Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Limited-Angle CBCT Reconstruction via Geometry-Integrated Cycle-domain Denoising Diffusion Probabilistic Models

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Two linked diffusion models reconstruct cone-beam CT volumes from a single 90-degree scan.

desk verdict A sensible dual-domain diffusion pipeline for 90-degree CBCT, but the evaluation split does not support the claim of patient-independent generalization, and the lack of real data and strong baselines makes the clinical claims premature. read the letter →

arxiv 2506.13545 v1 pith:FZ2N7WKC submitted 2025-06-16 cs.CV

Yuan Gao , Shaoyan Pan , Mingzhe Hu , Huiqiao Xie , Jill Remick , Chih-Wei Chang , Justin Roper , Zhen Tian
show 1 more author
Xiaofeng Yang
This is my paper · ORCID
classification cs.CV
keywords cone-beamCTlimited-anglereconstructiondenoisingdiffusionprobabilisticmodelprojectioncompletioncycle-domainconsistencyimage-guidedradiotherapyshort-arcacquisitionHDRbrachytherapy
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

LA-GICD claims that a pair of denoising diffusion models, one that completes missing projection views and one that refines the reconstructed volume, connected by a fixed analytic cone-beam projector and back-projector, can turn a single 90-degree cone-beam CT acquisition into a volume nearly matching a full 360-degree scan. On 18 held-out gynecologic HDR brachytherapy volumes simulated from planning CTs, the pipeline reports a mean absolute error of 35.5 HU, SSIM of 0.84, and PSNR of 29.8 dB, with visibly fewer streak and shading artifacts than the analytic FDK baseline. If this carries over to real scanners, short-arc CBCT would reduce acquisition time and imaging dose four-fold while preserving image quality and requiring no hardware changes. That would make volumetric image guidance practical in time-constrained, collision-prone, or motion-prone radiotherapy settings.

What carries the argument

The load-bearing object is the Geometry Transformation Module (GTM), a fixed, differentiable analytic cone-beam forward- and back-projector using an FDK-style ramp filter, with parameters matched to a gantry-mounted CBCT system. The GTM connects two learned diffusion modules: the Projection-DDPM synthesizes full-view projections conditioned on the limited-angle data, the GTM converts them into an intermediate volume, and the Image-DDPM refines that volume conditioned on the GTM output. A geometry-integrated cycle-domain (GICD) loss enforces agreement between the reconstructed volume, the ground-truth full-angle volume, and the re-projected projections, so every denoising update is anchored to measured rays rather than to an unconstrained image prior.

What would settle it

Acquire a real 90-degree CBCT scan on a gantry-mounted system for a phantom or patient with a known full-arc reference, run LA-GICD as described, and compare against the full-arc reconstruction; if MAE, SSIM, and PSNR depart substantially from the reported values or visible streaks and shading remain, the central claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that limited-angle CBCT reconstruction can be solved by enforcing geometry-aware cycle consistency between projection and image domains. A Projection-DDPM inpaints the missing projections from a 90-degree input; a fixed geometry transformation module back-projects the completed sinogram using exact scanner geometry; an Image-DDPM denoises and sharpens the resulting volume; and a cycle loss re-projects the refined volume and compares it with measured data. The authors claim this closed loop suppresses hallucinated anatomy and preserves soft-tissue boundaries, producing artifact-free, high-contrast reconstructions from a single 90-degree arc with a four-fold reduction in acquisition time and dose.

Load-bearing premise

The central claim rests on simulated projections produced by a CT simulator faithfully matching real cone-beam acquisitions, including scatter, noise, detector response, and patient motion; the same simulator generates the training data, test inputs, and ground truth, so a real-gantry discrepancy would change the reported numbers.

Editorial extensions

If this is right

  • Clinicians could acquire CBCT over a 90-degree arc and reconstruct volumes with reported accuracy of 35.5 HU MAE, 0.84 SSIM, and 29.8 dB PSNR relative to full-arc CBCT, cutting scan time and imaging dose about four-fold without hardware changes.
  • Short-arc acquisition would reduce motion artifacts from respiration, peristalsis, and patient movement, and would avoid gantry collisions in crowded settings such as gynecologic HDR brachytherapy with metallic applicators.
  • Because LA-GICD acts as a software-only post-acquisition reconstruction module on existing flat-panel CBCT systems, it could be deployed across diverse clinical centers without scanner modification.
  • The authors claim the trained model works across patients without retraining in the tested 90-degree configuration, and suggest the dual-domain design could generalize across different angular spans and orientations.

