Pith. sign in

REVIEW 3 major objections 5 minor 71 references

Spectral Consistent Flow for One-step 3D Medical Image Translation

T0 review · 3 major / 5 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read One latent-space step, with spectral gain correction, translates 3D medical volumes more accurately than multi-step diffusion or voxel-space baselines.

desk verdict Practical 1-NFE latent Brownian mean-flow plus radial PSD gain for 3D MIT; metric and PSD gains look real across four datasets, but rest on the AE retaining the high-frequency content SCC later amplifies. read the letter →

arxiv 2607.10627 v1 pith:ZXVJDHK3 submitted 2026-07-12 cs.CV

Haoqing Li , Jun Shi , Mingchao Li , Zehua Zhu , Qiwei Jia , Jiong Shi , Hong An This is my paper · ORCID
classification cs.CV
keywords medicalimagetranslationflowmatchingBrownianbridgepowerspectraldensityone-stepgeneration3Dlatentspaceconsistency
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

Medical image translation needs to map one 3D scan modality into another while keeping anatomy and fine texture faithful, yet multi-step diffusion and full-volume regression are too slow and memory-heavy for clinical use. This paper claims that the mapping can be learned as a single-step mean-velocity field of a Brownian bridge between paired latent codes, and that the usual over-smoothing and modality mixing can be fixed by a lightweight spectral corrector. The corrector predicts radial frequency-bin gains so that the power spectrum of the generated volume is forced to match the target, recovering high-frequency detail that average-velocity flow otherwise attenuates. On four multi-modal datasets the resulting one-evaluation model beats both multi-step diffusion and other single-step flow methods on voxel error, structural similarity, and spectral fidelity, and the synthetic images improve downstream disease classification. A sympathetic reader cares because the method removes the usual speed–quality trade-off: it is fast enough for real-time use yet still preserves the fine anatomical cues clinicians rely on.

What carries the argument

Spectral Consistency Corrector (SCC): a lightweight encoder–MLP that predicts per-bin radial frequency gains applied to the 3D FFT of the decoder output, aligning the generated power spectral density with the target and thereby restoring high-frequency texture lost by average-velocity flow and latent compression.

What would settle it

On a held-out paired volume whose latent already discards fine cortical or lesion texture, measure whether SCC still recovers that texture (e.g., by high-frequency PSD match and radiologist scoring); if the corrected image remains over-smoothed relative to ground truth, the recovery claim fails.

Watch

Extended reading notes

Core claim

SC-Flow shows that a support-regularized mean-velocity field of a Brownian bridge in latent space, combined with a learnable radial power-spectrum gain corrector, yields one-step 3D medical image translation that is more accurate, spectrally consistent, and robust than both multi-NFE diffusion and prior single-NFE flow baselines across MRI–PET, PET–MRI, and CT–MRI/CBCT tasks.

Load-bearing premise

The method assumes that a pre-trained latent autoencoder still retains the clinically important high-frequency anatomy, so that a post-hoc spectral gain map can restore it rather than invent missing detail.

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 / 5 minor

Summary. The paper proposes Spectral Consistent Flow (SC-Flow), a latent-space 1-NFE framework for paired 3D medical image translation. It reformulates the mapping as an endpoint-conditioned stochastic Brownian bridge whose mean velocity field is learned by a MeanFlow-style network (Eqs. 1–5), thereby exploiting paired anatomical correspondence without modality-specific priors. To counteract the low-pass bias of average-velocity transport and latent compression, a Spectral Consistency Corrector (SCC) predicts learnable radial frequency-bin gains that re-weight the power spectral density of the decoded volume (Eqs. 6–8, L_spec). End-to-end training combines the mean-flow objective with voxel-domain and log-spectral losses. Experiments on ADNI, A4, NACC and synthRAD2025 (MRI↔PET, CT→MRI/CBCT) report consistent gains in MAE/RMSE/PSNR/SSIM over multi-NFE diffusion and single-NFE baselines, supported by PSD curves, component ablations (Table 4), bin-count sensitivity (Table 6), reverse-direction tasks and a downstream AD-classification study (Table 3).

