Controllable Histopathology Image Synthesis with Training-free Structural Initialization and Textural Modulation

Chenfei Ye; Jianfeng Cao; Jingyi Luo; Ting Ma; Yuheng Qiu

arxiv: 2606.27935 · v2 · pith:CIKWMTMKnew · submitted 2026-06-26 · 💻 cs.CV

Controllable Histopathology Image Synthesis with Training-free Structural Initialization and Textural Modulation

Yuheng Qiu , Jingyi Luo , Chenfei Ye , Ting Ma , Jianfeng Cao This is my paper

Pith reviewed 2026-07-01 06:42 UTC · model grok-4.3

classification 💻 cs.CV

keywords histopathology image synthesisdiffusion modelstraining-free controlstructural initializationwavelet modulationpaired data generationsegmentation augmentation

0 comments

The pith

A training-free plug-in lets any pretrained diffusion model generate histopathology images that follow given structural masks while keeping the style of reference tissue.

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

The paper presents CHIS, a method that steers the sampling process of a diffusion model already trained only on unlabeled images. It does so in two steps: first by constructing an initial noise image whose phase spectrum matches a supplied mask, then by adjusting coarse and fine textures at different wavelet scales during the reverse diffusion steps. This combination produces outputs whose layout respects the mask without any retraining or paired mask-image data. The authors show the outputs improve the performance of downstream segmentation models that are trained on the synthetic pairs.

Core claim

By replacing the starting Gaussian noise with a frequency-domain mixture that takes its phase from the structural mask and its amplitude from noise, and by then performing adaptive modulation of wavelet sub-bands at multiple decomposition levels throughout the denoising trajectory, a diffusion model pretrained solely on unlabeled histopathology images can be made to produce images whose spatial layout matches an arbitrary input mask while retaining the textural statistics of a chosen reference image.

What carries the argument

structural initialization by phase-amplitude fusion in the frequency domain together with adaptive multi-scale wavelet-band modulation during reverse diffusion

If this is right

Any diffusion model trained on unlabeled histopathology slides can immediately be used for paired data generation without collecting new annotations.
The same mask can be paired with different reference images to vary tissue appearance while keeping layout fixed.
Downstream segmentation accuracy rises when models are trained on the synthetic pairs produced by this process.
The approach requires no gradient updates or fine-tuning at inference time.

Where Pith is reading between the lines

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

The same initialization and modulation steps could be tested on diffusion models for other medical imaging modalities such as radiology or microscopy.
If the phase-fusion step generalizes, the method might reduce the need for large annotated datasets in any domain where structural masks are cheaper to obtain than full images.
A natural extension would be to replace the supplied mask with an automatically extracted one from a different modality to test cross-modal structural transfer.

Load-bearing premise

The assumption that phase information taken from the mask and inserted into the initial noise, plus later wavelet-level texture adjustments, will be enough to force structural fidelity without any training on paired examples.

What would settle it

Generate a set of images with CHIS using masks whose structures are known to be absent from the pretraining distribution; if independent segmentation networks trained on real data achieve no higher Dice scores on the CHIS outputs than on standard diffusion samples, the structural-control claim is falsified.

Figures

Figures reproduced from arXiv: 2606.27935 by Chenfei Ye, Jianfeng Cao, Jingyi Luo, Ting Ma, Yuheng Qiu.

**Figure 1.** Figure 1: Overview of our proposed CHIS for controllable histopathology image synthesis. depends critically on the quantity and quality of labeled training data, which is expensive and time-consuming to obtain in routine clinical workflows. Synthetic data generation therefore offers a practical solution to alleviate annotation scarcity. While generative adversarial networks (GANs) can learn with weak or unpaired su… view at source ↗

**Figure 2.** Figure 2: Comparison of synthesized images from different methods. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

read the original abstract

Deep learning has demonstrated remarkable success in high-throughput histopathology image analysis. However, the performance of learning-based models critically depends on the quality and size of annotations by expert pathologists, which is a resource-intensive and time-consuming process. To address the limitations of data scarcity and annotation burden, several methods have been proposed to synthesize paired histopathology data. Nevertheless, these frameworks typically still require annotation data, albeit in reduced quantities, to impose structural constraints during training. In this work, we present CHIS, a plug-in framework that guides the sampling trajectory of a pretrained diffusion model through two key stages: structural initialization at the start and textural modulation during generation. The initial noise state is refined by fusing the phase information from a prior mask with the amplitude of Gaussian noise in the frequency domain, yielding a structurally informed starting point. During the reverse diffusion process, we adaptively modulate both coarse-grained and fine-grained textures at different wavelet decomposition levels. This enables a diffusion model pretrained solely on unlabeled images to generate outputs that align with prior structural masks while preserving the reference tissue style. We conducted extensive experiments demonstrating the superiority of CHIS in generation fidelity and its substantial benefits for downstream segmentation tasks. Code is available at https://github.com/IBIL-Code/CHIS.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

