Pith. sign in

REVIEW 3 major objections 6 minor 36 references

SS-CTML: Self-Supervised Cross-Task Mutual Learning for CT Image Reconstruction

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read Three CT reconstruction tasks train each other without any normal-dose paired data.

desk verdict Clever self-supervised multi-task CT setup with strong numbers, but the loss design leaves the mutual-learning mechanism underdetermined; worth a serious look with a requested control ablation. read the letter →

arxiv 2501.01456 v1 pith:QVA7CZQA submitted 2024-12-31 eess.IV cs.CVcs.LG

classification eess.IVcs.CVcs.LG
keywords CTimagereconstructionself-supervisedlearningcross-taskmutualsparse-viewlimited-viewlow-dosedual-domainnetworkknowledgedistillation
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

This paper tries to establish that low-dose full-view CT, sparse-view CT, and limited-view CT reconstruction can be improved simultaneously by a self-supervised mutual-learning framework that never uses normal-dose images as training targets. The authors derive sparse-view and limited-view sinograms from a single low-dose full-view scan, build three subnetworks for the three tasks, and constrain the networks to agree on shared anatomy. The payoff, if the claim holds, is that routine low-dose scans alone can train networks for several CT artifact-removal tasks, removing a major obstacle to clinical use of supervised deep learning, which is hard to obtain in practice.

What carries the argument

The load-bearing machinery is a trio of subnetworks, one per task, each made of a prior neural module (PNM) and a dual-domain neural module (DDNM). The PNM takes the contaminated FBP image and produces a cleaned prior image; that prior is forward-projected into the sinogram domain, where for sparse- and limited-view tasks a binary mask fills the missing angular positions with the forward projection of the prior while retaining measured data elsewhere, and for the full-view task the projection is concatenated with the measured sinogram. The compensated sinogram then passes through the DDNM, which is a sinogram network, a back-propagatable FBP layer, and an image network. The cross-task mutual-learning losses $L_{\text{prior}}^{\text{ml}}$ and $L_{\text{out}}^{\text{ml}}$ are $\ell^2$ distances among the three priors and among the three final outputs, and the reconstruction-consistency loss $L_{\text{rc}}$ ties the full-view branch to the low-mAs FBP image $\mu_{ld}$; the total loss is their sum. These objectives are what let one low-dose dataset supervise three tasks at once.

What would settle it

Feed the framework a test slice in which the low-mAs FBP image carries a strong known bias that the sparse- and limited-view degradations do not share; if the mutual-learning outputs reproduce that biased structure instead of removing it, the reconstruction-consistency anchor is pulling the consensus toward the artifact rather than toward true anatomy.

Watch

Extended reading notes

Core claim

The central claim is that the FBP reconstructions of the same slice from low-mAs full-view, sparse-view, and limited-view scans carry complementary information about true anatomy, so forcing three subnetworks to produce mutually consistent priors and final outputs, while anchoring the full-view branch to the low-mAs FBP image, optimizes all three tasks without any normal-dose reference. On simulated quarter-, sixth-, and eighth-dose data from a public dataset, the paper reports that SS-CTML outperforms the supervised baselines FBPConvNet, iRadonMap, and FreeSeed for sparse-view CT, outperforms FBPConvNet, Sam's Net, and DOLCE for limited-view CT, and is competitive with the supervised RED-CNN and diffusion-based Dn-Dp for low-mAs full-view CT. On real clinical low-dose scans, the trained model reduces noise and streak artifacts qualitatively for all three tasks, with no quantitative reference available.

Load-bearing premise

The method assumes that the three degraded FBP images of the same slice are complementary but consistent views of true anatomy, so that forcing them to agree pushes each network toward a clean image; in particular, every task is anchored to the low-mAs FBP image $\mu_{ld}$, and if that image is too corrupted the mutual-learning consensus could converge to a degraded average rather than the truth.