Significance. If the empirical claims hold, SC-Flow would be a practically useful advance for clinical MIT: it delivers high-fidelity volumetric translation at a single function evaluation, with substantially lower memory and latency than multi-step diffusion or voxel-space networks (Fig. 1). The combination of a support-regularized Brownian-bridge MeanFlow with an explicit spectral-energy regularizer is novel for paired medical data and is validated across eight translation directions, four datasets and a diagnostic downstream task. Strengths that should be credited include the multi-dataset quantitative tables, the PSD alignment analysis (Figs. 6–7), the controlled ablations of the diffusion term / SCC / L_spec, and the demonstration that improved perceptual metrics translate into better classifier performance on synthetic images. These elements make the work a solid candidate for a methods-oriented medical-imaging venue once the load-bearing recovery assumption is clarified.

major comments (3)
  1. Sec. 3.2, Eqs. (6)–(8) and Limitations: SCC multiplies existing Fourier coefficients by a learned radial gain map g_map. Gain modulation can only re-weight residual energy; it cannot restore coefficients that the latent encoder has already driven near zero. The abstract and introduction claim that SCC “enables the model to recover fine-grained anatomical fidelity,” yet the manuscript never quantifies how much high-frequency energy survives the AE before correction. Without such a measurement (or an AE-capacity ablation), the headline superiority of latent 1-NFE over multi-NFE / voxel baselines rests on an untested recovery premise. A short experiment reporting residual PSD of encoded latents, or a comparison against a higher-capacity AE, is needed to make the central claim load-bearing.
  2. Tables 1–2 and Fig. 4: All quantitative claims of “significantly more accurate” performance are reported as single point estimates without standard deviations, confidence intervals or statistical tests across subjects or folds. Given the multi-dataset, multi-modality scope of the strongest claim, the absence of any measure of variability undermines the robustness assertion. Adding subject-level error bars or a paired significance test would strengthen the evidence without altering the experimental design.
  3. Sec. 3.1 and Fig. 1: The paper repeatedly asserts that latent-space 1-NFE SC-Flow “matches or even surpasses voxel-space and multi-NFE approaches.” The only voxel-space single-NFE baseline shown with full metrics is Causal-PETs; the memory/latency comparison in Fig. 1 is not accompanied by a matched voxel-space accuracy table under identical training budgets. A clearer side-by-side of reconstruction metrics for a pure voxel 1-NFE MeanFlow (or Causal-PETs under the same AE-free setting) is required to substantiate the efficiency–accuracy trade-off claim.
minor comments (5)
  1. Keywords and abstract: “Power Spectral Densit” is missing the final “y”; several compound adjectives lack hyphens (e.g., “support regularized,” “frequency-domain”).
  2. Eq. (9): the spectral loss is written as a sum of (log P̂ − log P) without an absolute value or square; the surrounding text calls it a “weighted log mean square error.” Clarify whether the absolute value / square is omitted by typesetting error.
  3. Sec. 3.3: it is stated that the decoder is jointly trained, yet the encoder E is described as “pre-trained.” Explicitly state whether E is frozen, fine-tuned, or also end-to-end.
  4. Fig. 3 caption and Table 6: radial bin count is ablated, but the soft-mask construction m_b(k) and the precise radial bin edges are never defined; a short formula or appendix note would aid reproducibility.
  5. Related Work: the discussion of Consistency Models and distillation methods is accurate but could briefly note that none of those works address paired cross-modal medical constraints, sharpening the novelty claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical supervised method whose claims rest on external metrics and held-out ground truth, not on self-definitional reductions or load-bearing self-citations.

full rationale

SC-Flow is a standard empirical architecture paper. The Brownian-bridge mean-velocity construction (Eqs. 1–5) and Spectral Consistency Corrector (Eqs. 6–8) are design choices trained end-to-end against paired ground-truth volumes via the composite loss (Eq. 11: Lmf + Lvoxel + Lspec). All reported superiority claims (Tables 1–2, Figs. 4–8) are measured by external, non-circular metrics (MAE/RMSE/PSNR/SSIM/FID/MMD) on held-out data from four public datasets. No equation reduces a claimed “prediction” to a fitted constant by construction; no uniqueness theorem is imported from the authors’ prior work; citations to MeanFlow, Brownian bridges, and VAEs are ordinary architectural references, not load-bearing self-citations that force the result. The Limitations section itself flags the AE-recovery assumption without circularly assuming it away. Consequently the derivation chain contains no circular steps.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The central performance claim rests on standard flow-matching and Fourier identities plus several modeling choices (latent AE fidelity, radial isotropy of PSD, hand-chosen loss weights and bin counts) that are not independently validated outside the reported experiments. No new physical entities; the invented modules are engineering constructs whose only evidence is the paper’s own ablations.