CHIS offers a training-free plug-in for mask-controlled histopathology synthesis via phase-amplitude init and wavelet modulation, but the structure-preservation claim rests on unverified assumptions about diffusion trajectories.

read the letter

The core idea is a plug-in that steers a diffusion model pretrained only on unlabeled histopathology slides so its outputs follow an input structural mask. It starts by swapping the phase spectrum of the mask into the initial noise while keeping the amplitude from Gaussian noise, then applies adaptive wavelet-band modulation at multiple decomposition levels during the reverse process to handle coarse and fine textures separately.

This is genuinely new in the training-free aspect: most prior controllable synthesis still needs some paired mask-image data or fine-tuning to enforce structure. The approach directly targets the annotation bottleneck in pathology, where getting expert labels is slow and costly, and it ships code, which helps.

The experiments claim better fidelity and downstream segmentation gains, which is the practical payoff if true. The frequency-domain initialization plus wavelet steps are a reasonable attempt to inject structure without retraining.

The main soft spot is whether the injected structure actually survives the full denoising trajectory. The base model was trained without any mask supervision, so nothing in its prior enforces boundary alignment. A single initialization step plus modulation could easily lose the mask signal over hundreds of steps, and the paper does not appear to provide analytic bounds, targeted ablations on structure retention, or controls that isolate how much alignment comes from the mask versus the reference style image. Without those, the central claim is hard to assess.

This is for researchers working on data augmentation for computational pathology or diffusion-based medical image synthesis. A reader already familiar with frequency-domain tricks in diffusion would get the most out of it.

It deserves peer review because the problem is real and the method is distinct, even if the evidence needs strengthening on the structure-preservation side.

Referee Report

2 major / 2 minor

Summary. The paper presents CHIS, a training-free plug-in framework for controllable synthesis of histopathology images using a diffusion model pretrained on unlabeled data. The method involves structural initialization by fusing the phase spectrum from a structural mask with the amplitude spectrum of Gaussian noise in the frequency domain at the initial noise state, followed by adaptive textural modulation at different wavelet decomposition levels during the reverse diffusion process. This is claimed to produce images that align with the input structural masks while preserving reference tissue style, with demonstrated superiority in fidelity and benefits for downstream segmentation tasks.

Significance. If the claims hold, the approach would be significant as it enables controllable generation without requiring paired mask-image training data, addressing data scarcity and annotation burden in histopathology analysis. The training-free nature and use of pretrained models on unlabeled images could facilitate broader adoption in medical imaging.

major comments (2)

[Abstract] Abstract: The abstract asserts that 'extensive experiments' demonstrate superiority in generation fidelity and substantial benefits for downstream segmentation tasks, yet supplies no quantitative metrics, ablation results, or experimental protocol. This absence is load-bearing because the central claim of effective structural control cannot be verified.
[Method] Method description: No analytic bound or ablation is provided to show that phase-amplitude fusion at initialization plus wavelet-band modulation during reverse diffusion suffices to enforce mask alignment throughout the full denoising trajectory. Since the underlying model was trained only on unlabeled images, this directly tests whether injected structure survives without drift or is supplied by the reference style image.

minor comments (2)

[Abstract] The code repository link is provided, supporting reproducibility.
Notation for frequency-domain operations and specific wavelet levels could be clarified for readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments point by point below, clarifying the manuscript content and proposing targeted revisions.

read point-by-point responses

Referee: [Abstract] Abstract: The abstract asserts that 'extensive experiments' demonstrate superiority in generation fidelity and substantial benefits for downstream segmentation tasks, yet supplies no quantitative metrics, ablation results, or experimental protocol. This absence is load-bearing because the central claim of effective structural control cannot be verified.

Authors: We agree the abstract is too high-level. The full manuscript (Section 4) reports quantitative metrics including FID, LPIPS, and downstream Dice/IoU improvements, along with the evaluation protocol on public histopathology datasets. We will revise the abstract to include key numerical results and a concise protocol summary. revision: yes
Referee: [Method] Method description: No analytic bound or ablation is provided to show that phase-amplitude fusion at initialization plus wavelet-band modulation during reverse diffusion suffices to enforce mask alignment throughout the full denoising trajectory. Since the underlying model was trained only on unlabeled images, this directly tests whether injected structure survives without drift or is supplied by the reference style image.