Editorial extensions

If this is right

  • A single self-supervised training run on low-dose sinograms produces networks for all three reconstruction tasks at once, reducing the need for paired normal-dose training data in each task.
  • Per the paper's results, the sparse-view and limited-view tasks can match or exceed several dedicated supervised methods, so self-supervised mutual learning is a viable alternative where paired data are unavailable.
  • The ablation study implies that each task contributes: removing the full-view task sharply degrades the remaining two, so the cross-task correlation is the source of the gain, not a side effect.
  • The trained model transfers to real clinical low-dose scans of different anatomy and reduces visible noise and artifacts, suggesting the learned consensus is not overfit to the simulated training geometry.
  • Both the prior-image consensus and the final-output consensus contribute, since ablating either the PNM or the DDNM lowers performance; the dual-domain structure carries part of the benefit.

Reading between the lines

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

  • I would expect the same mutual-agreement recipe to extend beyond CT geometries: any set of image-degradation tasks derived from one measurement, such as different sparse-view rates or masked sinogram regions, could be jointly self-supervised, with retraining per geometry as the paper acknowledges.
  • The weakest spot I see is the anchor: if the full-view low-dose FBP image itself contains a structured artifact, Eq. (7) gives that artifact a privileged position in the consensus, so I would test the method under extreme noise or with a deliberately biased reference.
  • One network with task-specific heads might replace the three separate subnetworks and preserve most of the mutual-learning benefit while cutting memory; the paper itself flags the three-network design as redundant.
  • Because the three tasks share only the underlying anatomy, the mutual losses act as a form of self-distillation across geometries; comparing against a single-network denoiser trained with the same total compute would isolate what the cross-task coupling actually buys.
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

3 major / 6 minor

Summary. The paper proposes SS-CTML, a self-supervised framework for low-mAs full-view CT (FVCT), sparse-view CT (SVCT), and limited-view CT (LVCT) reconstruction. From a low-mAs full-view sinogram it extracts sparse-view and limited-view sinograms, trains three dual-domain subnetworks with prior neural modules and compensation modules, and optimizes them with pairwise mutual-learning losses on prior images and final outputs plus a reconstruction-consistency loss anchored to the low-mAs FBP image. Experiments on the Mayo dataset at quarter/sixth/eighth dose and on ELCAP real data report PSNR/NMSE/SSIM; SS-CTML outperforms the unsupervised baselines and, for SVCT and LVCT, the supervised baselines, while for FVCT it is below RED-CNN. The authors conclude that cross-task mutual learning improves all three tasks without normal-dose paired training.

Significance. If the central claim holds, SS-CTML would be a practically valuable self-supervised alternative to paired-data reconstruction, since it learns from low-mAs sinograms alone and improves three tasks simultaneously; the paper provides extensive quantitative tables across three dose levels, a real-data demonstration, and an ablation of the main modules. The evaluation is not circular because the normal-dose references are held out from training. However, the proposed mutual-learning mechanism is not isolated by the current experiments, and the baseline comparisons need more transparency; with those gaps closed, the work would be a solid contribution.