free parameters (4)
  • loss weights λ1, λ2, λ3
    Balance mean-flow, voxel, and spectral losses (Eq. 11); values not fixed by theory and affect reported metrics.
  • radial bin count B (default 24)
    Discretization of frequency axis for SCC; Table 6 shows performance depends on this choice.
  • diffusion schedule σ(t) with σ(0)=σ(1)=0
    Controls support expansion of the Brownian bridge; functional form is design choice.
  • spectral bin weights wb
    Weight log-energy errors per band in Lspec (Eq. 9); not derived from first principles.
assumptions (5)
  • domain assumption Paired source–target volumes are anatomically aligned enough that a voxel-wise and spectral supervised loss is meaningful.
    Stated in Introduction and Method; required for Lvoxel and Lspec to supervise the bridge endpoint.
  • standard math Parseval / Wiener–Khinchin: spatial texture fidelity is adequately captured by radial power spectral density energy.
    Invoked in Sec. 3.2 to justify SCC; standard Fourier analysis under stationarity assumptions that medical volumes only approximately satisfy.
  • domain assumption A pre-trained latent AE (VAE/VQGAN) preserves sufficient information for clinically faithful reconstruction after flow and spectral correction.
    Sec. 3.1 and Limitations; latent compression is required for 3D feasibility but is an unproven fidelity bound.
  • domain assumption Average velocity of a Brownian bridge can be learned by a neural net and evaluated in one step (r=0,t=1) without multi-step ODE integration.
    Inherited from MeanFlow [17] and used for 1-NFE claim (Sec. 3.3).
  • ad hoc to paper Radial (rotation-averaged) binning is an adequate spectral representation despite anisotropic medical sampling.
    Sec. 3.2 explicitly chooses radial averaging for compactness; direction-dependent structures may be lost.
invented entities (2)
  • Spectral Consistency Corrector (SCC) with learnable radial gain field g_b / g_map
    purpose: Compensate high-frequency attenuation of average velocity and reduce modality entanglement by aligning generated PSD to target PSD.
    Core novel module; independent_evidence is false because the only validation is internal ablation and PSD plots on the same datasets.
  • Support-regularized Brownian-bridge MeanFlow (SC-Flow) for paired MIT
    purpose: Construct a stochastic endpoint-conditioned trajectory between source and target latents for stable 1-NFE translation under limited data.
    Composition of known bridge and MeanFlow ideas specialized to MIT; evidence is the paper’s own tables.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spectral Consistent Flow for One-step 3D Medical Image Translation." pith.science (2026). https://pith.science/paper/ZXVJDHK3

@misc{pith2026260710627,
  author       = {Pith},
  title        = {Pith review of: Spectral Consistent Flow for One-step 3D Medical Image Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZXVJDHK3}},
  note         = {Machine review of arXiv:2607.10627}
}
read the original abstract

We present Spectral Consistent Flow (SC-Flow), a 3D medical image translation framework with a single function evaluation (1-NFE) in the latent space. This approach reformulates medical image translation as a stochastic Brownian bridge process that directly constructs a mapping between source and target modalities by predicting the support regularized mean velocity field. To mitigate modality entanglement, over-smoothing, and artifacts induced by the implicit low-pass modulation of the latent average velocity, we introduce a Spectral Consistency Corrector that dynamically regularizes the evolution of the power spectral density via learnable frequency-domain gain modulation. This mechanism establishes an explicit bridge between spatial textures and spectral energy flow, enabling the model to recover fine-grained anatomical fidelity while maintaining global structural coherence. Extensive experiments on four datasets demonstrate that SC-Flow delivers significantly more accurate, consistent, and robust performance across various translation scenarios.

Figures

Figures reproduced from arXiv: 2607.10627 by the authors.

