Pith. sign in

REVIEW 3 major objections 6 minor 24 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

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

T0 review · glm-5.2

PET-guided diffusion bridge translates whole-body MRI across regions and lesions

2026-07-09 11:55 UTC pith:6IOZFJX5

load-bearing objection Whole-body MRI translation via Schrödinger bridge with VLM region conditioning and PET-guided noise modulation; bridge validity under spatially modulated forward process is unproven. the 3 major comments →

arxiv 2607.07401 v1 pith:6IOZFJX5 submitted 2026-07-08 cs.CV cs.AIcs.LG

Heterogeneity-Adaptive Diffusion Schrodinger Bridge for PET-Guided Whole-Body MRI Translation

Chengbo Wang , Jiacheng Yu , Linjie Bian , Ming Qi , Xiaosheng Liu , Tongtong Che , Jichang Zhang , Shuyu Li
show 2 more authors
Shaoli Song Xiuying Wang
This is my paper
classification cs.CV cs.AIcs.LG
keywords translationwhole-bodybridgediffusionregionsacrossanatomicalchallenges
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper proposes a method called HA-DSB for translating one type of MRI scan into another across the entire body, a task complicated by the fact that different body regions (head, thorax, abdomen, pelvis, thighs) have very different image characteristics and lesions look different from healthy tissue. The authors frame translation as a Diffusion Schrödinger Bridge — a stochastic process that transports images from a source distribution to a target distribution — and make two key modifications. First, they inject region context embeddings generated by a vision-language model into the bridge's conditioning, so the transport dynamics adapt to which body region is being processed. Second, they use co-registered PET scans as a metabolic prior: during the forward (noise-injection) phase, PET uptake spatially modulates how much corruption each pixel receives, and during the reverse (denoising) phase, multi-scale PET-guided attention amplifies lesion-relevant features. The result is that the method achieves the best average structural similarity (SSIM 91.1%) and signal-to-noise ratio (PSNR 26.25 dB) across five body regions compared to five baselines, with PET guidance yielding substantially larger gains on lesion-containing slices than on healthy tissue.

Core claim

The central discovery is that spatially modulating the noise schedule of a Schrödinger bridge using PET uptake — so that regions with high metabolic activity receive different perturbation strengths than surrounding tissue — combined with region-aware conditioning from a vision-language model, produces a single model that handles whole-body MRI translation with improved fidelity in both normal anatomy and pathological regions. The PET guidance effect is concentrated where it matters: lesion-containing slices see a PSNR improvement of 0.80 dB from PET guidance, versus only 0.16 dB on the overall test set, indicating the noise modulation and attention mechanisms are selectively helping diseas-

What carries the argument

The core object is the Diffusion Schrödinger Bridge (DSB), specifically the I2SB instantiation, which defines a stochastic transport between two endpoint image distributions via an analytic Gaussian posterior q(X_t|X_0,X_1). The authors modify the standard bridge in two ways: (1) they replace the spatially uniform forward noise with a PET-guided spatial scaling map S that modulates per-pixel perturbation strength, and (2) they condition the reverse denoising network on region context embeddings derived from a vision-language model (body-location and organ labels encoded by PubMedBERT) fused with diffusion timestep embeddings.

Load-bearing premise

The PET-guided noise modulation modifies the forward noise schedule of the Schrödinger bridge without a corresponding proof that the resulting process still defines a valid bridge with a consistent training objective. If the spatial modulation breaks the bridge formulation, the lesion improvements could come from the PET attention module alone rather than from the noise modulation.

