Pith. sign in

REVIEW 4 major objections 6 minor 30 references

Energy-Based Prior Latent Space Diffusion model for Reconstruction of Lumbar Vertebrae from Thick Slice MRI

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A latent-space diffusion energy-based model reconstructs lumbar vertebrae from thick-slice MRI more faithfully than VAE and latent-EBM baselines, achieving DSC 0.8304 and VS 0.8627 against CT ground truth.

desk verdict A useful empirical comparison undermined by a broken theoretical derivation and a sign error in the training objective. read the letter →

arxiv 2412.00511 v1 pith:64Z6UN5X submitted 2024-11-30 eess.IV cs.AIcs.CV

Yanke Wang , Yolanne Y. R. Lee , Aurelio Dolfini , Markus Reischl , Ender Konukoglu , Kyriakos Flouris This is my paper · ORCID
classification eess.IVcs.AIcs.CV
keywords latentspacediffusionenergy-basedmodelthicksliceMRIlumbarvertebraereconstructionvariationalautoencoderDicesimilaritycoefficient3Dmedicalimagegenerationprobabilistic
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

The paper is trying to establish that a generative model trained on high-quality CT vertebrae segmentations can recover fine 3D anatomy from coarse segmentations derived from thick-slice MRI. To do this cheaply, it runs denoising diffusion not on the whole 3D image but on a learned latent code, and makes each reverse diffusion step an energy-based model. The authors report that this design, which they call LSD-EBM, beats a VAE and a latent-space energy-based model on Dice and volumetric similarity against CT ground truth, and that its reconstructions retain detail across sampling steps. If this holds, thick-slice MRI plus a learned prior could deliver CT-like bone models without ionizing radiation.

What carries the argument

The engine is a latent-space diffusion chain whose reverse denoising steps are Boltzmann distributions: the encoder produces $z_0 \sim q_\varphi(z_0|x)$, a forward chain adds Gaussian noise $q(z_{t+1}|z_t)=\mathcal{N}(z_{t+1};\sqrt{1-\sigma_{t+1}^2}z_t,\sigma_{t+1}^2 I)$, and each reverse transition $p_\alpha(z_t|z_{t+1})$ is proportional to $\exp(-E_\alpha(z_t,t)-\|z_{t+1}-z_t\|^2/(2\sigma_{t+1}^2))$, sampled with Langevin dynamics. The time-indexed energy function and the denoising structure are what stabilize sampling and keep the model trainable on $128^3$ volumes in 17 hours, where a full-image DDPM exceeds 40 GB of GPU memory after two steps.

What would settle it

Evaluate Eq. (15) on a simple tractable case, for instance a quadratic energy $E_\alpha(z,t)=\|z\|^2/2$, and compare $\log \tilde{Z}_\alpha$ against the expectation term under $p_\alpha$; any mismatch invalidates the claimed ELBO. Separately, run the trained pipeline on paired thin-slice MRI and CT of the same vertebrae and check whether the reported DSC and VS gains persist outside the specific L-MRI test set.

Watch

Extended reading notes

Core claim

The paper claims that performing denoising diffusion in the latent space of an autoencoder, with each reverse transition defined by a conditional energy-based model, yields lumbar-vertebra reconstructions from thick-slice MRI that are closer to high-quality CT ground truth than the two comparators. On the 80-volume test set, LSD-EBM attains DSC 0.8304 and VS 0.8627, against 0.7626/0.7887 for the VAE and 0.7619/0.7866 for the LEBM, and also higher normalized mutual information and Cohen's kappa. The authors interpret these scores, together with qualitative close-ups, as evidence that the latent energy-based diffusion process preserves fine 3D anatomical detail instead of smoothing it away.

Load-bearing premise

The whole training objective rests on Appendix C.1's claim that replacing the log-partition function $\log \tilde{Z}_\alpha$ with an expectation under the model's own distribution $p_\alpha$ is a valid ELBO step; the identity is asserted without proof, and if it fails the gradient used to train the energy function is not the gradient of a true lower bound.

Editorial extensions

If this is right

  • If correct, thick-slice MRI post-processing can recover vertebral detail that approaches CT quality without exposing patients to ionizing radiation.
  • The latent-space diffusion design brings DDPM-like fidelity within a feasible budget: 17 hours of training and 6.25 seconds per-sample reconstruction on the vertebrae data, where full-image DDPM was infeasible on a 40 GB GPU.
  • LSD-EBM retains detail consistently across 2, 15, and 20 sampling steps, unlike the LEBM, so it can be used efficiently with fewer denoising steps.
  • The framework also improves FID scores on MNIST, FashionMNIST, and CelebA relative to the EBM and LEBM baselines, suggesting the method generalizes beyond vertebrae.
  • The reported higher NMI and Cohen's kappa indicate reconstructions share more information with the CT ground truth than the baselines, not merely higher overlap.