Figure 1
Figure 1. SC-Flow enables accurate, fast, and robust one-step 3D medical image transla￾tion. It achieves a lower memory footprint and higher training and inference efficiency compared with other paradigms. Benefiting from the Spectral Consistency Correc￾tor, SC-Flow achieves latent-space 1-NFE performance that matches or even surpasses voxel-space and multi-NFE approaches. SCC enforces spectral-domain consistency complementar… view at source ↗
Figure 2
Figure 2. Architecture Overview. (a) Brownian Bridge Mean Flows construct a stochastic trajectory between source and target latents under the mean velocity field for one-step translation. (b) Spectral Consistency Corrector predicts adaptive frequency￾domain gains to align the power spectral density and preserve texture fidelity. (c) Spectrum Energy Distribution estimates and aggregates 3D FFT magnitudes via ra￾dial binning to… view at source ↗
Figure 3
Figure 3. Effectiveness of the Radial Gain (gb) for spectral correction. Smaller bins are low frequencies, while larger bins are high frequencies. (a) SC-Flow corrects the baseline’s artifacts and high-frequency attenuation. (b) The baseline gain confirms this attenuation (gb ≫ 1), while SC-Flow remains near the ideal gb = 1. (c) 3D Frequency difference maps show the baseline’s widespread spectral errors (left) are effectivel… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Comparison of all the MRI (T1WI and FLAIR) to FDG, Tau, and Aβ-PET scenarios on the NACC, A4 and ADNI datasets. 4 Experiments 4.1 Experimental Setting We evaluate our SC-Flow on eight 3D medical image translation tasks: (1) MRI to FDG-PET, (2) MRI to Tau-PET, (3) MRI t…
Figure 5
Figure 5. Figure 5: CT to MRI / CBCT translation scenarios on the synthRAD2025 dataset [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 7
Figure 7. Figure 7: Comparison of PSD curves across different methods. Our approach is closer to the ground truth. X-axis repre￾sents frequency from low to high values. Right: zoomed-in high-frequency region. Other details of the implementation and data processing are shown in the Supplem…
Figure 4
Figure 4. Figure 4: Our approach achieves the best medical image translation accuracy in [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison on MRI (T1WI and FLAIR) to FDG, Tau, and Aβ-PET translation scenarios on three datasets. stable performance across all three translation settings ( [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Evolution of inter￾mediate bridge states over timesteps, illustrating how the Brownian-bridge trajectory progressively transports the source modality toward the target distribution. method for Alzheimer’s Disease versus Cognitively Normal diagnosis. This set￾ting measu…
Figure 10
Figure 10. Figure 10: FDG, Tau, and Aβ-PET to MRI translation on the NACC dataset. contrast between gray and white matter, showing reduced halo artifacts and sharper regional metabolism compared to others. For the MRI→Tau-PET task, our method maintains a smooth yet anatomically faithful up…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

71 extracted references · 8 linked inside Pith

  1. [1]

    arXiv preprint arXiv:2303.08797 (2023)

    Albergo,M.S.,Boffi,N.M.,Vanden-Eijnden,E.:Stochasticinterpolants:Aunifying framework for flows and diffusions. arXiv preprint arXiv:2303.08797 (2023)

    arXiv 2023

  2. [2]

    In: The Eleventh International Conference on Learning Representations (2023)

    Albergo, M.S., Vanden-Eijnden, E.: Building normalizing flows with stochastic in- terpolants. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  3. [3]

    Academic press (2000)

    Ash, R.B., Doléans-Dade, C.A.: Probability and measure theory. Academic press (2000)

    2000

  4. [4]

    Alzheimer Disease & Associated Disorders21(3), 249–258 (2007)

    Beekly, D.L., Ramos, E.M., Lee, W.W., Deitrich, W.D., Jacka, M.E., Wu, J., Hubbard, J.L., Koepsell, T.D., Morris, J.C., Kukull, W.A., et al.: The national alzheimer’s coordinating center (nacc) database: the uniform data set. Alzheimer Disease & Associated Disorders21(3), 249–258 (2007)

    2007

  5. [5]

    Prentice-Hall, Inc

    Brigham, E.O.: The fast Fourier transform and its applications. Prentice-Hall, Inc. (1988)

    1988

  6. [6]

    arXiv preprint arXiv:2503.07535 (2025)

    Chadebec, C., Tasar, O., Sreetharan, S., Aubin, B.: Lbm: Latent bridge matching for fast image-to-image translation. arXiv preprint arXiv:2503.07535 (2025)

    arXiv 2025

  7. [7]

    In: International Workshop on Simulation and Synthesis in Medical Imaging

    Chartsias, A., Joyce, T., Dharmakumar, R., Tsaftaris, S.A.: Adversarial image synthesis for unpaired multi-modal cardiac data. In: International Workshop on Simulation and Synthesis in Medical Imaging. pp. 3–13 (2017)