What would settle it

If an ablation removing only the PET-guided noise modulation (but keeping PET attention) shows the same lesion-region improvements, the noise modulation mechanism would not be contributing the claimed effect.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the spatial noise modulation is valid, the same principle could extend to other paired-modalities settings where an auxiliary image provides a spatial prior for where translation difficulty is concentrated (e.g., CT-guided MR translation, or ultrasound-guided CT synthesis).
  • The region context embedding approach suggests that vision-language models can serve as lightweight anatomical priors for any medical imaging task with body-region heterogeneity, potentially replacing hand-crafted region labels.
  • If PET-guided noise modulation genuinely helps lesion fidelity, it could reduce the need for full multi-sequence MRI acquisition in PET/MR protocols, shortening clinical scan times.
  • The concentration of PET-guidance gains in lesion regions (5x the overall improvement) suggests that evaluation on whole-image metrics underestimates clinically relevant gains, and pathology-specific evaluation protocols should become standard.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The modest overall gains from PET guidance (PSNR +0.16) combined with large lesion-specific gains (PSNR +0.80) imply that the noise modulation is doing something localized rather than globally transformative — which is consistent with the design intent but also means the method's value proposition is narrow: it helps mainly when lesions are present.
  • The paper does not prove that spatially modulating the forward noise preserves the Schrödinger bridge's theoretical guarantees. If the bridge formulation is broken, the method may functionally reduce to a conditional diffusion model with PET attention — still potentially useful, but the bridge framing would be incidental rather than load-bearing.
  • The reliance on a specific VLM (Google Gemini 3 Pro) for label generation introduces an external dependency whose failure modes on unusual anatomy (e.g., post-surgical changes, congenital anomalies) are not characterized.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes HA-DSB, a Diffusion Schrödinger Bridge framework for whole-body MRI translation in integrated PET/MR settings. The method addresses two sources of heterogeneity: (1) cross-region anatomical variability, handled via region context embeddings derived from a vision-language model (PubMedBERT + Gemini), and (2) pathological tissue fidelity, addressed through a dual-stage PET guidance mechanism comprising forward-process noise modulation (Eq. 7) and reverse-process multi-scale attention (Eq. 8). Experiments on 246 whole-body PET/MR subjects across five anatomical regions show improvements over five baselines (Table 1), and a lesion-specific ablation (Table 2) isolates the PET guidance effect on pathological slices. Code is stated to be available.

Significance. The clinical motivation is strong: whole-body PET/MR acquisition times of 60–90 minutes are a real bottleneck, and MR-to-MR translation from fast sequences is a practical solution. The paper provides falsifiable quantitative predictions: per-region SSIM/PSNR gains (Table 1) and a lesion-subset ablation (Table 2) showing that PET guidance yields +0.80 PSNR / +1.7 SSIM on lesion slices versus +0.16 / +0.3 on overall data. The region-conditioning design via VLM-generated labels is a reasonable engineering contribution. The code availability statement is a positive for reproducibility. However, the theoretical framing of the PET-guided noise modulation as a modification to the Schrödinger bridge dynamics is not rigorously justified, which weakens the contribution's depth.