Reading between the lines

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

  • If the simulator-to-real gap is small, the same framework could be trained once on mixed arc lengths and applied to previously unseen arcs such as 60 or 45 degrees, a test the paper flags as future work.
  • The reported uncertainty maps, with low but nonzero variability near bone, soft-tissue boundaries, and applicator tips, suggest clinical deployment should display confidence maps alongside reconstructions rather than treat every voxel as certain.
  • The cycle-domain, geometry-integrated design could transfer to other ill-posed inverse problems with a fixed analytic forward model, such as sparse-view CT or metal artifact reduction, though the paper does not test these settings.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes LA-GICD, a dual-domain diffusion framework for limited-angle cone-beam CT (CBCT) reconstruction, in which a Projection-DDPM completes missing 90-degree projections, an analytic cone-beam forward/back-projector (GTM) maps them to image space, and an Image-DDPM refines the volume. The framework is trained on simulated projections from 78 planning CT volumes of 18 gynecologic HDR brachytherapy patients and evaluated against full-angle reconstructions using MAE, SSIM, and PSNR, plus visual review by four medical physicists. The authors report MAE 35.5 HU, SSIM 0.84, PSNR 29.8 dB and conclude that the method provides artifact-free, clinically applicable reconstructions from a single 90-degree scan, reducing acquisition time and dose four-fold without hardware changes.

Significance. If the results are validated, the proposed dual-domain, geometry-integrated diffusion approach is a potentially useful contribution to limited-angle CBCT reconstruction, addressing a real clinical need in brachytherapy and other time- or geometry-constrained workflows. The paper's strengths include a fully differentiable pipeline, an explicit geometry module, a clear description of the diffusion formulation, and the use of a clinically motivated dataset. However, the current evidence is insufficient to support the paper's central claims of patient-independent generalization, artifact-free reconstruction, and four-fold dose reduction, because the evaluation is compromised by the data-split design and the absence of state-of-the-art baseline comparisons and ablations. The clinical significance claimed in the abstract and conclusion substantially exceeds what the experiments demonstrate.