Reading between the lines

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

  • If the energy-update step in Eq. (15) is repaired or shown to be correct, the same latent conditional-EBM diffusion recipe should transfer to other large 3D medical structures where full-image diffusion is memory-prohibitive, such as cardiac or brain volumes.
  • The reported test set is 80 vertebrae, so per-level (L1-L5) and per-acquisition-protocol breakdowns would tell whether the DSC gain is uniform or driven by particular anatomies.
  • Because the prior is trained on CT segmentations, the method's viability in clinical MRI workflows still depends on how far MRI contrast and artefact distributions can shift before reconstruction fidelity drops; paired thin-slice MRI would provide the direct test.
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 LSD-EBM, a latent-space diffusion model with a conditional energy-based prior, as a replacement for the VAE post-processing step in an existing pipeline for reconstructing lumbar vertebrae from thick-slice MRI. The model encodes a 3D input into a latent variable, runs a diffusion process in latent space with an energy-based prior sampled via Langevin dynamics, and decodes the refined latent back into a segmentation. The authors report that LSD-EBM outperforms VAE and LEBM baselines on Dice and volumetric similarity in an 80-sample paired MRI/CT test set, and also report favorable FID scores on standard 2D datasets. The paper includes derivations of an ELBO objective in Appendix C.1 and gives training and inference algorithms in Appendix C.2.

Significance. If the method and its theoretical justification were sound, this would be a practically useful contribution: it addresses a real clinical imaging problem (thick-slice MRI reconstruction), proposes a computationally feasible latent-space diffusion energy-based model for 3D medical data, and provides code and quantitative comparisons on both vertebrae and standard 2D datasets. The reported improvements in DSC (0.8304 vs 0.7626/0.7619) and VS are large and internally consistent across 80 test samples, and the 2D FID results in Appendix F support the general generative capability of the architecture. However, the significance is substantially undercut by serious flaws in the theoretical derivation and in the stated training algorithm, which call into question whether the reported results are produced by the mechanism the paper claims.

major comments (4)
  1. [Appendix C.1, Eq. (15)] The identity used to eliminate the log-partition function is false. Equation (15) replaces log Z̃_α(z_{t+1}, t+1) with E_{pα(zt|zt+1)}[-Eα(zt,t) - (1/2σ²)||z_{t+1}-z_t||²]. In general, for any density p(e) = exp(f(e))/Z, one has E_p[f] = log Z - H(p), where H(p) is the differential entropy. The equality in Eq. (15) therefore holds only if the conditional density is degenerate, which is not the case. Consequently, the quantity called L(α, φ, β) in Eq. (16) is not an evidence lower bound, and the gradient in Eq. (17) is not the gradient of any ELBO. This is a load-bearing error: the theoretical grounding of the energy-prior training step collapses unless a corrected derivation is provided.
  2. [Algorithm 1, energy loss update] The sign of the energy loss in Algorithm 1 is reversed relative to maximum-likelihood training of a Boltzmann distribution. The algorithm states "Update α by minimizing the energy loss -Eα(zt,t) - (-Eα(˜zt,t))", i.e. minimize -E(data) + E(sample). For pα ∝ exp(-Eα), the maximum-likelihood gradient is E_data[∇_α Eα] - E_p[∇_α Eα], which corresponds to minimizing E(data) - E(sample). The stated update minimizes the negative of this objective, increasing energy at data latents and decreasing it at MCMC samples, the opposite of what the paper claims. If the released code implements the stated sign, the reported performance gain cannot be attributed to a learned energy prior; if the code uses the correct sign, the paper substantially misreports the algorithm. Either way, the paper must be corrected and the experiments reconciled with the actual training objective.
  3. [Section 3, Eq. (2) vs Eq. (4)] The definition of the reverse process is internally inconsistent. Equation (2) defines pα(˜zt|zt+1) with ˜zt = sqrt(1-σ²_{t+1}) z_t and the quadratic term ||z_{t+1}-˜zt||², but the text then states that "in practice pα(zt|zt+1) can be used instead" and Eq. (4) writes the gradient as -∇_z Eα(zt,t) + (1/σ²_{t+1})(z_{t+1}-z_t). This change of variables is not derived; if z_t is the latent before the scaling, the energy should be evaluated at the scaled argument and the quadratic term should contain the scaled variable. As written, the sampling distribution used for MCMC is not well defined, and this ambiguity compounds the error in Eq. (15).
  4. [Section 4.2, Table 1] The central empirical claim that LSD-EBM "outperforms" VAE and LEBM is not supported by any statistical significance testing. The table reports means and standard deviations over 80 samples, but no confidence intervals, paired tests, or effect-size measures are given. Although the reported DSC differences appear large relative to the standard deviations, a significance test is necessary to support the claim, especially because the same test set is used to select and evaluate models. The paper should report paired tests across the 80 samples for DSC, VS, and the other metrics.
minor comments (6)
  1. [Appendix C.1, Eq. (15)] Equation (15) contains a typesetting inconsistency: the first line uses ||z_{t+1} - z_t||² in the quadratic term, while the second line uses ||z_{t+1} - ˜zt||². This should be fixed and aligned with the explicit definition of the conditional distribution.
  2. [Section 3] The phrase "Evidence-based Lower BOund" in Section 3 should read "Evidence Lower Bound".
  3. [Fig. 6 caption] The caption of Fig. 6 appears to contain a long stray paragraph that is duplicated from the appendix text and is not a proper caption. This should be removed or rewritten.
  4. [References] Reference [22] is incomplete: it lists only "Turner, R.: (2005)" with no title, venue, or other bibliographic information.
  5. [Appendix F] The first paragraph of Appendix F says the models were trained on "MNIST, CIFAR10, and CelebA", while the preceding sentence mentions FashionMNIST as well. The dataset list should be made consistent.
  6. [Fig. 5] The axis label "2(z)" in Fig. 5 appears to be a typo for the variance σ²(z); it should be clarified to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical DSC/VS comparison is benchmarked on held-out data; self-citation supplies dataset, not conclusion.

