong, and G. Wang, “Diffusion prior regularized iterative reconstruction for low-dose ct,”arXiv preprint arXiv:2310.06949, 2023

    2023 arXiv

  71. [79]

    Poisson flow generative models,

    Y . Xu, Z. Liu, M. Tegmark, and T. Jaakkola, “Poisson flow generative models,”Advances in Neural Information Processing Systems, vol. 35, pp. 16 782–16 795, 2022

    2022

  72. [80]

    Pfgm++: Unlocking the potential of physics-inspired generative models,

    Y . Xu, Z. Liu, Y . Tian, S. Tong, M. Tegmark, and T. Jaakkola, “Pfgm++: Unlocking the potential of physics-inspired generative models,” in International Conference on Machine Learning. PMLR, 2023, pp. 38 566–38 591

    2023

  73. [81]

    Score-based generative modeling through stochastic differ- ential equations,

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differ- ential equations,”arXiv preprint arXiv:2011.13456, 2020

    2011 arXiv

  74. [82]

    Elucidating the design space of diffusion-based generative models,

    T. Karras, M. Aittala, T. Aila, and S. Laine, “Elucidating the design space of diffusion-based generative models,”Advances in neural infor- mation processing systems, vol. 35, pp. 26 565–26 577, 2022

    2022

  75. [83]

    Solving inverse problems in medical imaging with score-based generative models,

    Y . Song, L. Shen, L. Xing, and S. Ermon, “Solving inverse problems in medical imaging with score-based generative models,”arXiv preprint arXiv:2111.08005, 2022

    2022 arXiv

  76. [84]

    Improving diffusion models for inverse problems using manifold constraints,

    H. Chung, B. Sim, D. Ryu, V . Dumoulin, and J. C. Ye, “Improving diffusion models for inverse problems using manifold constraints,” in Advances in Neural Information Processing Systems, vol. 35, 2022, pp. 25 683–25 696

    2022

  77. [85]

    The little engine that could: Regularization by denoising (red),

    Y . Romano, M. Elad, and P. Milanfar, “The little engine that could: Regularization by denoising (red),”SIAM Journal on Imaging Sciences, vol. 10, no. 4, pp. 1804–1844, 2017

    2017

  78. [86]

    Ordered-subset simultaneous algebraic re- construction techniques (OS-SART),

    G. Wang and M. Jiang, “Ordered-subset simultaneous algebraic re- construction techniques (OS-SART),”Journal of X-ray Science and Technology, vol. 12, no. 3, pp. 169–177, 2004

    2004

  79. [87]

    Latent consistency models: Synthesizing high-resolution images with few-step inference,

    S. Luo, Y . Tan, L. Huang, J. Li, and H. Zhao, “Latent consistency models: Synthesizing high-resolution images with few-step inference,” arXiv preprint arXiv:2310.04378, 2023

    2023 arXiv

  80. [88]

    Neural discrete representation learning,

    A. Van Den Oord, O. Vinyalset al., “Neural discrete representation learning,”Advances in neural information processing systems, vol. 30, 2017

    2017

  81. [89]

    In-context learning unlocked for diffusion models,

    Z. Wang, Y . Jiang, Y . Lu, P. He, W. Chen, Z. Wang, M. Zhouet al., “In-context learning unlocked for diffusion models,”Advances in Neural Information Processing Systems, vol. 36, pp. 8542–8562, 2023

    2023

  82. [90]

    Seggpt: Segmenting everything in context,

    X. Wang, X. Zhang, Y . Cao, W. Wang, C. Shen, and T. Huang, “Seggpt: Segmenting everything in context,”arXiv preprint arXiv:2304.03284, 2023

    2023 arXiv

  83. [91]

    Catsim: a new computer assisted tomography simulation environment,

    B. De Man, S. Basu, N. Chandra, B. Dunham, P. Edic, M. Iatrou, S. McOlash, P. Sainath, C. Shaughnessy, B. Toweret al., “Catsim: a new computer assisted tomography simulation environment,” inMedical Imaging 2007: Physics of Medical Imaging, vol. 6510. SPIE, 2007, pp. 856–863

    2007

  84. [92]

    Dukesim: a realistic, rapid, and scanner-specific simulation framework in computed tomography,

    E. Abadi, B. Harrawood, S. Sharma, A. Kapadia, W. P. Segars, and E. Samei, “Dukesim: a realistic, rapid, and scanner-specific simulation framework in computed tomography,”IEEE transactions on medical imaging, vol. 38, no. 6, pp. 1457–1465, 2018. 10

    2018

  85. [93]

    A gpu tool for efficient, accurate, and realistic simulation of cone beam ct projections,

    X. Jia, H. Yan, L. Cervi ˜no, M. Folkerts, and S. B. Jiang, “A gpu tool for efficient, accurate, and realistic simulation of cone beam ct projections,” Medical physics, vol. 39, no. 12, pp. 7368–7378, 2012

    2012

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.