major comments (4)
  1. [Section 2.2.2] The data split is not patient-independent, and the statement 'No patient overlap exists between the training and testing subsets' is incompatible with the described procedure. The paper states that 60 volumes were randomly chosen for training and one representative volume per patient (18 volumes) was held out for testing, with 78 volumes total from 18 patients. Because each patient contributes 4 or 5 volumes, the 60 training volumes necessarily include additional fractions from the same 18 patients whose one held-out volume forms the test set. Within-patient similarity across treatment fractions (same body habitus, similar applicator geometry, correlated soft-tissue layout) can be memorized, inflating the reported MAE, SSIM, and PSNR. Consequently, Table 1 and the conclusion that LA-GICD works 'without retraining across patients' do not establish generalization to unseen patients. A proper patient-level split (or leave-one-patient-out evaluation) is required.
  2. [Table 1 and Section 3] The quantitative comparison is limited to FDK, a classical analytical baseline. No comparison is made against learned iterative reconstruction methods, post-processing CNNs, transformer-based approaches, or other diffusion-based limited-angle methods discussed in the Introduction. Without such baselines, the claim that the dual-domain, geometry-integrated design yields superior performance is not supported. Furthermore, there is no ablation of the proposed components (Projection-DDPM alone, Image-DDPM alone, GICD cycle losses, or the GTM as a fixed operator). The reader cannot determine which part of the framework drives the reported improvement, and the abstract's attribution of the gains to the GICD strategy remains unsubstantiated.
  3. [Section 2.2.1 and Abstract] All training, validation, and test projections are generated by LEAP-CT simulation from planning CT volumes of 18 patients; no real CBCT acquisitions are used. The central clinical claim—that LA-GICD produces 'artifact-free, high-contrast reconstructions' and reduces dose four-fold in clinical practice—therefore rests on an untested assumption that the simulator faithfully reproduces real gantry behavior including scatter, noise, detector response, and patient motion. The manuscript should either add a real-data validation study (or a domain-shift experiment with real measured projections) or substantially temper the claims about clinical applicability and dose reduction. The four-fold dose reduction is also not measured; it is inferred from the 90-degree arc length, and no dosimetric analysis is provided.
  4. [Figure 5 and Discussion, 'Hallucination' paragraph] The claim that LA-GICD effectively suppresses hallucination is not quantitatively supported. Figure 5 is described in the text as showing variability 'typically less than 0.01 normalized units,' but the caption reports a displayed range of 0–0.06 normalized units; this inconsistency is not explained. No statistical summary, threshold, or comparison against a baseline is given for the uncertainty maps. The term 'artifact-free' in the abstract and conclusion is also inconsistent with the acknowledgment of minor inconsistencies near applicators and bone–soft-tissue interfaces in the Discussion. The wording throughout the paper should be aligned with the actual evidence.
minor comments (6)
  1. [Running title] The running title reads 'Limited-Angel CBCT Reconstruction' instead of 'Limited-Angle'; this typo should be corrected.
  2. [Section 2.1.2] In the first sentence, 'is de designed to generate' contains a typo ('de designed' should be 'designed').
  3. [Figure 5 caption] The caption uses 'LA-GACD' in the first sentence, which is inconsistent with the framework name LA-GICD.
  4. [Equations (5)–(10)] The notation for the Image-DDPM losses is inconsistent: Eq. (7) defines ℒ_Recon-DDPM, but Eq. (10) refers to ℒ_θ_I^μ and ℒ_θ_I^Σ without defining them or relating them to Eq. (7). Please clarify the relationship among these loss terms.
  5. [Section 2.1.5] The cycle-domain loss in Eq. (10) uses weights γ1=0.05, γ2=0.5, and γ3=0.5 described as 'empirically set,' but no sensitivity analysis or justification is provided. A brief rationale or a supplementary experiment varying these weights would strengthen the presentation.
  6. [Section 3, Table 1] The paired t-test is reported only as p < 0.01 for each metric; the number of paired samples (18 test volumes) should be stated in the table or text, and the test should be described precisely (e.g., two-sided paired t-test) to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; reported reconstructions are not forced by construction, though simulator fidelity and data-split consistency are separate concerns.

full rationale

The derivation chain is self-contained. Projection-DDPM and Image-DDPM use the standard DDPM objectives (Eqs. 4 and 7), the GTM is a fixed analytic forward/back-projector (Section 2.1.4) with parameters taken from the scanner geometry, and final supervision is against original full-scan CT volumes (Eq. 9). The limited-angle input, synthesized full-view sinogram, GTM backprojection, and refined volume are distinct quantities; no reported metric is a fitted parameter renamed as a prediction, and no equation-level reduction equates the target to the input. The self-citations to CG-DDPM [21] and PC-DDPM [22] in Section 2.1.1 only state design lineage and are not load-bearing. Two non-circular concerns remain. First, Section 2.2.2 says a 'volume-level split' with 60 training and 18 test volumes was used 'to ensure patient-level independence' and that 'no patient overlap exists,' but 78 volumes from 18 patients make those statements mutually inconsistent; this threatens generalization claims, not circularity. Second, training projections, inference inputs, and the GTM all derive from the LEAP-CT simulator, so transfer to real CBCT acquisitions is unvalidated; this is a simulator-fidelity limitation, not a definitional circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim does not introduce new physical entities, but it leans on a closed simulation loop: the same LEAP-CT simulator generates training projections, ground-truth reconstructions, and the GTM operator used at inference. The model weights and loss weights are fitted quantities, not derived constants, so the clinical transfer rests entirely on the domain assumption that the simulation matches real scanners.