major comments (3)
  1. §2.3, Eq. (7): The PET-guided noise modulation replaces the standard bridge forward process x_t = μ_t + Σ_t^{1/2} ε with x_t = μ_t + Σ_t^{1/2}(S⊙ε), changing the marginal from N(μ_t, Σ_t I) to N(μ_t, Σ_t diag(S²)). However, the training objective in Eq. (3) uses the target (X_t − X_0)/σ_t with no S-dependent correction. For the modulated marginal, the correct denoising target should account for the spatially varying diffusion coefficient (e.g., predict S⊙ε or adjust the score by 1/S²). The paper does not prove that the modified forward process, the training objective, and the reverse process remain mutually consistent. This is load-bearing for the claim that the noise modulation modifies 'bridge dynamics' rather than functioning as a form of spatially adaptive data augmentation. The authors should either (a) derive the consistent objective for the modulated bridge and show that Eq. (3)近似
  2. §2.3–2.4, Table 2: The lesion improvement (PSNR +0.80, SSIM +1.7) is attributed to both the forward noise modulation (Eq. 7) and the reverse multi-scale PET attention (Eq. 8), but no ablation isolates these two components. Since the attention module (Eq. 8) with zero-initialized W_out is a standard conditioning mechanism whose lesion benefit is plausible on its own, the marginal contribution of the noise modulation cannot be determined from the current experiments. An ablation with attention-only (no noise modulation) vs. full PET guidance would clarify whether the forward-process modification is necessary or whether the attention module suffices.
  3. §3, Table 1: The 'HA-DSB (no PET)' variant already outperforms all baselines substantially (e.g., Head/Neck SSIM 94.0 vs. ResViT 91.0). However, the paper does not describe what 'no PET' entails—whether the noise modulator S is set to uniform (S=1), whether the PET attention module is removed, or whether both are disabled. This ambiguity makes it difficult to interpret the PET contribution. The variant configuration should be explicitly stated.
minor comments (6)
  1. §2.2: The VLM used is 'Google Gemini 3 Pro'—this should be cited properly if a reference exists, or noted as a proprietary model with version and access date.
  2. §2.2: The 96% accuracy for body-location labels is mentioned but the validation protocol is vague (how many samples, which rater, inter-rater agreement). A brief clarification would help.
  3. Table 1: The dagger (†) notation is defined as 'second-best result' but is applied to HA-DSB (no PET) rows, which is confusing since it is the authors' own ablation, not a competing method. Consider relabeling.
  4. Fig. 2 caption: The figure shows translated T2w results and error maps but does not label which method each column corresponds to. Adding method labels would improve readability.
  5. §3, Dataset: The text mentions 'three co-registered modalities (LAVA, T2, and PET)' but the translation direction (LAVA→T2) is only clear from Fig. 1. Stating this explicitly in the Dataset paragraph would help.
  6. References [9, 10, 14, 15] appear to be from 2025, which is fine, but [10] is an arXiv preprint with a 2507.xxxx number suggesting July 2025—verify this is the correct and citable version.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback. The referee raises three points: (1) theoretical consistency of the PET-guided noise modulation with the DSB training objective, (2) lack of an ablation isolating the forward noise modulation from the reverse attention module, and (3) ambiguity in the 'no PET' variant configuration. We address each below and describe the revisions we will make.

read point-by-point responses

  1. Referee: §2.3, Eq. (7): The PET-guided noise modulation replaces the standard bridge forward process x_t = μ_t + Σ_t^{1/2} ε with x_t = μ_t + Σ_t^{1/2}(S⊙ε), changing the marginal from N(μ_t, Σ_t I) to N(μ_t, Σ_t diag(S²)). However, the training objective in Eq. (3) uses the target (X_t − X_0)/σ_t with no S-dependent correction. For the modulated marginal, the correct denoising target should account for the spatially varying diffusion coefficient (e.g., predict S⊙ε or adjust the score by 1/S²). The paper does not prove that the modified forward process, the training objective, and the reverse process remain mutually consistent.

    Authors: The referee is correct that the modulated forward process in Eq. (7) changes the marginal covariance to N(μ_t, Σ_t diag(S²)), and that the training objective in Eq. (3) does not include an S-dependent correction. We acknowledge this as a genuine gap in the current presentation. In practice, our implementation trains the network to predict the standard target (X_t − X_0)/σ_t while the forward process uses modulated noise, which means the model implicitly learns to compensate for the spatially varying perturbation through the conditioning signal c (which includes region context and, during reverse sampling, PET features via the attention module). However, we agree that this falls short of a rigorous derivation showing mutual consistency of the modified forward process, training objective, and reverse process. We will revise the manuscript in two ways: (1) We will add a derivation showing that when the network is trained to predict S⊙ε (or equivalently, when the score is adjusted by the spatially varying coefficient), the objective remains a valid ELBO for the modulated marginal, and we will clarify that our implementation approximates this by absorbing the S-dependence into the conditioning. (2) We will temper the language from 'modifying bridge dynamics' to 'spatially adaptive forward corruption with approximate reverse consistency,' and explicitly discuss the theoretical limitation that the reverse process does not use a corrected score. We believe this is the honest characterization: the noise modulation functions as a principled form of spatially adaptive perturbation that is empirically effective, but the full theoretical consistency proof is not provided and we will not overclaim it. revision: partial

  2. Referee: §2.3–2.4, Table 2: The lesion improvement (PSNR +0.80, SSIM +1.7) is attributed to both the forward noise modulation (Eq. 7) and the reverse multi-scale PET attention (Eq. 8), but no ablation isolates these two components. Since the attention module (Eq. 8) with zero-initialized W_out is a standard conditioning mechanism whose lesion benefit is plausible on its own, the marginal contribution of the noise modulation cannot be determined from the current experiments. An ablation with attention-only (no noise modulation) vs. full PET guidance would clarify whether the forward-process modification is necessary or whether the attention module suffices.