major comments (3)
  1. [III-C, Eqs. (5)-(8); V-C, Table IV] The central claim that cross-task mutual learning is the active mechanism is not established by the loss design or the ablations. Lprior_ml and Lout_ml are symmetric pairwise agreement terms whose gradients vanish at any consensus configuration, and Lrc in Eq. (7) is the only asymmetric term, anchoring all outputs to the corrupted FBP image mu_ld. The objective therefore admits a near-zero-loss solution in which f_ld is identity and f_sv/f_lv map their inputs to mu_ld, so the improvements over FBP reported in Tables I-III could in principle come from the U-net backbone's implicit regularization rather than from cross-task consensus. Table IV removes whole tasks or modules (w/o FVCT, w/o SVCT, w/o LVCT, w/o PNM, w/o DDNM) but never removes the mutual-learning losses while keeping the same anchor and architecture; notably, w/o FVCT drops SVCT PSNR to 29.10 dB, which is consistent with losing the Lrc anchor rather than with losing mutual learning. Please add an ablation training the same three subnetworks with only Lrc (or only per-task consistency toward mu_ld) and report the three tasks, and specify which loss terms remain in the w/o FVCT variant.
  2. [IV-C, Tables I-III] The quantitative comparison to supervised baselines is currently difficult to interpret because the training protocols of the competing methods are not specified. It is not stated whether RED-CNN, FBPConvNet, iRadonMap, FreeSeed, Sam's Net, Dn-Dp, and DOLCE were retrained on the same simulated low-dose data with the same view geometries, or with which hyperparameters and implementations. The very low DOLCE scores (18.84/18.29/19.03 dB across dose levels versus 25-34 dB for other LVCT methods) suggest the baseline may not have been adapted to the mixed limited-view and low-mAs setting, which would bias the comparison in favor of SS-CTML. Please provide the training protocol for each baseline and report paired statistical significance tests (for example, Wilcoxon signed-rank or paired t-test over slices) for the main comparisons.
  3. [III-A, Eq. (7); V-A] The framework's assumption that the three FBP images contain complementary but consistent anatomy so that consensus via Eqs. (5)-(6) drives toward true anatomy is not directly tested. Since Eq. (7) anchors the entire objective to the low-mAs FBP image mu_ld, a heavily corrupted mu_ld could make the consensus a degraded average; the paper does not report how close the converged consensus is to the normal-dose reference on a case-by-case basis, or compare with an alternative anchor such as a filtered or iterative reconstruction of pld. An additional experiment on a numerical phantom, or with a deliberately corrupted mu_ld, would show whether mutual learning recovers truth rather than a common artifact pattern, and would directly address the load-bearing assumption.
minor comments (6)
  1. [I, VII] Section I contains the typo 'real clinal practices' and should read 'real clinical practices'; Section VII contains 'the the effectiveness' and should read 'the effectiveness'.
  2. [Fig. 9] The caption says 'SVCT reconstruction task' but the figure shows LVCT results; the caption should read 'LVCT reconstruction task'.
  3. [IV-B] The phrase 'the batch number was 1000' is ambiguous; please specify the batch size, the number of training epochs, and the learning-rate schedule, and justify the unusual Adam beta values (0.5, 0.9).
  4. [Fig. 3] The axis label 'Epoches' should be 'Epochs', and the legends for the three tasks in panel (b) are difficult to read; please enlarge the fonts and clarify the panel labels.
  5. [Eq. (3)] The use of concatenation (the symbol ⊕) only for the FVCT branch, while SVCT and LVCT use masking, is not explained; please clarify how concatenation acts as sinogram compensation in the low-mAs case.
  6. [Reproducibility] The paper does not release code or data; adding a reproducibility statement and the trained models would strengthen the empirical claims.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the Eq. (7) anchor to the input FBP is self-referential but the reported improvements are externally evaluated against held-out normal-dose references.

full rationale

The central claim, that the three subnetworks can improve low-mAs FVCT, SVCT, and LVCT reconstruction without normal-dose paired training data, is not derived by definition from the loss functions. The only self-referential element is the reconstruction consistency loss in Eq. (7), which anchors the FVCT prior and final output to mu_ld, the FBP image of the low-dose sinogram pld. Together with the symmetric mutual-learning losses in Eqs. (5) and (6), the total loss in Eq. (8) admits a degenerate global minimum where all outputs equal mu_ld; this means the mutual-learning terms alone do not logically force improvement over FBP, and any observed gain must come partly from the U-net backbone's implicit regularization. This is a genuine weakness in the claimed mechanism's identifiability, and the ablations in Table IV do not isolate the mutual losses while keeping the anchor and architecture fixed, so the specific contribution of cross-task consensus is under-supported. However, this is not circularity in the prediction sense: the quantitative results in Tables I-III are measured against normal-dose FBP references that are explicitly held out and used only for evaluation (Section IV-A), not fitted or used as training targets. The cited prior works [33], [34] are used as architectural references and competing methods rather than as an unverified load-bearing premise, and no uniqueness theorem is imported from the authors' earlier work. The Discussion's stated limitations concern fixed imaging geometries, retraining requirements, and architectural redundancy, which are generalization and efficiency concerns, not circularity. Overall, the empirical performance claim is self-contained against external benchmarks, so no significant circularity is present.