free parameters (3)
  • Trained weights of Projection-DDPM and Image-DDPM = not released
    Learned from 60 training volumes; the central reconstruction performance depends entirely on them. The checkpoint is not provided.
  • GICD loss weights gamma1, gamma2, gamma3 = 0.05, 0.5, 0.5
    Set empirically (Section 2.1.5) and not ablated; the final reconstruction quality depends on these choices.
  • Diffusion timesteps and inference schedule = 1000 training steps, 50 inference steps
    Chosen cosine schedule and DDIM-style 50-step sampling; the paper shows no sensitivity analysis for these choices.
assumptions (4)
  • domain assumption LEAP-CT analytic forward/backprojector accurately simulates clinical CBCT acquisition
    All training projections and ground truth volumes are generated by LEAP-CT from planning CTs (Section 2.2.1); transfer to real scanners is unverified.
  • domain assumption FDK reconstruction of full-angle simulated projections is a valid ground-truth CBCT
    The reference images are FDK reconstructions of the same simulator, not clinical gold-standard CBCT (Section 2.2 and Table 1).
  • standard math DDPM reverse process can learn conditional image distribution with a U-Net denoiser
    Equations 1-7 rely on standard DDPM theory (Ho et al., ref 23); accepted as background.
  • domain assumption Training on 60 pelvic HDR volumes generalizes to unseen patients and geometries
    Only 18 patients from one anatomical site are used; no multi-site or multi-scanner data are included (Section 4 limitations).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Limited-Angle CBCT Reconstruction via Geometry-Integrated Cycle-domain Denoising Diffusion Probabilistic Models." pith.science (2026). https://pith.science/paper/FZ2N7WKC

@misc{pith2026250613545,
  author       = {Pith},
  title        = {Pith review of: Limited-Angle CBCT Reconstruction via Geometry-Integrated Cycle-domain Denoising Diffusion Probabilistic Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FZ2N7WKC}},
  note         = {Machine review of arXiv:2506.13545}
}
read the original abstract

Cone-beam CT (CBCT) is widely used in clinical radiotherapy for image-guided treatment, improving setup accuracy, adaptive planning, and motion management. However, slow gantry rotation limits performance by introducing motion artifacts, blurring, and increased dose. This work aims to develop a clinically feasible method for reconstructing high-quality CBCT volumes from consecutive limited-angle acquisitions, addressing imaging challenges in time- or dose-constrained settings. We propose a limited-angle (LA) geometry-integrated cycle-domain (LA-GICD) framework for CBCT reconstruction, comprising two denoising diffusion probabilistic models (DDPMs) connected via analytic cone-beam forward and back projectors. A Projection-DDPM completes missing projections, followed by back-projection, and an Image-DDPM refines the volume. This dual-domain design leverages complementary priors from projection and image spaces to achieve high-quality reconstructions from limited-angle (<= 90 degrees) scans. Performance was evaluated against full-angle reconstruction. Four board-certified medical physicists conducted assessments. A total of 78 planning CTs in common CBCT geometries were used for training and evaluation. The method achieved a mean absolute error of 35.5 HU, SSIM of 0.84, and PSNR of 29.8 dB, with visibly reduced artifacts and improved soft-tissue clarity. LA-GICD's geometry-aware dual-domain learning, embedded in analytic forward/backward operators, enabled artifact-free, high-contrast reconstructions from a single 90-degree scan, reducing acquisition time and dose four-fold. LA-GICD improves limited-angle CBCT reconstruction with strong data fidelity and anatomical realism. It offers a practical solution for short-arc acquisitions, enhancing CBCT use in radiotherapy by providing clinically applicable images with reduced scan time and dose for more accurate, personalized treatments.

Review history (2 revisions) →

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. Tooth-Diffusion: Guided 3D CBCT Synthesis with Fine-Grained Tooth Conditioning

    cs.CV 2025-08 conditional novelty 6.0 of 10

    A conditional wavelet diffusion model with FiLM conditioning generates and edits 3D CBCT volumes from per-tooth presence vectors.