    Authors: This is a fair and important point. The current experiments do not isolate the individual contributions of the forward noise modulation (Eq. 7) and the reverse multi-scale PET attention (Eq. 8). We will conduct the requested ablation on the lesion-containing subset (Table 2 cohort) with three configurations: (a) attention-only (S = 1, PET attention active), (b) noise modulation only (S active, PET attention removed), and (c) full dual-stage PET guidance. This will allow us to report the marginal contribution of each component. Based on our understanding of the method, we expect both components to contribute—the noise modulation redistributes perturbation strength to emphasize pathological regions during training, while the attention module provides explicit PET-conditioned feature recovery during inference—but we will report whatever the data shows. We will add these results to a revised Table 2 (or an additional table) and update the discussion accordingly. revision: yes

  3. Referee: §3, Table 1: The 'HA-DSB (no PET)' variant already outperforms all baselines substantially (e.g., Head/Neck SSIM 94.0 vs. ResViT 91.0). However, the paper does not describe what 'no PET' entails—whether the noise modulator S is set to uniform (S=1), whether the PET attention module is removed, or whether both are disabled. This ambiguity makes it difficult to interpret the PET contribution. The variant configuration should be explicitly stated.

    Authors: The referee is correct that the manuscript does not explicitly define the 'no PET' configuration. To clarify: in the 'HA-DSB (no PET)' variant, both PET-dependent components are disabled—the noise modulator S is set to uniform (S = 1 everywhere, equivalent to the standard bridge forward process), and the multi-scale PET attention modules are removed from the UNet. The model retains the region context embeddings and all other architectural components. We will add this explicit description to Section 3 (Experiments) where the variant is introduced, so readers can unambiguously interpret the PET contribution shown in Tables 1 and 2. revision: yes

Circularity Check

0 steps flagged

No circularity: HA-DSB's claims are tested against external baselines on held-out data, with independent ablation isolating PET guidance effects.

full rationale

The paper's central claims are evaluated against external baselines (Pix2Pix, Palette, I2SB, PPT, SelfRDB, ResViT) on a held-out test set of 42 subjects (4,162 slices), and the PET guidance contribution is isolated via ablation (HA-DSB with/without PET) on a separate lesion-confirmed cohort (Table 2). The VLM labels are generated by an external model (Gemini 3 Pro) and validated by a radiologist at ~96% accuracy. The I2SB instantiation [18] and DSB formulation [17] are cited from independent prior work by different author groups. The PET-guided noise modulation (Eq. 7) and multi-scale PET attention (Eq. 8) are novel architectural components, not definitions that reduce to their own outputs. The training objective (Eq. 3) is the standard denoising score-matching loss from the I2SB framework, not a quantity defined in terms of the paper's own predictions. While the reader raises a valid correctness concern about whether the modified forward process (Eq. 7) yields a valid Schrödinger bridge given the unmodified training objective (Eq. 3), this is a theoretical consistency issue, not circularity: the empirical results are measured against ground-truth MRI on held-out data, not against a quantity defined by the method's own construction. No step in the derivation chain reduces to its inputs by definition or self-citation.

Axiom & Free-Parameter Ledger

10 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities, particles, forces, or dimensions. All components (Schrödinger bridge, U-Net, VLM, PET, attention, FiLM) are established techniques. The 'region context embedding' c_reg is a new combination of existing components (PubMedBERT + cross-attention + VLM labels) but not a new entity. The 'PET-guided noise modulator' M_ϕ is a lightweight CNN, not a new architectural entity. The free parameters are standard hyperparameters (learning rate, channel counts, diffusion steps) plus two hand-chosen values (K=11, s_min/s_max) without sensitivity analysis. The most critical axiom is the unproven validity of the Schrödinger bridge under spatially modulated forward noise (Eq. 7).