Authors: The paper supplies empirical evidence via alignment metrics, visual trajectory inspections, and ablations isolating initialization versus modulation, showing structure persists independently of the style reference. No analytic bound is derived, as the work is empirical; we can add further trajectory ablations in revision to strengthen this point. revision: partial

Circularity Check

0 steps flagged

No circularity: method is a proposed heuristic without definitional reduction or self-citation chains

full rationale

The paper describes a training-free plug-in technique (phase-amplitude fusion at initialization plus wavelet-band modulation) applied to an externally pretrained diffusion model. No equations, fitted parameters, or predictions are presented that reduce to the inputs by construction. No self-citations are invoked as load-bearing uniqueness theorems. The central claim is an empirical assertion about the sufficiency of the described operations, which remains independent of the target outputs and does not rename or smuggle prior results. This is the common case of a self-contained engineering proposal.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; all technical details are deferred to the full manuscript.

pith-pipeline@v0.9.1-grok · 5767 in / 1146 out tokens · 31646 ms · 2026-07-01T06:42:16.228841+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

28 extracted references · 7 canonical work pages · 1 internal anchor

[1]

In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

Bansal, A., Chu, H.M., Schwarzschild, A., Sengupta, S., Goldblum, M., Geiping, J., Goldstein, T.: Universal guidance for diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 843–852 (2023)

2023
[2]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Bhosale, M., Wasi, A., Zhai, Y., Tian, Y., Border, S., Xi, N., Sarder, P., Yuan, J., Doermann, D., Gong, X.: Pathdiff: Histopathology image synthesis with un- paired text and mask conditions. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 22415–22424 (2025)

2025
[3]

In: 2022 IEEE 19th International sympo- sium on biomedical imaging (ISBI)

Butte, S., Wang, H., Xian, M., Vakanski, A.: Sharp-gan: Sharpness loss regularized gan for histopathology image synthesis. In: 2022 IEEE 19th International sympo- sium on biomedical imaging (ISBI). pp. 1–5. IEEE (2022)

2022
[4]

Improved Regularization of Convolutional Neural Networks with Cutout

DeVries, T., Taylor, G.W.: Improved regularization of convolutional neural net- works with cutout. arXiv preprint arXiv:1708.04552 (2017),https://arxiv.org/ abs/1708.04552

work page internal anchor Pith review Pith/arXiv arXiv 2017
[5]

Nature medicine pp

Ding, T., Wagner, S.J., Song, A.H., Chen, R.J., Lu, M.Y., Zhang, A., Vaidya, A.J., Jaume,G.,Shaban,M.,Kim,A.,etal.:Amultimodalwhole-slidefoundationmodel for pathology. Nature medicine pp. 1–13 (2025)

2025
[6]

In: European Congress on Digital Pathology

Gamper, J., Koohbanani, N.A., Benet, K., Khuram, A., Rajpoot, N.: Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classifica- tion. In: European Congress on Digital Pathology. pp. 11–19. Springer (2019)

2019
[7]

IEEE Transactions on Medical Imaging 39(12), 4124–4136 (2020)

Graham, S., Epstein, D., Rajpoot, N.: Dense steerable filter cnns for exploiting rotational symmetry in histology images. IEEE Transactions on Medical Imaging 39(12), 4124–4136 (2020)

2020
[8]

Medical image analysis58, 101563 (2019)

Graham, S., Vu, Q.D., Raza, S.E.A., Azam, A., Tsang, Y.W., Kwak, J.T., Rajpoot, N.: Hover-net: Simultaneous segmentation and classification of nuclei in multi- tissue histology images. Medical image analysis58, 101563 (2019)

2019
[9]

Signal processing: Image communication13(3), 171–181 (1998)

Huang, W.C., Chang, L.W.: Predictive subband image coding with wavelet trans- form. Signal processing: Image communication13(3), 171–181 (1998)

1998
[10]

Nature methods18(2), 203–211 (2021)

Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H.: nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods18(2), 203–211 (2021)

2021
[11]

Journal of pathology informatics 7(1), 29 (2016)

Janowczyk,A.,Madabhushi,A.:Deeplearningfordigitalpathologyimageanalysis: A comprehensive tutorial with selected use cases. Journal of pathology informatics 7(1), 29 (2016)