Reference graph

Works this paper leans on

37 extracted references · 30 canonical work pages · cited by 1 Pith paper

  1. [1]

    geometry -integrated

    Introduction Cone-beam computed tomography (CBCT) has become an essential imaging modality in modern radiation therapy. Mounted on linear accelerator and C -arm gantries, flat -panel CBCT systems provide in -room volumetric imaging that enables sub-millimeter patient setup verification, adaptive replanning based on the anatomy-of-the-day (intraoperative i...

    work page

  2. [2]

    Materials and Methods 2.1 LA -GICD framework

    work page

  3. [3]

    1. 1. Overview We propose a novel deep generative framework, named LA-GICD, for high-fidelity CBCT reconstruction from limited -angle projections. The proposed architecture mitigates the ill -posedness of limited -angle reconstruction by integrating data-driven priors with explicit geometric consistency constraints. The pipeline contains three fully diffe...

    work page

  4. [4]

    2 Projection -DDPM The Projection-DDPM module is de designed to generate a complete set of synthetic projections 𝑃̂1:360∘ from limited-angle projections 𝑃̂135∘:225∘

    1. 2 Projection -DDPM The Projection-DDPM module is de designed to generate a complete set of synthetic projections 𝑃̂1:360∘ from limited-angle projections 𝑃̂135∘:225∘. This task is formulated as a conditional generative process within the DDPM framework. The model learns to map from Gaussian noise to high-fidelity, full-view projections conditioned on th...

    work page

  5. [5]

    3 Image -DDPM The second DDPM module is Image-DDPM module ( Image-DDPM shown in Figure 1

    1. 3 Image -DDPM The second DDPM module is Image-DDPM module ( Image-DDPM shown in Figure 1. (c)), which is designed to refine intermediate CBCT volumes reconstructed from synthesized projections, yielding high- fidelity CBCT images with enhanced anatomical consistency. Let 𝐼𝑡 syn denote the noisy volumetric CBCT at diffusion timestep 𝑡, and let 𝐼𝑡 det be...

    work page

  6. [6]

    1. 5 GICD Strategy: Geometry -Integrated Cycle -Domain Supervision The GICD supervision strategy integrates physical acquisition geometry and bidirectional consistency to guide training under limited -angle settings. It consists of three core components: Geometry-Conditioned Forward Path . Geometry-Conditioned reconstruction, the Projection -DDPM generate...

    work page

  7. [7]

    1. 7. Implementation Detail All models were implemented using PyTorch and trained on a single NVIDIA A100 GPU with 80 GB memory. Both Projection -DDPM and Image-DDPM were trained using 1000 diffusion steps during learning and sampled using 50 inference steps at test time. A cosine-based noise schedule [β𝑡]𝑡=1 𝑇 was used, with cumulative noise factor α𝑡̅̅̅...

    work page

  8. [8]

    1 Dataset description We studied 18 anonymized patients who received gynecologic HDR brachytherapy

    2. 1 Dataset description We studied 18 anonymized patients who received gynecologic HDR brachytherapy. The dataset comprises a total of 78 CT volumes, acquired from 18 patients who received 4 or 5 CBCT scans during the course of treatments. Each case provided a 512 × 512 × 256 pelvic CBCT volume with voxel spacing of 1 × 1 × 2 mm in native resolution; we ...

    work page

Show all 37 references
  1. [9]

    As shown in Table 1, the proposed method demonstrates strong reconstruction performance under severe angular limitations

    Quantitative and qualitative evaluation We evaluated LA-GICD on a limited-angle CBCT dataset using three standard metrics: MAE, SSIM, and PSNR. As shown in Table 1, the proposed method demonstrates strong reconstruction performance under severe angular limitations. These resul...

  2. [10]

    However, reconstructing high - quality volumes from such limited data remains a major challenge

    Discussion Limited-angle CBCT has the potential to significantly improve clinical workflows by reducing scan time, minimizing motion artifacts, and avoiding physical gantry collisions—particularly in RT settings such as equipment packed HDR rooms, intra-fractional motion manag...

  3. [11]

    The method reconstructs high - quality CBCT volumes from a 90° short-arc projection data, without retraining across angles or patients

    Conclusion We proposed a new LA-GICD framework for LA CBCT reconstruction. The method reconstructs high - quality CBCT volumes from a 90° short-arc projection data, without retraining across angles or patients. Tested on gynecologic HDR datasets, LA-GICD achieved strong quanti...