    2017

  8. [8]

    IEEE Transactions on Medical Imag- ing37(3), 803–814 (2017)

    Chartsias, A., Joyce, T., Giuffrida, M.V., Tsaftaris, S.A.: Multimodal mr synthesis via modality-invariant latent representation. IEEE Transactions on Medical Imag- ing37(3), 803–814 (2017)

    2017

Show all 71 references
  1. [9]

    103605 (2025)

    Chen, J., Ye, Z., Zhang, R., Li, H., Fang, B., Zhang, L.b., Wang, W.: Medical image translationwithdeeplearning:Advances,datasetsandperspectives.MedicalImage Analysis p. 103605 (2025)

    2025

  2. [10]

    In: Advances in Neural Information Processing Systems

    Chen, R.T., Rubanova, Y., Bettencourt, J., Duvenaud, D.K.: Neural ordinary dif- ferential equations. In: Advances in Neural Information Processing Systems. vol. 31 (2018)

    2018

  3. [11]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Choo, K., Jun, Y., Yun, M., Hwang, S.J.: Slice-consistent 3d volumetric brain ct-to-mri translation with 2d brownian bridge diffusion model. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 657–667 (2024)

    2024

  4. [12]

    Cohen, M., Quispe, G., Le Corff, S., Ollion, C., Moulines, E.: Diffusion bridges vectorquantizedvariationalautoencoders.In:InternationalConferenceonMachine Learning. No. 39 (2022)

    2022

  5. [13]

    arXiv preprint arXiv:2307.08698 (2023)

    Dao, Q., Phung, H., Nguyen, B., Tran, A.: Flow matching in latent space. arXiv preprint arXiv:2307.08698 (2023)

    2023 arXiv

  6. [14]

    IEEE Transactions on Medical Imaging38(10), 2375–2388 (2019)

    Dar, S.U., Yurt, M., Karacan, L., Erdem, A., Erdem, E., Cukur, T.: Image synthe- sis in multi-contrast mri with conditional generative adversarial networks. IEEE Transactions on Medical Imaging38(10), 2375–2388 (2019)

    2019

  7. [15]

    Medical Image Analysis 92, 103046 (2024) 16 H

    Dayarathna, S., Islam, K.T., Uribe, S., Yang, G., Hayat, M., Chen, Z.: Deep learn- ing based synthesis of mri, ct and pet: Review and analysis. Medical Image Analysis 92, 103046 (2024) 16 H. Li et al

    2024

  8. [16]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Esser, P., Rombach, R., Ommer, B.: Taming transformers for high-resolution image synthesis. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 12873–12883 (2021)

    2021

  9. [17]

    arXiv preprint arXiv:2505.13447 (2025)

    Geng,Z.,Deng,M.,Bai,X.,Kolter,J.Z.,He,K.:Meanflowsforone-stepgenerative modeling. arXiv preprint arXiv:2505.13447 (2025)

    2025 arXiv

  10. [18]

    Geng,Z.,Pokle,A.,Kolter,J.Z.:One-stepdiffusiondistillationviadeepequilibrium models.In:AdvancesinNeuralInformationProcessingSystems.vol.36,pp.41914– 41931 (2023)

    2023

  11. [19]

    In: The Thirteenth International Conference on Learning Representations (2025)

    Geng, Z., Pokle, A., Luo, W., Lin, J., Kolter, J.Z.: Consistency models made easy. In: The Thirteenth International Conference on Learning Representations (2025)

    2025

  12. [20]

    In: Advances in Neural Information Processing Systems

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial nets. In: Advances in Neural Information Processing Systems. vol. 27 (2014)

    2014

  13. [21]

    arXiv preprint arXiv:1810.01367 (2018)

    Grathwohl, W., Chen, R.T., Bettencourt, J., Sutskever, I., Duvenaud, D.: Ffjord: Free-form continuous dynamics for scalable reversible generative models. arXiv preprint arXiv:1810.01367 (2018)

    2018 arXiv

  14. [22]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Han, K., Xiong, Y., You, C., Khosravi, P., Sun, S., Yan, X., Duncan, J.S., Xie, X.: Medgen3d: A deep generative framework for paired 3d image and mask generation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 759–769 (2023)

    2023

  15. [23]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 770–778 (2016)

    2016

  16. [24]

    In: Advances in Neural Information Processing Systems

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. In: Advances in Neural Information Processing Systems. vol. 33, pp. 6840–6851 (2020)