2016
[12]

IEEE transactions on medical imaging43(3), 980–993 (2023)

Li, Y., Shao, H.C., Liang, X., Chen, L., Li, R., Jiang, S., Wang, J., Zhang, Y.: Zero-shot medical image translation via frequency-guided diffusion models. IEEE transactions on medical imaging43(3), 980–993 (2023)

2023
[13]

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
[14]

In: International conference on machine learning

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: International conference on machine learning. pp. 8748–8763. PmLR (2021) 10 Qiu et al

2021
[15]

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
[16]

Advances in neural information processing systems29(2016)

Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved techniques for training gans. Advances in neural information processing systems29(2016)

2016
[17]

Nature Reviews Bioengineering1(12), 930–949 (2023)

Song, A.H., Jaume, G., Williamson, D.F., Lu, M.Y., Vaidya, A., Miller, T.R., Mah- mood, F.: Artificial intelligence for digital and computational pathology. Nature Reviews Bioengineering1(12), 930–949 (2023)

2023
[18]

net/forum?id=PxTIG12RRHS

Song,J.,Meng,C.,Ermon,S.:Denoisingdiffusionimplicitmodels.In:International Conference on Learning Representations (ICLR) (2020),https://openreview. net/forum?id=PxTIG12RRHS

2020
[19]

Patterns1(6) (2020)

Tschuchnig,M.E.,Oostingh,G.J.,Gadermayr,M.:Generativeadversarialnetworks in digital pathology: a survey on trends and future potential. Patterns1(6) (2020)

2020
[20]

IEEE Transactions on Medical Imaging40(12), 3413–3423 (2021).https://doi.org/10.1109/TMI.2021.3085712

Verma, R., Kumar, N., Patil, A., Kurian, N.C., Rane, S., Graham, S., Vu, Q.D., Zwager, M., Raza, S.E.A., Rajpoot, N., Wu, X., Chen, H., Huang, Y., Wang, L., Jung, H., Brown, G.T., Liu, Y., Liu, S., Jahromi, S.A.F., Khani, A.A., Montahaei, E., Baghshah, M.S., Behroozi, H., Semkin, P., Rassadin, A., Dutande, P., Lodaya, R., Baid, U., Baheti, B., Talbar, S.,...

work page doi:10.1109/tmi.2021.3085712 2021
[21]

arXiv preprint arXiv:2207.00050 (2022)

Wang, W., Bao, J., Zhou, W., Chen, D., Chen, D., Yuan, L., Li, H.: Semantic image synthesis via diffusion models. arXiv preprint arXiv:2207.00050 (2022)

work page arXiv 2022
[22]

Knudsen, Tahsin Kurc, Rajarsi R

Yellapragada, S., Graikos, A., Li, Z., Triaridis, K., Belagali, V., Kapse, S., Nandi, T.N., Madduri, R.K., Prasanna, P., Kurc, T., et al.: Pixcell: A generative foun- dation model for digital histopathology images. arXiv preprint arXiv:2506.05127 (2025)

work page arXiv 2025
[23]

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

Yu, X., Li, G., Lou, W., Liu, S., Wan, X., Chen, Y., Li, H.: Diffusion-based data augmentation for nuclei image segmentation. In: International Conference on Med- ical Image Computing and Computer-Assisted Intervention. pp. 592–602. Springer (2023)

2023
[24]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Yun, S., Han, D., Oh, S.J., Chun, S., Choe, J., Yoo, Y.: Cutmix: Regularization strategy to train strong classifiers with localizable features. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 6023–6032 (2019).https://doi.org/10.1109/ICCV.2019.00612

work page doi:10.1109/iccv.2019.00612 2019
[25]

Neuralremaster: Phase-preserving diffusion for structure-aligned generation.arXiv preprint arXiv:2512.05106, 2025

Zeng, Y., Ochoa, C., Zhou, M., Patel, V.M., Guizilini, V., McAllister, R.: Neuralre- master: Phase-preserving diffusion for structure-aligned generation. arXiv preprint arXiv:2512.05106 (2025)

work page arXiv 2025
[26]

In: Proceedings of the Computer Vision and Pattern Recogni- tion Conference

Zhang, Y., Liu, Z., Li, Z., Li, Z., Clark, J.J., Si, X.: Decoupling training-free guided diffusion by admm. In: Proceedings of the Computer Vision and Pattern Recogni- tion Conference. pp. 23292–23302 (2025)