  4. [12]

    Image -guided radiotherapy: from current concept to future perspectives

    Jaffray DA. Image -guided radiotherapy: from current concept to future perspectives. Nat Rev Clin Oncol. Dec 2012;9(12):688-99. doi:10.1038/nrclinonc.2012.194

    2012 doi

  5. [13]

    Mitigation of motion-induced artifacts in cone beam computed tomography using deep convolutional neural networks

    Amirian M, Montoya-Zegarra JA, Herzig I, et al. Mitigation of motion-induced artifacts in cone beam computed tomography using deep convolutional neural networks. Med Phys. Oct 2023;50(10):6228-

    2023

  6. [14]

    Limited-Angle Computed Tomography Reconstruction using Combined FDK-Based Neural Network and U-Net

    Wang Y , Yang T, Huang W. Limited-Angle Computed Tomography Reconstruction using Combined FDK-Based Neural Network and U-Net. 2020:1572-1575

    2020

  7. [15]

    Cone beam computed tomography -based monitoring and management of target and organ motion during external beam radiotherapy in cervical cancer

    Jensen NBK, Assenholt MS, Fokdal LU, et al. Cone beam computed tomography -based monitoring and management of target and organ motion during external beam radiotherapy in cervical cancer. Phys Imaging Radiat Oncol. Jan 2019;9:14-20. doi:10.1016/j.phro.2018.12.002

    2019 doi

  8. [16]

    Calculating tumor trajectory and dose‐of‐the‐day using cone‐beam CT projections

    Jones BL, Westerly D, Miften M. Calculating tumor trajectory and dose‐of‐the‐day using cone‐beam CT projections. Medical physics. 2015;42(2):694-702

    2015

  9. [17]

    Evaluation of the positional reproducibility of sedation versus non-sedation state in pediatric radiotherapy: a retrospective study

    Lei Z, Shi Y , Liu Y , et al. Evaluation of the positional reproducibility of sedation versus non-sedation state in pediatric radiotherapy: a retrospective study. Original Research. Frontiers in Oncology. 2024- October-29 2024;V olume 14 - 2024doi:10.3389/fonc.2024.1475060

    2024

  10. [18]

    An open-source platform for interactive collision prevention in photon and particle beam therapy treatment planning

    Hueso-González F, Wohlfahrt P, Craft D, Remillard K. An open-source platform for interactive collision prevention in photon and particle beam therapy treatment planning. Biomedical Physics & Engineering Express. 2020;6(5):055013

    2020

  11. [19]

    Practical cone -beam algorithm

    Feldkamp LA, Davis LC, Kress JW. Practical cone -beam algorithm. J Opt Soc Am A . 1984/06/01 1984;1(6):612-619. doi:10.1364/JOSAA.1.000612

    1984 doi

  12. [20]

    Limited-angle computed tomography with deep image and physics priors

    Barutcu S, Aslan S, Katsaggelos AK, Gürsoy D. Limited-angle computed tomography with deep image and physics priors. Scientific Reports. 2021/09/06 2021;11(1):17740. doi:10.1038/s41598-021-97226- 2

    2021 doi

  13. [21]

    Improved total variation -based CT image reconstruction applied to clinical data

    Ritschl L, Bergner F, Fleischmann C, Kachelrieß M. Improved total variation -based CT image reconstruction applied to clinical data. Physics in Medicine & Biology. 2011;56(6):1545

    2011

  14. [22]

    Deep learning image reconstruction for improving image quality of contrast -enhanced dual -energy CT in abdomen

    Sato M, Ichikawa Y , Domae K, et al. Deep learning image reconstruction for improving image quality of contrast -enhanced dual -energy CT in abdomen. European Radiology . 2022;32(8):5499 -5507. doi:10.1007/s00330-022-08647-0

    2022 doi

  15. [23]

    On Hallucinations in Tomographic Image Reconstruction

    Bhadra S, Kelkar V A, Brooks FJ, Anastasio MA. On Hallucinations in Tomographic Image Reconstruction. IEEE Trans Med Imaging . Nov 2021;40(11):3249 -3260. doi:10.1109/tmi.2021.3077857

    2021

  16. [24]