free parameters (10)
  • K (number of body regions) = 11
    Chosen by hand in §2.2 to partition the body into regions; no sensitivity analysis provided.
  • s_min, s_max (noise scaling bounds) = 0.2, 2.0
    Stated in §2.3; control the range of the spatial scaling map S. No justification or sensitivity analysis given.
  • T (diffusion steps) = 200
    Stated in Implementation Details; standard hyperparameter.
  • β_max (variance schedule) = 0.35
    Stated in Implementation Details; controls noise schedule.
  • H (attention heads) = 4
    Stated in §2.2 for cross-attention; standard.
  • d (embedding dimension) = Not specified
    Used throughout §2.2 but value not stated.
  • Learning rate = 1e-4
    Standard AdamW hyperparameter.
  • EMA decay = 0.9998
    Standard hyperparameter.
  • Base channels = 128
    U-Net architecture choice.
  • Channel multipliers = [1,1,2,2,4,4]
    U-Net architecture choice.
axioms (5)
  • standard math The I2SB instantiation of the Schrödinger Bridge with drift f=0 and paired samples (X_0, X_1) yields a valid analytic posterior q(X_t|X_0,X_1) = N(μ_t, Σ_t).
    Invoked in §2.1, Eq. 2; this is the standard I2SB result from Liu et al. 2023.
  • ad hoc to paper Modifying the forward noise from uniform ε to spatially scaled S⊙ε (Eq. 7) yields a valid bridge process that can be trained with the standard objective (Eq. 3).
    Invoked in §2.3, Eq. 7. The paper does not prove that the Schrödinger bridge remains valid under spatially non-uniform noise modulation, nor that the training objective is consistent. This is the load-bearing assumption.
  • domain assumption VLM-generated body-location labels (~96% accuracy) and organ labels are sufficiently reliable for conditioning.
    Invoked in §2.2; the 96% accuracy is stated but the validation methodology is not described. Label noise could degrade region conditioning.
  • domain assumption PET uptake is a reliable proxy for lesion location and can guide noise modulation to improve pathological fidelity.
    Invoked in §2.3 and §2.4; assumes focal high-uptake PET signals correspond to lesions. This is clinically reasonable but not universally true (e.g., inflammation, physiologic uptake).
  • domain assumption SSIM and PSNR are adequate metrics for evaluating MRI translation quality, including in pathological regions.
    Used throughout §3; these are standard but may not capture clinical utility or lesion detectability. No radiologist reader study is reported.

pith-pipeline@v1.1.0-glm · 12335 in / 3746 out tokens · 439277 ms · 2026-07-09T11:55:16.394339+00:00 · methodology

0 comments
read the original abstract

While whole-body multimodal medical imaging scanners have been increasingly recognized for more effective medical applications, the excessive long acquisition time in PET-MR scanning is a major obstacle in more efficient clinical practice. Deep learning-based MRI translation provides a potential solution to reduce scan duration. However, current models often focus on specific anatomical regions and face challenges for whole-body scans that consists of highly heterogeneous feature distributions mainly due to (1) different anatomical regions across whole-body, and (2) lesions or pathological tissues. This paper tackles the challenges through a novel Heterogeneity-Adaptive Diffusion Schrodinger Bridge (HA-DSB) framework. By explicitly modeling translation as stochastic transport between source and target distributions, HA-DSB incorporates region context embeddings derived from a vision-language model (VLM) to enable region-specific modeling. To enhance fidelity of the pathological tissue, lesion-aware metabolic prior from PET is integrated directly into the bridge dynamics through a dual-stage guidance mechanism. Specifically, a PET-guided noise modulation module adaptively scales spatial diffusion perturbations during the forward process, while PET features are leveraged during the reverse process to selectively amplify lesion-relevant structures via an attention mechanism. Experiments demonstrate the superiority of our method across different body regions in whole-body MRI translation and show improved translation quality in lesion areas under PET guidance. Our code is available at Github.

Figures

Figures reproduced from arXiv: 2607.07401 by Chengbo Wang, Jiacheng Yu, Jichang Zhang, Linjie Bian, Ming Qi, Shaoli Song, Shuyu Li, Tongtong Che, Xiaosheng Liu, Xiuying Wang.