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

This is an empirical deep-learning paper. The central performance claim depends on hand-chosen imaging geometries, training hyperparameters, and the assumption that mutual consistency among degraded views provides a useful training signal. No new physical entities are introduced.

free parameters (5)
  • SVCT exposure view count = 144
    Chosen protocol for the sparse-view task; it controls artifact level and affects all reported SVCT results.
  • LVCT angular range and view count = 120 degrees, 384 views
    Chosen limited-view geometry; combined with dose level it sets the difficulty of the LVCT task.
  • Training hyperparameters = Initial LR 1e-5, Adam betas 0.5/0.9, batch size 1000, about 400 epochs
    Hand-selected training schedule; epoch count is inferred from Fig. 3 rather than stated in the text.
  • Five-stage U-net backbone = Residual U-net, five stages
    Architecture choice for PNM and DDNM; the authors note that more advanced backbones could improve performance.
  • Loss weighting = Implicit 1:1:1 weights
    Lprior_ml, Lout_ml, and Lrc are summed as equal terms in Eq. (8) with no sensitivity analysis.
assumptions (5)
  • domain assumption Fan-beam forward projection and FBP layers accurately model the acquisition geometry and are differentiable for backpropagation.
    Invoked in Eq. (3) and the DDNM via FPLayer and FBPLayer; if the projector mismatches the scanner, the sinogram compensation is invalid.
  • domain assumption SVCT and LVCT data extracted by subsampling the low-mAs fan-beam sinogram faithfully represent those acquisition protocols.
    Section IV-A.1 describes extraction of p_sv and p_lv from p_ld; this assumes the simulated geometries capture real clinical sparse-view and limited-view acquisition.
  • domain assumption The same anatomic slice across FVCT, SVCT, and LVCT tasks contains identical anatomy, so output consistency is a valid training signal.
    Section III-A and Eqs. (5)-(6); this premise is the basis for the mutual learning mechanism.
  • domain assumption The degraded FBP image μ_ld is a sufficiently good anchor that pulling outputs toward it does not cap achievable quality or inject bias.
    Eq. (7) uses μ_ld as the only reconstruction consistency target; the self-supervised behavior depends on this anchor being informative.
  • domain assumption L2 losses provide adequate gradients for suppressing CT noise, streak artifacts, and limited-view artifacts.
    Eqs. (5)-(7) use only L2 terms; no perceptual or edge-preserving terms are included, and no analysis of the loss landscape is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SS-CTML: Self-Supervised Cross-Task Mutual Learning for CT Image Reconstruction." pith.science (2026). https://pith.science/paper/QVA7CZQA

@misc{pith2026250101456,
  author       = {Pith},
  title        = {Pith review of: SS-CTML: Self-Supervised Cross-Task Mutual Learning for CT Image Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QVA7CZQA}},
  note         = {Machine review of arXiv:2501.01456}
}
read the original abstract

Supervised deep-learning (SDL) techniques with paired training datasets have been widely studied for X-ray computed tomography (CT) image reconstruction. However, due to the difficulties of obtaining paired training datasets in clinical routine, the SDL methods are still away from common uses in clinical practices. In recent years, self-supervised deep-learning (SSDL) techniques have shown great potential for the studies of CT image reconstruction. In this work, we propose a self-supervised cross-task mutual learning (SS-CTML) framework for CT image reconstruction. Specifically, a sparse-view scanned and a limited-view scanned sinogram data are first extracted from a full-view scanned sinogram data, which results in three individual reconstruction tasks, i.e., the full-view CT (FVCT) reconstruction, the sparse-view CT (SVCT) reconstruction, and limited-view CT (LVCT) reconstruction. Then, three neural networks are constructed for the three reconstruction tasks. Considering that the ultimate goals of the three tasks are all to reconstruct high-quality CT images, we therefore construct a set of cross-task mutual learning objectives for the three tasks, in which way, the three neural networks can be self-supervised optimized by learning from each other. Clinical datasets are adopted to evaluate the effectiveness of the proposed framework. Experimental results demonstrate that the SS-CTML framework can obtain promising CT image reconstruction performance in terms of both quantitative and qualitative measurements.