    Data consistent artifact reduction for limited angle tomography with deep learning prior

    Huang Y , Preuhs A, Lauritsch G, Manhart M, Huang X, Maier A. Data consistent artifact reduction for limited angle tomography with deep learning prior. Springer; 2019:101-112

    2019

  17. [25]

    A deep convolutional neural network using directional wavelets for low ‐dose X‐ray CT reconstruction

    Kang E, Min J, Ye JC. A deep convolutional neural network using directional wavelets for low ‐dose X‐ray CT reconstruction. Medical Physics. 2017;44(10):e360-e375. doi:10.1002/mp.12344

    2017 doi

  18. [27]

    Learned primal -dual reconstruction

    Adler J, Öktem O. Learned primal -dual reconstruction. IEEE transactions on medical imaging . 2018;37(6):1322-1332

    2018

  19. [28]

    Deep Learning Computed Tomography: Learning Projection- Domain Weights From Image Domain in Limited Angle Problems

    Würfl T, Hoffmann M, Christlein V , et al. Deep Learning Computed Tomography: Learning Projection- Domain Weights From Image Domain in Limited Angle Problems. IEEE Transactions on Medical Imaging. 2018;37(6):1454-1463. doi:10.1109/TMI.2018.2833499

    2018

  20. [29]

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

    Wu W, Hu D, Niu C, Y u H, Vardhanabhuti V , Wang G. DRONE: Dual -domain residual -based optimization network for sparse -view CT reconstruction. IEEE Transactions on Medical Imaging . 2021;40(11):3002-3014

    2021

  21. [30]

    Hybrid U -Net and Swin -transformer network for limited-angle cardiac computed tomography

    Xu Y , Han S, Wang D, Wang G, Maltz JS, Yu H. Hybrid U -Net and Swin -transformer network for limited-angle cardiac computed tomography. Phys Med Biol . Apr 30 2024;69(10)doi:10.1088/1361 - 6560/ad3db9

    2024 doi

  22. [31]

    Dolce: A model -based probabilistic diffusion framework for limited-angle ct reconstruction

    Liu J, Anirudh R, Thiagarajan JJ, et al. Dolce: A model -based probabilistic diffusion framework for limited-angle ct reconstruction. 2023:10498-10508

    2023

  23. [32]

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

    Yang C, Sheng D, Yang B, Zheng W, Liu C. A Dual -Domain Diffusion Model for Sparse -View CT Reconstruction. IEEE Signal Processing Letters. 2024;31:1279-1283. doi:10.1109/LSP.2024.3392690

    2024

  24. [33]

    CBCT Reconstruction using Single X-ray Projection with Cycle-domain Geometry-integrated Denoising Diffusion Probabilistic Models

    Pan S, Peng J, Gao Y , et al. CBCT Reconstruction using Single X-ray Projection with Cycle-domain Geometry-integrated Denoising Diffusion Probabilistic Models. IEEE Transactions on Medical Imaging. 2025

    2025

  25. [34]

    Patient-Specific CBCT Synthesis for Real-time Tumor Tracking in Surface- guided Radiotherapy

    Pan S, Su V , Peng J, et al. Patient-Specific CBCT Synthesis for Real-time Tumor Tracking in Surface- guided Radiotherapy. ArXiv. Nov 1 2024

    2024

  26. [35]

    Denoising diffusion probabilistic models

    Ho J, Jain A, Abbeel P . Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems. 2020;33:6840-6851

    2020

  27. [36]

    Differentiable forward projector for X-ray computed tomography

    Kim H, Champley K. Differentiable forward projector for X-ray computed tomography. arXiv preprint arXiv:230705801. 2023

    2023

  28. [37]

    Physics -informed Score -based Diffusion Model for Limited -angle Reconstruction of Cardiac Computed Tomography

    Han S, Xu Y , Wang D, et al. Physics -informed Score -based Diffusion Model for Limited -angle Reconstruction of Cardiac Computed Tomography. IEEE Trans Med Imaging . Nov 8 2024;PPdoi:10.1109/TMI.2024.3494271

    2024

  29. [6242]

    doi:10.1002/mp.16405

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.