    2020

  17. [25]

    New York (2011)

    Hsu, H.P.: Signals and systems. New York (2011)

    2011

  18. [26]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 7132–7141 (2018)

    2018

  19. [27]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with condi- tional adversarial networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 1125–1134 (2017)

    2017

  20. [28]

    Jeong, J., Han, S., Kim, J., Kim, S.J.: Latent space super-resolution for higher- resolutionimagegenerationwithdiffusionmodels.In:ProceedingsoftheComputer Vision and Pattern Recognition Conference. pp. 2355–2365 (2025)

    2025

  21. [29]

    In: European Conference on Computer Vision

    Johnson, J., Alahi, A., Fei-Fei, L.: Perceptual losses for real-time style transfer and super-resolution. In: European Conference on Computer Vision. pp. 694–711 (2016)

    2016

  22. [30]

    SIAM Journal on Mathematical Analysis29(1), 1–17 (1998)

    Jordan, R., Kinderlehrer, D., Otto, F.: The variational formulation of the fokker– planck equation. SIAM Journal on Mathematical Analysis29(1), 1–17 (1998)

    1998

  23. [31]

    springer (2014)

    Karatzas, I., Shreve, S.: Brownian motion and stochastic calculus. springer (2014)

    2014

  24. [32]

    Math- ematische Annalen109(1), 604–615 (1934)

    Khintchine, A.: Korrelationstheorie der stationären stochastischen prozesse. Math- ematische Annalen109(1), 604–615 (1934)

    1934

  25. [33]

    arXiv preprint arXiv:1312.6114 (2013)

    Kingma, D.P., Welling, M.: Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

    2013 arXiv

  26. [34]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Konz, N., Gu, H., Dong, H., Mazurowski, M.A.: The intrinsic manifolds of radiolog- ical images and their role in deep learning. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 684–694 (2022)

    2022

  27. [35]

    arXiv preprint arXiv:2504.10483 (2025) Spectral Consistent Flow 17

    Leng, X., Singh, J., Hou, Y., Xing, Z., Xie, S., Zheng, L.: Repa-e: Unlock- ing vae for end-to-end tuning with latent diffusion transformers. arXiv preprint arXiv:2504.10483 (2025) Spectral Consistent Flow 17

    2025

  28. [36]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, B., Xue, K., Liu, B., Lai, Y.K.: Bbdm: Image-to-image translation with brow- nian bridge diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1952–1961 (2023)

    1952

  29. [37]

    In: Proceedings of Medical Image Computing and Computer Assisted Intervention

    Li, Y., Yakushev, I., Hedderich, D.M., Wachinger, C.: Pasta: Pathology-aware mri to pet cross-modal translation with diffusion models. In: Proceedings of Medical Image Computing and Computer Assisted Intervention. vol. 15007 (2024)

    2024

  30. [38]

    In: Medical Imaging with Deep Learning (2025)

    Li, Y., Li, H., Zhou, S.K.: Causal pets: Causality-informed pet synthesis from multi-modal data. In: Medical Imaging with Deep Learning (2025)

    2025

  31. [39]

    Frontiers in Neuroscience15, 646013 (2021)

    Lin, W., Lin, W., Chen, G., Zhang, H., Gao, Q., Huang, Y., Tong, T., Du, M., Initiative, A.D.N.: Bidirectional mapping of brain mri and pet with 3d reversible gan for the diagnosis of alzheimer’s disease. Frontiers in Neuroscience15, 646013 (2021)

    2021

  32. [40]

    In: The Eleventh International Conference on Learning Rep- resentations (2023)

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. In: The Eleventh International Conference on Learning Rep- resentations (2023)

    2023

  33. [41]

    arXiv preprint arXiv:2209.14577 (2022)

    Liu, Q.: Rectified flow: A marginal preserving approach to optimal transport. arXiv preprint arXiv:2209.14577 (2022)

    2022 arXiv

  34. [42]

    In: The Eleventh International Conference on Learning Representations (2023)

    Liu, X., Gong, C., Liu, Q.: Flow straight and fast: Learning to generate and transfer data with rectified flow. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  35. [43]

    In: The Thirteenth International Conference on Learning Representations (2025)

    Lu, C., Song, Y.: Simplifying, stabilizing and scaling continuous-time consistency models. In: The Thirteenth International Conference on Learning Representations (2025)