Figures

Figures reproduced from arXiv: 2501.01456 by the authors.

Figure 1
Figure 1. Overview of the proposed SS-CTML framework. The mutual learning of cross-task (i.e., low-mAs FVCT, SVCT, and LVCT reconstruction tasks) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The comparison of different reconstruction tasks. From (a) to (c): low [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (a) Total loss curves for training and validation; and (b) PSNR curves [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The simulated data results of low-mAs FVCT reconstruction task for different methods with different dose levels. The three rows show quarter-dose, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The simulated data results of SVCT reconstruction task for different methods with different dose levels. The exposure view number of SVCT is [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The simulated data results of LVCT reconstruction task for different methods with different dose levels. The exposure view range of LVCT is set to [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: The absolute difference images of SVCT reconstruction task in [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: The absolute difference images of SVCT reconstruction task in Fig. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: The real clinical data results of different CT reconstruction tasks. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 29 canonical work pages

  1. [1]

    Computed tomography–an increasing source of radiation exposure,

    D. J. Brenner and E. J. Hall, “Computed tomography–an increasing source of radiation exposure,” New England Journal of Medicine , vol. 357, no. 22, pp. 2277–2284, 2007

  2. [2]

    Deep convolutional framelet denosing for low-dose CT via wavelet residual network,

    E. Kang, W. Chang, J. Yoo, and J. C. Ye, “Deep convolutional framelet denosing for low-dose CT via wavelet residual network,” IEEE Trans- actions on Medical Imaging , vol. 37, no. 6, pp. 1358–1369, 2018

  3. [3]

    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. 10 D D D E E E F F F )%3 )%3 )%3 'Q '3 )UHH6HHG '2/&( 66 &70/ 66 &70/ 66 &70/ )9&7 69&7 /9&7 Fig. 10. The real clinical data results of different CT reconstr...

  4. [4]

    Sam’s net: a self-augmented multistage deep-learning network for end-to-end recon- struction of limited angle CT,

    C. Chen, Y . Xing, H. Gao, L. Zhang, and Z. Chen, “Sam’s net: a self-augmented multistage deep-learning network for end-to-end recon- struction of limited angle CT,” IEEE Transactions on Medical Imaging , vol. 41, no. 10, pp. 2912–2924, 2022

  5. [5]

    Semi-supervised learned sinogram restoration network for low-dose CT image reconstruction,

    M. Meng, S. Li, L. Yao, D. Li, M. Zhu, Q. Gao, Q. Xie, Q. Zhao, Z. Bian, J. Huang et al., “Semi-supervised learned sinogram restoration network for low-dose CT image reconstruction,” in Medical Imaging 2020: Physics of Medical Imaging , vol. 11312. SPIE, 2020, pp. 67– 73

  6. [6]

    Noise characteristics modeled unsupervised network for robust CT image reconstruction,

    D. Li, Z. Bian, S. Li, J. He, D. Zeng, and J. Ma, “Noise characteristics modeled unsupervised network for robust CT image reconstruction,” IEEE Transactions on Medical Imaging , vol. 41, no. 12, pp. 3849–3861, 2022

  7. [7]

    IntraTomo: self-supervised learning-based tomography via sinogram synthesis and prediction,

    G. Zang, R. Idoughi, R. Li, P. Wonka, and W. Heidrich, “IntraTomo: self-supervised learning-based tomography via sinogram synthesis and prediction,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 1960–1970

  8. [8]

    Noise2Noise: Learning image restora- tion without clean data,

    J. Lehtinen, J. Munkberg, J. Hasselgren, S. Laine, T. Karras, M. Aittala, and T. Aila, “Noise2Noise: Learning image restora- tion without clean data,” arXiv preprint arXiv:1803.04189, 2018, https://arxiv.org/abs/1803.04189