2025
[27]

Medical image analysis p

Zhu, P., Liu, C., Fu, Y., Chen, N., Qiu, A.: Cycle-conditional diffusion model for noise correction of diffusion-weighted images using unpaired data. Medical image analysis p. 103579 (2025)

2025
[28]

arXiv preprint arXiv:2508.06625 (2025)

Zou, S., Huang, Y., Yi, R., Zhu, C., Xu, K.: Cyclediff: Cycle diffusion models for unpaired image-to-image translation. arXiv preprint arXiv:2508.06625 (2025)

work page arXiv 2025

[1] [1]

In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

Bansal, A., Chu, H.M., Schwarzschild, A., Sengupta, S., Goldblum, M., Geiping, J., Goldstein, T.: Universal guidance for diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 843–852 (2023)

2023

[2] [2]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision

Bhosale, M., Wasi, A., Zhai, Y., Tian, Y., Border, S., Xi, N., Sarder, P., Yuan, J., Doermann, D., Gong, X.: Pathdiff: Histopathology image synthesis with un- paired text and mask conditions. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 22415–22424 (2025)

2025

[3] [3]

In: 2022 IEEE 19th International sympo- sium on biomedical imaging (ISBI)

Butte, S., Wang, H., Xian, M., Vakanski, A.: Sharp-gan: Sharpness loss regularized gan for histopathology image synthesis. In: 2022 IEEE 19th International sympo- sium on biomedical imaging (ISBI). pp. 1–5. IEEE (2022)

2022

[4] [4]

Improved Regularization of Convolutional Neural Networks with Cutout

DeVries, T., Taylor, G.W.: Improved regularization of convolutional neural net- works with cutout. arXiv preprint arXiv:1708.04552 (2017),https://arxiv.org/ abs/1708.04552

work page internal anchor Pith review Pith/arXiv arXiv 2017

[5] [5]

Nature medicine pp

Ding, T., Wagner, S.J., Song, A.H., Chen, R.J., Lu, M.Y., Zhang, A., Vaidya, A.J., Jaume,G.,Shaban,M.,Kim,A.,etal.:Amultimodalwhole-slidefoundationmodel for pathology. Nature medicine pp. 1–13 (2025)

2025

[6] [6]

In: European Congress on Digital Pathology

Gamper, J., Koohbanani, N.A., Benet, K., Khuram, A., Rajpoot, N.: Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classifica- tion. In: European Congress on Digital Pathology. pp. 11–19. Springer (2019)

2019

[7] [7]

IEEE Transactions on Medical Imaging 39(12), 4124–4136 (2020)

Graham, S., Epstein, D., Rajpoot, N.: Dense steerable filter cnns for exploiting rotational symmetry in histology images. IEEE Transactions on Medical Imaging 39(12), 4124–4136 (2020)

2020

[8] [8]

Medical image analysis58, 101563 (2019)

Graham, S., Vu, Q.D., Raza, S.E.A., Azam, A., Tsang, Y.W., Kwak, J.T., Rajpoot, N.: Hover-net: Simultaneous segmentation and classification of nuclei in multi- tissue histology images. Medical image analysis58, 101563 (2019)

2019

[9] [9]

Signal processing: Image communication13(3), 171–181 (1998)

Huang, W.C., Chang, L.W.: Predictive subband image coding with wavelet trans- form. Signal processing: Image communication13(3), 171–181 (1998)

1998

[10] [10]

Nature methods18(2), 203–211 (2021)

Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H.: nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods18(2), 203–211 (2021)

2021

[11] [11]

Journal of pathology informatics 7(1), 29 (2016)

Janowczyk,A.,Madabhushi,A.:Deeplearningfordigitalpathologyimageanalysis: A comprehensive tutorial with selected use cases. Journal of pathology informatics 7(1), 29 (2016)

2016

[12] [12]

IEEE transactions on medical imaging43(3), 980–993 (2023)

Li, Y., Shao, H.C., Liang, X., Chen, L., Li, R., Jiang, S., Wang, J., Zhang, Y.: Zero-shot medical image translation via frequency-guided diffusion models. IEEE transactions on medical imaging43(3), 980–993 (2023)

2023

[13] [13]

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

[14] [14]

In: International conference on machine learning

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: International conference on machine learning. pp. 8748–8763. PmLR (2021) 10 Qiu et al

2021

[15] [15]

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

[16] [16]

Advances in neural information processing systems29(2016)

Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., Chen, X.: Improved techniques for training gans. Advances in neural information processing systems29(2016)