Figure 1
Figure 1. Figure 1: Overview of HA-DSB. (a) Whole-body MR translation setting. (b) Diffusion Schrödinger bridge from LAVA to T2 with noise modulation jointly conditioned on PET and region context creg. (c) UNet translation backbone with region-conditioned residual blocks and multi-scale PET attention. (d) Construction of the region context embedding creg. where σ 2 t := R t 0 βτ dτ and σ¯ 2 t := R 1 t βτ dτ denote the varianc… view at source ↗
Figure 2
Figure 2. Figure 2: Qualitative comparison across five anatomical regions. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effect of PET guidance on pathological cases (liver lesions). [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages · 1 internal anchor

  1. [1]

    Masselli, G., Di Bella, C.: Will PET/MR Imaging Replace PET/CT for Pediatric Applications? Diagnostics 15(9), 1070 (2025)

    work page 2025

  2. [2]

    Academic Press, Cambridge (2022)

    Catalano, O.A.: Clinical PET/MRI. Academic Press, Cambridge (2022)

    work page 2022

  3. [3]

    IEEE Transactions on Medical Imaging 42(12), 3524–3539 (2023)

    Özbey, M., et al.: Unsupervised Medical Image Translation with Adversarial Dif- fusion Models. IEEE Transactions on Medical Imaging 42(12), 3524–3539 (2023)

    work page 2023

  4. [4]

    In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2024

    Xing, Z., et al.: Cross-Conditioned Diffusion Model for Medical Image to Image Translation. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2024. Springer, Cham (2024)

    work page 2024

  5. [5]

    IEEE Transactions on Medical Imaging 41(10), 2598–2614 (2022)

    Dalmaz, O., Yurt, M., Cukur, T.: ResViT: Residual Vision Transformers for Mul- timodal Medical Image Synthesis. IEEE Transactions on Medical Imaging 41(10), 2598–2614 (2022). https://doi.org/10.1109/TMI.2022.3167808

    work page doi:10.1109/tmi.2022.3167808 2022

  6. [6]

    Medical Image Analysis 106, 103747 (2025)

    Arslan, F., Kabas, B., Dalmaz, O., Ozbey, M., Cukur, T.: Self-Consistent Recur- sive Diffusion Bridge for Medical Image Translation. Medical Image Analysis 106, 103747 (2025). https://doi.org/10.1016/j.media.2025.103747

    work page doi:10.1016/j.media.2025.103747 2025

  7. [7]

    In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2023

    Jiang, L., et al.: Cola-Diff: Conditional Latent Diffusion Model for Multi-Modal MRI Synthesis. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2023. Springer, Cham (2023)

    work page 2023

  8. [8]

    In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2013

    Dzyubachyk, O., et al.: Joint Intensity Inhomogeneity Correction for Whole-Body MR Data. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2013. Springer, Berlin, Heidelberg (2013)

    work page 2013

  9. [9]

    Medical Image Analysis 103, 103622 (2025) 10 C

    Zheng, J., et al.: Structure-Guided MR-to-CT Synthesis with Spatial and Semantic Alignments for Attenuation Correction of Whole-Body PET/MR Imaging. Medical Image Analysis 103, 103622 (2025) 10 C. Wang et al

    work page 2025

  10. [10]

    3D Wavelet Latent Diffusion Model for Whole-Body MR-to-CT Modality Translation

    Zheng, J., et al.: 3D Wavelet Latent Diffusion Model for Whole-Body MR-to-CT Modality Translation. arXiv preprint arXiv:2507.11557 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  11. [11]

    In: Proceedings of the International Conference on Artificial Intelligence and Statistics

    Gong, Y., et al.: Testing Generated Distributions in GANs to Penalize Mode Col- lapse. In: Proceedings of the International Conference on Artificial Intelligence and Statistics. PMLR (2024)

    work page 2024

  12. [12]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 46(8), 5493–5503 (2024)

    Luo, Y., Yang, Z.: DynGAN: Solving Mode Collapse in GANs with Dynamic Clus- tering. IEEE Transactions on Pattern Analysis and Machine Intelligence 46(8), 5493–5503 (2024)

    work page 2024

  13. [13]

    In: International Conference on Learning Representations, ICLR (2023)