Show all 36 references
  1. [9]

    Neighbor2neighbor: Self- supervised denoising from single noisy images,

    T. Huang, S. Li, X. Jia, H. Lu, and J. Liu, “Neighbor2neighbor: Self- supervised denoising from single noisy images,” in Proceedings of The IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 14 781–14 790

  2. [10]

    Noise2self: Blind denoising by self- supervision,

    J. Batson and L. Royer, “Noise2self: Blind denoising by self- supervision,” in International Conference on Machine Learning . PMLR, 2019, pp. 524–533

  3. [11]

    Noise2inverse: Self-supervised deep convolutional denoising for tomography,

    A. A. Hendriksen, D. M. Pelt, and K. J. Batenburg, “Noise2inverse: Self-supervised deep convolutional denoising for tomography,” IEEE Transactions on Computational Imaging , vol. 6, pp. 1320–1335, 2020

  4. [12]

    Noise2Filter: fast, self-supervised learning and real-time reconstruction for 3D computed tomography,

    M. J. Lagerwerf, A. A. Hendriksen, J.-W. Buurlage, and K. J. Batenburg, “Noise2Filter: fast, self-supervised learning and real-time reconstruction for 3D computed tomography,” Machine Learning: Science and Tech- nology, vol. 2, no. 1, p. 015012, 2020

  5. [13]

    SACNN: Self-attention convolutional neural network for low-dose CT denoising with self- supervised perceptual loss network,

    M. Li, W. Hsu, X. Xie, J. Cong, and W. Gao, “SACNN: Self-attention convolutional neural network for low-dose CT denoising with self- supervised perceptual loss network,” IEEE Transactions on Medical Imaging, vol. 39, no. 7, pp. 2289–2301, 2020

  6. [14]

    Unsharp structure guided filtering for self-supervised low-dose CT imaging,

    Q. Wu, X. Ji, Y . Gu, J. Xiang, G. Quan, B. Li, J. Zhu, G. Coatrieux, J.-L. Coatrieux, and Y . Chen, “Unsharp structure guided filtering for self-supervised low-dose CT imaging,” IEEE Transactions on Medical Imaging, vol. 42, no. 11, pp. 3283–3294, 2023

  7. [15]

    Self-Supervised Coordi- nate Projection Network for Sparse-View Computed Tomography,

    Q. Wu, R. Feng, H. Wei, J. Yu, and Y . Zhang, “Self-Supervised Coordi- nate Projection Network for Sparse-View Computed Tomography,”IEEE Transactions on Computational Imaging , vol. 9, pp. 517–529, 2023

  8. [16]

    Iterative reconstruction for limited-angle CT using implicit neural representation,

    J. Lee and J. Baek, “Iterative reconstruction for limited-angle CT using implicit neural representation,” Physics in Medicine & Biology , vol. 69, no. 10, p. 105008, 2024

  9. [17]

    DOLCE: A model-based probabilistic diffusion framework for limited-angle CT reconstruction,

    J. Liu, R. Anirudh, J. J. Thiagarajan, S. He, K. A. Mohan, U. S. Kamilov, and H. Kim, “DOLCE: A model-based probabilistic diffusion framework for limited-angle CT reconstruction,” in 2023 IEEE/CVF International Conference on Computer Vision (ICCV) , 2023, pp. 10 464–10 474

  10. [18]

    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

  11. [19]

    Diffusion model for gen- erative image denoising,

    Y . Xie, M. Yuan, B. Dong, and Q. Li, “Diffusion model for gen- erative image denoising,” arXiv preprint arXiv:2302.02398, 2023, https://arxiv.org/abs/2302.02398

  12. [20]

    Denoising diffusion restoration models,

    B. Kawar, M. Elad, S. Ermon, and J. Song, “Denoising diffusion restoration models,” Advances in Neural Information Processing Sys- tems, vol. 35, pp. 23 593–23 606, 2022