2016

[17] [17]

Nature Reviews Bioengineering1(12), 930–949 (2023)

Song, A.H., Jaume, G., Williamson, D.F., Lu, M.Y., Vaidya, A., Miller, T.R., Mah- mood, F.: Artificial intelligence for digital and computational pathology. Nature Reviews Bioengineering1(12), 930–949 (2023)

2023

[18] [18]

net/forum?id=PxTIG12RRHS

Song,J.,Meng,C.,Ermon,S.:Denoisingdiffusionimplicitmodels.In:International Conference on Learning Representations (ICLR) (2020),https://openreview. net/forum?id=PxTIG12RRHS

2020

[19] [19]

Patterns1(6) (2020)

Tschuchnig,M.E.,Oostingh,G.J.,Gadermayr,M.:Generativeadversarialnetworks in digital pathology: a survey on trends and future potential. Patterns1(6) (2020)

2020

[20] [20]

IEEE Transactions on Medical Imaging40(12), 3413–3423 (2021).https://doi.org/10.1109/TMI.2021.3085712

Verma, R., Kumar, N., Patil, A., Kurian, N.C., Rane, S., Graham, S., Vu, Q.D., Zwager, M., Raza, S.E.A., Rajpoot, N., Wu, X., Chen, H., Huang, Y., Wang, L., Jung, H., Brown, G.T., Liu, Y., Liu, S., Jahromi, S.A.F., Khani, A.A., Montahaei, E., Baghshah, M.S., Behroozi, H., Semkin, P., Rassadin, A., Dutande, P., Lodaya, R., Baid, U., Baheti, B., Talbar, S.,...

work page doi:10.1109/tmi.2021.3085712 2021

[21] [21]

arXiv preprint arXiv:2207.00050 (2022)

Wang, W., Bao, J., Zhou, W., Chen, D., Chen, D., Yuan, L., Li, H.: Semantic image synthesis via diffusion models. arXiv preprint arXiv:2207.00050 (2022)

work page arXiv 2022

[22] [22]

Knudsen, Tahsin Kurc, Rajarsi R

Yellapragada, S., Graikos, A., Li, Z., Triaridis, K., Belagali, V., Kapse, S., Nandi, T.N., Madduri, R.K., Prasanna, P., Kurc, T., et al.: Pixcell: A generative foun- dation model for digital histopathology images. arXiv preprint arXiv:2506.05127 (2025)

work page arXiv 2025

[23] [23]

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

Yu, X., Li, G., Lou, W., Liu, S., Wan, X., Chen, Y., Li, H.: Diffusion-based data augmentation for nuclei image segmentation. In: International Conference on Med- ical Image Computing and Computer-Assisted Intervention. pp. 592–602. Springer (2023)

2023

[24] [24]

In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

Yun, S., Han, D., Oh, S.J., Chun, S., Choe, J., Yoo, Y.: Cutmix: Regularization strategy to train strong classifiers with localizable features. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 6023–6032 (2019).https://doi.org/10.1109/ICCV.2019.00612

work page doi:10.1109/iccv.2019.00612 2019

[25] [25]

Neuralremaster: Phase-preserving diffusion for structure-aligned generation.arXiv preprint arXiv:2512.05106, 2025

Zeng, Y., Ochoa, C., Zhou, M., Patel, V.M., Guizilini, V., McAllister, R.: Neuralre- master: Phase-preserving diffusion for structure-aligned generation. arXiv preprint arXiv:2512.05106 (2025)

work page arXiv 2025

[26] [26]

In: Proceedings of the Computer Vision and Pattern Recogni- tion Conference

Zhang, Y., Liu, Z., Li, Z., Li, Z., Clark, J.J., Si, X.: Decoupling training-free guided diffusion by admm. In: Proceedings of the Computer Vision and Pattern Recogni- tion Conference. pp. 23292–23302 (2025)

2025

[27] [27]

Medical image analysis p

Zhu, P., Liu, C., Fu, Y., Chen, N., Qiu, A.: Cycle-conditional diffusion model for noise correction of diffusion-weighted images using unpaired data. Medical image analysis p. 103579 (2025)

2025

[28] [28]

arXiv preprint arXiv:2508.06625 (2025)

Zou, S., Huang, Y., Yi, R., Zhu, C., Xu, K.: Cyclediff: Cycle diffusion models for unpaired image-to-image translation. arXiv preprint arXiv:2508.06625 (2025)

work page arXiv 2025