    2025

  36. [44]

    Advances in Neural Information Processing Systems36, 76525–76546 (2023)

    Luo, W., Hu, T., Zhang, S., Sun, J., Li, Z., Zhang, Z.: Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models. Advances in Neural Information Processing Systems36, 76525–76546 (2023)

    2023

  37. [45]

    arXiv preprint arXiv:1411.1784 (2014)

    Mirza, M., Osindero, S.: Conditional generative adversarial nets. arXiv preprint arXiv:1411.1784 (2014)

    2014 arXiv

  38. [46]

    In: Interna- tional Conference on Medical Image Computing and Computer-Assisted Interven- tion

    Nie, D., Trullo, R., Lian, J., Petitjean, C., Ruan, S., Wang, Q., Shen, D.: Medical image synthesis with context-aware generative adversarial networks. In: Interna- tional Conference on Medical Image Computing and Computer-Assisted Interven- tion. pp. 417–425 (2017)

    2017

  39. [47]

    IEEE Journal of Biomedical and Health Informatics29(2), 1221–1231 (2024)

    Ou, Z., Pan, Y., Xie, F., Guo, Q., Shen, D.: Image-and-label conditioning latent diffusion model: Synthesizing Abeta-pet from mri for detecting amyloid status. IEEE Journal of Biomedical and Health Informatics29(2), 1221–1231 (2024)

    2024

  40. [48]

    IEEE Transactions on Medical Imaging42(12), 3524–3539 (2023)

    Özbey, M., Dalmaz, O., Dar, S.U., Bedel, H.A., Özturk, Ş., Güngör, A., Cukur, T.: Unsupervised medical image translation with adversarial diffusion models. IEEE Transactions on Medical Imaging42(12), 3524–3539 (2023)

    2023

  41. [49]

    In: European Conference on Computer Vision

    Park, T., Efros, A.A., Zhang, R., Zhu, J.Y.: Contrastive learning for unpaired image-to-image translation. In: European Conference on Computer Vision. pp. 319–345 (2020)

    2020

  42. [50]

    In: Advances in Neural Information Processing Systems

    Park, Y.H., Kwon, M., Choi, J., Jo, J., Uh, Y.: Understanding the latent space of diffusion models through the lens of riemannian geometry. In: Advances in Neural Information Processing Systems. vol. 36, pp. 24129–24142 (2023)

    2023

  43. [51]

    Neurology74(3), 201–209 (2010)

    Petersen, R.C., Aisen, P.S., Beckett, L.A., Donohue, M.C., Gamst, A.C., Harvey, D.J., Jack Jr, C., Jagust, W.J., Shaw, L.M., Toga, A.W., et al.: Alzheimer’s disease neuroimaging initiative (adni) clinical characterization. Neurology74(3), 201–209 (2010)

    2010

  44. [52]

    In: MICCAI Workshop on Deep Generative Models

    Pinaya, W.H., Tudosiu, P.D., Dafflon, J., Da Costa, P.F., Fernandez, V., Nachev, P., Ourselin, S., Cardoso, M.J.: Brain imaging generation with latent diffusion models. In: MICCAI Workshop on Deep Generative Models. pp. 117–126 (2022) 18 H. Li et al

    2022

  45. [53]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 10684–10695 (2022)

    2022

  46. [54]

    In: International Conference on Learning Representations (2022)

    Salimans, T., Ho, J.: Progressive distillation for fast sampling of diffusion models. In: International Conference on Learning Representations (2022)

    2022

  47. [55]

    arXiv preprint arXiv:2508.01292 (2025)

    Sargood, A., Puglisi, L., Cole, J.H., Oxtoby, N.P., Ravì, D., Alexander, D.C.: Co- colit: Controlnet-conditioned latent image translation for mri to amyloid pet syn- thesis. arXiv preprint arXiv:2508.01292 (2025)

    2025

  48. [56]

    In: European Conference on Computer Vision

    Sauer, A., Lorenz, D., Blattmann, A., Rombach, R.: Adversarial diffusion distilla- tion. In: European Conference on Computer Vision. pp. 87–103 (2024)

    2024

  49. [57]

    In: Advances in Neural Information Processing Systems

    Shi, Y., De Bortoli, V., Campbell, A., Doucet, A.: Diffusion schrödinger bridge matching. In: Advances in Neural Information Processing Systems. vol. 36, pp. 62183–62223 (2023)