full rationale

The central claim, that LSD-EBM outperforms VAE and LEBM on DSC and VS for lumbar vertebrae reconstruction, is an empirical benchmark claim. The models are trained on CT-Train and evaluated on 80 held-out paired L-MRI/H-CT samples from the Turella et al. MICCAI 2021 pipeline; the reported metrics are measured on a separate test set and are not fitted parameters renamed as predictions. The only author-overlap citations are [21] (Konukoglu is a co-author) and [8] (Flouris and Konukoglu). [21] is cited for the segmentation/reconstruction pipeline and the datasets, which is a legitimate external benchmark source rather than a load-bearing self-citation; the paper does not invoke any uniqueness theorem from its own authors to forbid alternatives. [8] is a passing reference to normalizing flows and is not load-bearing. The suspicious mathematical step in Appendix C.1, Eq. (15), replaces log \tilde Z with an expectation under p_alpha without accounting for the entropy term, and Algorithm 1 appears to use a contrastive-divergence-style energy loss with a sign that would train the energy in the wrong direction. These are correctness and reproducibility concerns, not circularity: they do not make the empirical result equal to its input by construction, and they do not turn a fitted parameter into a prediction. No known result is renamed or smuggled in via citation; prior latent diffusion energy-based work [25,9] is explicitly credited and differentiated. Therefore no significant circularity is present; the score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper's central training objective depends on a derived ELBO that includes an unproven identity for the log-partition function. The remaining assumptions are standard diffusion and domain transfer assumptions.

free parameters (4)
  • Noise schedule sigma_t
    Fixed schedule of Gaussian noise levels for the latent diffusion chain, chosen following DDPM conventions (Eq. 1).
  • MCMC steps K = 20
    Number of Langevin iterations per diffusion step, chosen as 20 during training and inference (Appendix E).
  • Diffusion steps T = 20
    Number of diffusion/denoising steps in the latent space, tested at 2, 15, 20 (Section 4.2).
  • Learning rates and batch sizes = 2e-5, 1e-4, 2e-5 for VAE/LEBM/LSD-EBM; batch 4, 2, 4
    Training hyperparameters in Appendix E.
assumptions (5)
  • standard math q(z_{t+1}|z_t) = N(z_{t+1}; sqrt(1-sigma^2_{t+1}) z_t, sigma^2_{t+1} I) defines a valid forward Markov chain.
    Sohl-Dickstein et al. Eq. (11).
  • standard math p_alpha(\tilde z_t | z_{t+1}) follows the Boltzmann distribution in Eq. (2) with a finite partition function.
    Definition of the conditional energy-based prior.
  • ad hoc to paper The identity in Eq. (15) relating -log Z to an expectation under p_alpha holds.
    Unproven and appears incorrect as written; the log of a partition function is not generally equal to an expectation of the energy under the model.
  • domain assumption CT-Train and L-MRI/H-CT datasets are representative and correctly paired.
    The model is trained on CT segmentations and evaluated on MRI-derived segmentations from the same pipeline [21].
  • domain assumption The noise schedule and K steps allow MCMC sampling to converge in practice.
    The paper relies on K=20 finite steps without convergence guarantees (Sec. 3, Eq. 3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Energy-Based Prior Latent Space Diffusion model for Reconstruction of Lumbar Vertebrae from Thick Slice MRI." pith.science (2026). https://pith.science/paper/64Z6UN5X