    Um, S., Lee, S., Ye, J.C.: Don’t Play Favorites: Minority Guidance for Diffusion Models. In: International Conference on Learning Representations, ICLR (2023)

    work page 2023

  14. [14]

    In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2025

    Zhang, X., et al.: Structure-Aware MRI Translation: Multi-Modal Latent Diffu- sion Model with Arbitrary Missing Modalities. In: Medical Image Computing and Computer-Assisted Intervention – MICCAI 2025. Springer, Cham (2025)

    work page 2025

  15. [15]

    npj Artificial Intelligence 1(1), 11 (2025)

    Dayarathna, S., et al.: MU-Diff: A Mutual Learning Diffusion Model for Synthetic MRI with Application for Brain Lesions. npj Artificial Intelligence 1(1), 11 (2025)

    work page 2025

  16. [16]

    Medical Image Analysis 88, 102872 (2023)

    Güngör, A., et al.: Adaptive Diffusion Priors for Accelerated MRI Reconstruction. Medical Image Analysis 88, 102872 (2023)

    work page 2023

  17. [17]

    In: International Conference on Learning Representations, ICLR (2022)

    Chen, T., Liu, G.H., Theodorou, E.: Likelihood Training of Schrödinger Bridge Using Forward-Backward SDEs Theory. In: International Conference on Learning Representations, ICLR (2022)

    work page 2022

  18. [18]

    In: Proceedings of the 40th International Conference on Machine Learning

    Liu, G.H., et al.: I2SB: Image-to-Image Schrödinger Bridge. In: Proceedings of the 40th International Conference on Machine Learning. PMLR (2023)

    work page 2023

  19. [19]

    ACM Transactions on Computing for Healthcare 3(1), 1–23 (2022)

    Gu, Y., Tinn, R., Cheng, H., Lucas, M., Usuyama, N., Liu, X., Naumann, T., Gao, J., Poon, H.: Domain-Specific Language Model Pretraining for Biomedical Natural Language Processing. ACM Transactions on Computing for Healthcare 3(1), 1–23 (2022). https://doi.org/10.1145/3458754

    work page doi:10.1145/3458754 2022

  20. [20]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Perez, E., Strub, F., de Vries, H., Dumoulin, V., Courville, A.: FiLM: Vi- sual Reasoning with a General Conditioning Layer. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 32, no. 1, pp. 3942–3951 (2018). https://doi.org/10.1609/aaai.v32i1.11671

    work page doi:10.1609/aaai.v32i1.11671 2018

  21. [21]

    In: Advances in Neural Information Processing Systems, vol

    Dhariwal, P., Nichol, A.: Diffusion Models Beat GANs on Image Synthesis. In: Advances in Neural Information Processing Systems, vol. 34, pp. 8780–8794 (2021)

    work page 2021

  22. [22]

    In: Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pp

    Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-Image Translation with Conditional Adversarial Networks. In: Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pp. 1125–1134 (2017). https://doi.org/10.1109/CVPR.2017.632

    work page doi:10.1109/cvpr.2017.632 2017

  23. [23]

    In: ACM SIGGRAPH 2022 Con- ference Proceedings, pp

    Saharia, C., Chan, W., Chang, H., Lee, C., Ho, J., Salimans, T., Fleet, D., Norouzi, M.: Palette: Image-to-Image Diffusion Models. In: ACM SIGGRAPH 2022 Con- ference Proceedings, pp. 1–10 (2022). https://doi.org/10.1145/3528233.3530757

    work page doi:10.1145/3528233.3530757 2022

  24. [24]

    In: Linguraru, M.G., Dou, Q., Feragen, A., Giannarou, S., Glocker, B., Lekadir, K., Schnabel, J.A

    Zhang, W., Hui, T.H., Tse, P.Y., Hill, F., Lau, C., Li, X.: High-Resolution Medical Image Translation via Patch Alignment-Based Bidirectional Contrastive Learning. In: Linguraru, M.G., Dou, Q., Feragen, A., Giannarou, S., Glocker, B., Lekadir, K., Schnabel, J.A. (eds.) Medical Image Computing and Computer-Assisted Inter- vention – MICCAI 2024. LNCS, vol. ...

    work page 2024