    2023

  50. [58]

    In: The Twelfth International Conference on Learning Representations (2024)

    Song, Y., Dhariwal, P.: Improved techniques for training consistency models. In: The Twelfth International Conference on Learning Representations (2024)

    2024

  51. [59]

    In: Interna- tional Conference on Machine Learning

    Song, Y., Dhariwal, P., Chen, M., Sutskever, I.: Consistency models. In: Interna- tional Conference on Machine Learning. pp. 32211–32252 (2023)

    2023

  52. [60]

    Song, Y., Ermon, S.: Generative modeling by estimating gradients of the data distribution.In:AdvancesinNeuralInformationProcessingSystems.vol.32(2019)

    2019

  53. [61]

    Sperling, R.A., Rentz, D.M., Johnson, K.A., Karlawish, J., Donohue, M., Salmon, D.P., Aisen, P.: The a4 study: stopping ad before symptoms begin? Science Trans- lational Medicine6(228), 228fs13–228fs13 (2014)

    2014

  54. [62]

    Medical Physics52(7), e17981 (2025)

    Thummerer, A., van der Bijl, E., Galapon, A.J., Kamp, F., Savenije, M., Muijs, C., Aluwini, S., Steenbakkers, R.J., Beuel, S., Intven, M.P., et al.: Synthrad2025 grand challenge dataset: Generating synthetic cts for radiotherapy from head to abdomen. Medical Physics52(7), e179...

    2025

  55. [63]

    In: International Conference on Artificial Intelligence and Statistics

    Tong, A.Y., Malkin, N., Fatras, K., Atanackovic, L., Zhang, Y., Huguet, G., Wolf, G., Bengio, Y.: Simulation-free schrödinger bridges via score and flow matching. In: International Conference on Artificial Intelligence and Statistics. pp. 1279–1287 (2024)

    2024

  56. [64]

    IEEE Transactions on Image Process- ing13(4), 600–612 (2004)

    Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Process- ing13(4), 600–612 (2004)

    2004

  57. [65]

    Acta Mathematica55(1), 117–258 (1930)

    Wiener, N.: Generalized harmonic analysis. Acta Mathematica55(1), 117–258 (1930)

    1930

  58. [66]

    In: International Workshop on Simulation and Synthesis in Medical Imaging

    Wolterink, J.M., Dinkla, A.M., Savenije, M.H., Seevinck, P.R., van den Berg, C.A., Išgum, I.: Deep mr to ct synthesis using unpaired data. In: International Workshop on Simulation and Synthesis in Medical Imaging. pp. 14–23 (2017)

    2017

  59. [67]

    In: International Con- ference on Medical Image Computing and Computer-Assisted Intervention

    Xing, Z., Yang, S., Chen, S., Ye, T., Yang, Y., Qin, J., Zhu, L.: Cross-conditioned diffusion model for medical image to image translation. In: International Con- ference on Medical Image Computing and Computer-Assisted Intervention. pp. 201–211 (2024)

    2024

  60. [68]

    IEEE Transactions on Medical Imaging39(12), 4249–4261 (2020)

    Yang, H., Sun, J., Carass, A., Zhao, C., Lee, J., Prince, J.L., Xu, Z.: Unsuper- vised mr-to-ct synthesis using structure-constrained cyclegan. IEEE Transactions on Medical Imaging39(12), 4249–4261 (2020)

    2020

  61. [69]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Yang, H., Sun, J., Yang, L., Xu, Z.: A unified hyper-gan model for unpaired multi- contrast mr image translation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 127–137 (2021)

    2021

  62. [70]

    Computer Methods and Programs in Biomedicine217, 106676 (2022) Spectral Consistent Flow 19

    Zhang, J., He, X., Qing, L., Gao, F., Wang, B.: Bpgan: Brain pet synthesis from mri using generative adversarial network for multi-modal alzheimer’s disease diagnosis. Computer Methods and Programs in Biomedicine217, 106676 (2022) Spectral Consistent Flow 19

    2022

  63. [71]

    In: Proceedings of the IEEE Interna- tional Conference on Computer Vision

    Zhu, J.Y., Park, T., Isola, P., Efros, A.A.: Unpaired image-to-image translation using cycle-consistent adversarial networks. In: Proceedings of the IEEE Interna- tional Conference on Computer Vision. pp. 2223–2232 (2017)

    2017

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.