@misc{pith2026241200511,
  author       = {Pith},
  title        = {Pith review of: Energy-Based Prior Latent Space Diffusion model for Reconstruction of Lumbar Vertebrae from Thick Slice MRI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/64Z6UN5X}},
  note         = {Machine review of arXiv:2412.00511}
}
read the original abstract

Lumbar spine problems are ubiquitous, motivating research into targeted imaging for treatment planning and guided interventions. While high resolution and high contrast CT has been the modality of choice, MRI can capture both bone and soft tissue without the ionizing radiation of CT albeit longer acquisition time. The critical trade-off between contrast quality and acquisition time has motivated 'thick slice MRI', which prioritises faster imaging with high in-plane resolution but variable contrast and low through-plane resolution. We investigate a recently developed post-acquisition pipeline which segments vertebrae from thick-slice acquisitions and uses a variational autoencoder to enhance quality after an initial 3D reconstruction. We instead propose a latent space diffusion energy-based prior to leverage diffusion models, which exhibit high-quality image generation. Crucially, we mitigate their high computational cost and low sample efficiency by learning an energy-based latent representation to perform the diffusion processes. Our resulting method outperforms existing approaches across metrics including Dice and VS scores, and more faithfully captures 3D features.

Figures

Figures reproduced from arXiv: 2412.00511 by the authors.

Figure 1
Figure 1. Schematic diagram of the segmentation and reconstruction of high-quality [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The schematic diagram of our network structure and proposed LSD-EBM. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of VAE, LEBM, and LSD-EBM reconstructions of the L3 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: The visualization of VAE, LEBM, and LSD-EBM on the reconstruction [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: The mean variance of the latent variables ( [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 8
Figure 8. Figure 8: Sketches of different methods with focus on the processes in Fig. 6: Schematics of different existing methods, with focus on the processes in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 7
Figure 7. Figure 7: Examples of generated images by the models (EBM, LEBM, and LSD [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: Examples of generated images by the models (EBM, LEBM, and LSD [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Examples of generated images by the models (EBM, LEBM, and LSD [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]
Figure 10
Figure 10. Figure 10: Reconstructions of the input (in yellow) using LEBM, VAE, and [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: Reconstructions of the input (in yellow) using LEBM, VAE, and [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Reconstructions of the input (in yellow) using LEBM, VAE, and [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Reconstructions of the input (in yellow) using LEBM, VAE, and [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 25 canonical work pages

  1. [1]

    Cognitive Science9(1), 147–169 (1985)

    Ackley, D.H., Hinton, G.E., Sejnowski, T.J.: A learning algorithm for boltz- mann machines. Cognitive Science9(1), 147–169 (1985). https://doi.org/https:// doi.org/10.1016/S0364-0213(85)80012-4, https://www.sciencedirect.com/science/ article/pii/S0364021385800124

    doi:10.1016/s0364-0213(85)80012-4 1985

  2. [2]

    In: Inter- national Conference on Medical Imaging with Deep Learning

    Amiranashvili, T., Lüdke, D., Li, H.B., Menze, B., Zachow, S.: Learning shape reconstruction from sparse measurements with neural implicit functions. In: Inter- national Conference on Medical Imaging with Deep Learning. pp. 22–34. PMLR (2022)

    work page 2022

  3. [3]

    In: Digital Image Computing: Techniques and Applications (DICTA)

    Bajger, M., To, M.S., Lee, G., Wells, A., Chong, C., Agzarian, M., Poonnoose, S.: Lumbar spine CT synthesis from MR images using CycleGAN-a preliminary study. In: Digital Image Computing: Techniques and Applications (DICTA). pp. 1–

    work page

  4. [4]