  13. [21]

    CoCoDiff: a contextual conditional diffusion model for low-dose CT image denoising,

    Q. Gao and H. Shan, “CoCoDiff: a contextual conditional diffusion model for low-dose CT image denoising,” in Developments in X-Ray Tomography XIV, vol. 12242. SPIE, 2022

  14. [22]

    Diffusion probabilistic priors for zero-shot low-dose CT image denoising,

    X. Liu, Y . Xie, J. Cheng, S. Diao, S. Tan, and X. Liang, “Diffusion probabilistic priors for zero-shot low-dose CT image denoising,” arXiv preprint arXiv:2305.15887, 2023, https://arxiv.org/abs/2305.15887

  15. [23]

    Generative modeling in sinogram domain for sparse-view CT reconstruction,

    B. Guan, C. Yang, L. Zhang, S. Niu, M. Zhang, Y . Wang, W. Wu, and Q. Liu, “Generative modeling in sinogram domain for sparse-view CT reconstruction,” IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 8, no. 2, pp. 195–207, 2024

  16. [24]

    A Dual-Domain Diffusion Model for Sparse-View CT Reconstruction,

    C. Yang, D. Sheng, B. Yang, W. Zheng, and C. Liu, “A Dual-Domain Diffusion Model for Sparse-View CT Reconstruction,” IEEE Signal Processing Letters, vol. 31, pp. 1279–1283, 2024

  17. [25]

    Solving 3d inverse problems using pre-trained 2d diffusion models,

    H. Chung, D. Ryu, M. T. McCann, M. L. Klasky, and J. C. Ye, “Solving 3d inverse problems using pre-trained 2d diffusion models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 22 542–22 551

  18. [26]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531, 2015, https://arxiv.org/abs/1503.02531

  19. [27]

    Efficient medical image segmentation based on knowledge distillation,

    D. Qin, J.-J. Bu, Z. Liu, X. Shen, S. Zhou, J.-J. Gu, Z.-H. Wang, L. Wu, and H.-F. Dai, “Efficient medical image segmentation based on knowledge distillation,” IEEE Transactions on Medical Imaging , vol. 40, no. 12, pp. 3820–3831, 2021

  20. [28]

    A self-supervised guided knowledge distillation framework for unpaired low-dose CT image denoising,

    J. Wang, Y . Tang, Z. Wu, Q. Du, L. Yao, X. Yang, M. Li, and J. Zheng, “A self-supervised guided knowledge distillation framework for unpaired low-dose CT image denoising,” Computerized Medical Imaging and Graphics , vol. 107, p. 102237, 2023

  21. [29]

    Deep mutual learning,

    Y . Zhang, T. Xiang, T. M. Hospedales, and H. Lu, “Deep mutual learning,” in Proceedings of The IEEE Conference on Computer Vision and Pattern Recognition , 2018, pp. 4320–4328

  22. [30]

    Incremental cross-view mutual distillation for self-supervised medical CT synthesis,

    C. Fang, L. Wang, D. Zhang, J. Xu, Y . Yuan, and J. Han, “Incremental cross-view mutual distillation for self-supervised medical CT synthesis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 20 677–20 686

  23. [31]

    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

  24. [32]

    DRONE: Dual-domain residual-based optimization network for sparse-view CT reconstruction,

    W. Wu, D. Hu, C. Niu, H. Yu, V . Vardhanabhuti, and G. Wang, “DRONE: Dual-domain residual-based optimization network for sparse-view CT reconstruction,” IEEE Transactions on Medical Imaging , vol. 40, no. 11, pp. 3002–3014, 2021

  25. [33]

    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

  26. [34]

    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

  27. [35]

    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

  28. [36]

    FreeSeed: Frequency- band-aware and self-guided network for sparse-view CT reconstruc- tion,

    C. Ma, Z. Li, J. Zhang, Y . Zhang, and H. Shan, “FreeSeed: Frequency- band-aware and self-guided network for sparse-view CT reconstruc- tion,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2023, pp. 250–259

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.