    Spine (Philadelphia, Pa

    Been, E., Barash, A., Pessah, H., Peleg, S.: A new look at the geometry of the lumbar spine. Spine (Philadelphia, Pa. : 1986)35(20), E1014–E1017 (2010)

    work page 2010

  5. [5]

    In: ECML PKDD Workshop: Languages for Data Mining and Machine Learning

    Buitinck, L., Louppe, G., Blondel, M., Pedregosa, F., Mueller, A., Grisel, O., Nic- ulae, V., Prettenhofer, P., Gramfort, A., Grobler, J., Layton, R., VanderPlas, J., Joly, A., Holt, B., Varoquaux, G.: API design for machine learning software: Expe- riences from the scikit-learn project. In: ECML PKDD Workshop: Languages for Data Mining and Machine Learni...

    work page 2013

  6. [6]

    IEEE Access : practical innovations, open solutions8, 83858–83870 (2020) 10 Y

    Chai, Y., Xu, B., Zhang, K., Lepore, N., Wood, J.C.: MRI restoration using edge- guided adversarial learning. IEEE Access : practical innovations, open solutions8, 83858–83870 (2020) 10 Y. Wang et al

    work page 2020

  7. [7]

    In: Wallach, H., Larochelle, H., Beygelzimer, A., dAlché-Buc, F., Fox, E., Garnett, R

    Du, Y., Mordatch, I.: Implicit generation and modeling with energy based models. In: Wallach, H., Larochelle, H., Beygelzimer, A., dAlché-Buc, F., Fox, E., Garnett, R. (eds.) Advances in Neural Information Processing Systems. vol. 32. Curran Associates, Inc. (2019)

    work page 2019

  8. [8]

    In: Oh, A., Neumann, T., Globerson, A., Saenko, K., Hardt, M., Levine, S

    Flouris, K., Konukoglu, E.: Canonical normalizing flows for manifold learning. In: Oh, A., Neumann, T., Globerson, A., Saenko, K., Hardt, M., Levine, S. (eds.) Advances in Neural Information Processing Systems. vol. 36, pp. 27294–27314. Curran Associates, Inc. (2023), https://proceedings.neurips.cc/paper_files/paper/ 2023/file/572a6f16ec44f794fb3e0f8a310a...

    work page 2023

Show all 30 references
  1. [9]

    arXiv preprint arXiv:2012.08125 (2020)

    Gao, R., Song, Y., Poole, B., Wu, Y.N., Kingma, D.P.: Learning energy-based models by diffusion recovery likelihood. arXiv preprint arXiv:2012.08125 (2020)

    2020 arXiv

  2. [10]

    Advances in Neural Information Processing Systems33, 6840–6851 (2020)

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems33, 6840–6851 (2020)

    2020

  3. [11]

    Proceedings of the National Academy of Sciences of the United States of America79 8, 2554–8 (1982), https://api.semanticscholar.org/CorpusID: 784288

    Hopfield, J.J.: Neural networks and physical systems with emergent collective com- putational abilities. Proceedings of the National Academy of Sciences of the United States of America79 8, 2554–8 (1982), https://api.semanticscholar.org/CorpusID: 784288

    1982

  4. [12]

    Neurocomputing341, 183– 194 (May 2019)

    Hou, X., Sun, K., Shen, L., Qiu, G.: Improving variational autoencoder with deep feature consistent and generative adversarial training. Neurocomputing341, 183– 194 (May 2019). https://doi.org/10.1016/j.neucom.2019.03.013

    2019 doi

  5. [13]

    In: International Conference on Information Processing in Medical Imaging

    Huang, S., Chen, G., Sun, K., Cui, Z., Zhang, X., Xue, P., Zhang, X., Zhang, H., Shen, D.: Super-resolution reconstruction of fetal brain MRI with prior anatomi- cal knowledge. In: International Conference on Information Processing in Medical Imaging. pp. 428–441. Springer (2023)

    2023

  6. [14]

    Magnetic Resonance Imaging 15(2), 263–265 (1997)

    Laakso,M.P.,Juottonen,K.,Partanen,K.,Vainio,P.,Soininen,H.:MRIvolumetry of the hippocampus: The effect of slice thickness on volume formation. Magnetic Resonance Imaging 15(2), 263–265 (1997)

    1997

  7. [15]

    Proceedings of MIE (2022)

    Müller, D., Hartmann, D., Meyer, P., Auer, F., Soto-Rey, I., Kramer, F.: MISeval: Ametriclibraryformedicalimagesegmentationevaluation.Challengesoftrustable AI and added-value on health. Proceedings of MIE (2022)

    2022

  8. [16]

    arXiv preprint arXiv:1907.08956 (2019)

    Odaibo, S.: Tutorial: Deriving the standard variational autoencoder (vae) loss func- tion. arXiv preprint arXiv:1907.08956 (2019)

    2019 arXiv

  9. [17]

    Advances in Neural Information Processing Systems33, 21994– 22008 (2020)

    Pang, B., Han, T., Nijkamp, E., Zhu, S.C., Wu, Y.N.: Learning latent space energy- based prior model. Advances in Neural Information Processing Systems33, 21994– 22008 (2020)

    2020

  10. [18]

    In: International Conference on Machine Learning

    Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N., Ganguli, S.: Deep unsuper- vised learning using nonequilibrium thermodynamics. In: International Conference on Machine Learning. pp. 2256–2265. PMLR (2015)

    2015

  11. [19]

    IEEE Transactions on Medical Imaging41(6), 1383–1399 (2022)

    Sui, Y., Afacan, O., Jaimes, C., Gholipour, A., Warfield, S.K.: Scan-Specific gener- ative neural network for MRI super-resolution reconstruction. IEEE Transactions on Medical Imaging41(6), 1383–1399 (2022)

    2022

  12. [20]

    BMC Medical Imaging15(1), 1–28 (2015)

    Taha, A.A., Hanbury, A.: Metrics for evaluating 3D medical image segmentation: Analysis, selection, and tool. BMC Medical Imaging15(1), 1–28 (2015)

    2015

  13. [21]

    In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part I 24

    Turella, F., Bredell, G., Okupnik, A., Caprara, S., Graf, D., Sutter, R., Konukoglu, E.: High-resolution segmentation of lumbar vertebrae from conventional thick slice mri. In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conferenc...

    2021

  14. [22]

    Turner, R.: (2005) Energy-based prior latent diffusion model for MRI vertebrae reconstruction 11

    2005

  15. [23]

    In: Proceedings of the 28th International Conference on Machine Learning (ICML-11)

    Welling, M., Teh, Y.W.: Bayesian learning via stochastic gradient Langevin dy- namics. In: Proceedings of the 28th International Conference on Machine Learning (ICML-11). pp. 681–688 (2011)

    2011

  16. [24]

    Wu, A., March, L., Zheng, X., Huang, J., Wang, X., Zhao, J., Blyth, F.M., Smith, E., Buchbinder, R., Hoy, D.: Global low back pain prevalence and years lived with disability from 1990 to 2017: Estimates from the Global Burden of Disease Study

    1990

  17. [25]

    In: International Conference on Machine Learning (ICML)

    Yu, P., Xie, S., Ma, X., Jia, B., Pang, B., Gao, R., Zhu, Y., Zhu, S.C., Wu, YN.: La- tent diffusion energy-based model for interpretable text modeling. In: International Conference on Machine Learning (ICML). (2022)

    2022

  18. [26]

    IEEE Transactions on Medical Imaging40(3), 805–817 (2020) 12 Y

    Zhao, C., Dewey, B.E., Pham, D.L., Calabresi, P.A., Reich, D.S., Prince, J.L.: SMORE: A self-supervised anti-aliasing and super-resolution algorithm for MRI using deep learning. IEEE Transactions on Medical Imaging40(3), 805–817 (2020) 12 Y. Wang et al. Appendix A Theoretical ...

    2020

  19. [28]

    The trained model can be used to generate new samples starting from pure noise

    introduces the diffusion probabilistic model (DPM), which artificially de- creases the quality of the data by adding increasing levels of noise, while training a model to reverse this process, both can be modeled with a Markov chain. The trained model can be used to generate n...

  20. [29]

    proposes DDPM which achieves remarkable results in image synthesis by fixing the variance and learning noise directly. The forward - also diffusion or noising - process starts with a data sample from a real distributionx ∼ q(x0), and Gaussian noise is added gradually to the sa...

  21. [30]

    relies on an information bottleneck in conjunction with geometric clustering for their symbol-vector coupling to avoid mode-collapse and generates more cre- ative text outputs. Their symbol-vector coupling EBM results in the distribution pα(y, z0:T , x) where the symbol vector...

  22. [2017]

    https://doi.org/10

    Annals of Translational Medicine8(6), 299 (Mar 2020). https://doi.org/10. 21037/atm.2020.02.175

    2020

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.