Pith. sign in

REVIEW 3 major objections 5 minor 43 references

Joint Holistic and Lesion Controllable Mammogram Synthesis via Gated Conditional Diffusion Model

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Adding gated radiomic-geometric conditioning to a mask-guided diffusion model yields more realistic mammograms with tighter lesion control than prior methods.

desk verdict The SOTA claim in Table 1 is not credible as-is because GCDM is conditioned on radiomic features extracted from the real target image, an oracle input the baselines never receive; the underlying method is still a reasonable candidate. read the letter →

arxiv 2507.19201 v1 pith:4GJ67BNB submitted 2025-07-25 eess.IV cs.AIcs.CV

classification eess.IVcs.AIcs.CV
keywords mammogramsynthesislesion-controllablegenerationlatentdiffusionmodelgatedfusionmechanismradiomicstop-kselectionVinDr-Mammo
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

This paper claims that a diffusion model can synthesize full mammograms and their lesions together, with precise control over the lesion's shape and texture. The proposed Gated Conditional Diffusion Model (GCDM) feeds the denoiser a soft semantic mask of breast and lesion regions, and adds a gated conditioning branch that selects the most relevant radiomic and geometric features of the mass and injects them through cross-attention. On the VinDr-Mammo test set, GCDM reports an FID of 26.77 versus 30.50 for the best prior method and a mass IoU of 86.30% versus 84.02%, improvements the paper calls statistically significant. If correct, the result matters because realistic, lesion-controllable mammogram synthesis could serve as data augmentation for breast cancer screening models, where annotated data is scarce.

What carries the argument

The load-bearing mechanism is the gated fusion of radiomic and geometric features: radiomic features (67 dims: shape, histogram, GLSZM, GLCM) extracted by PyRadiomics and geometric features (768 dims) from a CLIP embedding of the mass mask are cross-concatenated into m×n candidate pairs; a relevance score w_gate = AVG(f_comb) ⊙ MLP(f_comb) weights each candidate, the top-k are selected, and a convolution produces the conditioning vector c that enters the UNet via cross-attention. This gate is what the paper claims lets the model dynamically pick the most relevant lesion properties instead of naively concatenating features. The second mechanism is the soft mask: Gaussian blurring the lesion channel of the three-class mask before encoding and concatenating it with the noisy latent, which the paper says models smooth transitions between lesion and surrounding tissue.

What would settle it

Re-run the comparison where GCDM is conditioned on radiomic features drawn from a template or from the conditioning mask alone, instead of from the real test image; if FID and Mass IoU then fall near the baselines' level, the gains attributed to gated fusion would instead come from test-time leakage of real-image radiomics. A second check: recompute Mass IoU using radiologist-annotated lesion boundaries on both real and generated images instead of masks produced by the automatic segmenter.

Watch

Extended reading notes

Core claim

GCDM is a latent diffusion model built on a fine-tuned Stable Diffusion v1.5. Its central claim is that combining two conditioning channels yields state-of-the-art mammogram synthesis: (1) a soft three-channel mask (background, breast, lesion) whose lesion channel is Gaussian-blurred, concatenated with the noised latent to enforce anatomical coherence; and (2) a gated 'Mass Control Branch' that extracts 67-dimensional radiomic features via the open-source PyRadiomics library and geometric features via a CLIP embedding of the mass mask, cross-concatenates them into candidate pairs, scores each pair by a gate (average pooling times an MLP), keeps the top-k, and feeds the result through cross-attention as key and value. On the VinDr-Mammo dataset (CC views only), the paper reports FID 26.77 (vs 30.50 second-best Seg-Diff), Mass IoU 86.30% (vs 84.02% SR3), Breast IoU 97.63%, and pixel accuracy 98.41%, with p-values 0.025 and 0.039 for the two headline gains. The ablation study attributes the gains to the lesion control branch, the radiomic features, and especially the gated fusion, each adding a step of improvement.

Load-bearing premise

The reported lesion-control accuracy assumes the automatically generated lesion masks used for both training and evaluation coincide with true lesion boundaries, and that using real test-image radiomics to condition generation is a fair comparison against baselines that do not get that information.

Editorial extensions

If this is right

  • Synthesized mammograms can be generated conditionally from a hand-drawn mask plus template radiomic features, without a real source image (Appendix A), enabling on-demand creation of cases with specified mass geometry and texture.
  • Training a benign/malignant classifier on real data augmented with GCDM-synthesized images improves held-out accuracy, AUC, and F1 for both ResNet-50 and ViT-B/16 backbones (Appendix B).
  • The choice of segmentation tool (MedSAM, GroundedSAM2, MedSAM2) has limited influence on the reported metrics, suggesting the improvement is in the synthesis model rather than the mask generator (Appendix C).
  • Soft labels with moderate blur (σ=1.5) balance image quality and lesion-boundary fidelity: stronger blur lowers FID but weakens the mass IoU (Table 3).
  • The improvements over prior methods reach statistical significance under the reported test protocol (p=0.025 for FID, p=0.039 for Mass IoU).

Reading between the lines

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

  • The same gated radiomics-geometry conditioning could transfer to other lesion-bearing imaging tasks (e.g., lung nodules in CT or colorectal polyps in CT colonography) where radiomic features are well-defined, provided a mask source exists.
  • A cleaner test of the gate's contribution would feed the same radiomic vector to a baseline via plain concatenation with the mask; the paper ablates the gate but not the raw-feature conditioning channel, so the residual gain of top-k selection over simple fusion is not isolated.
  • Because the conditioning vector encodes per-example radiomics from the target lesion, the model is effectively learning a conditional distribution; a natural extension is unconditional sampling from the learned radiomics prior to probe distribution coverage and mode collapse.
  • Since GCDM receives test-image radiomics during evaluation, its FID may partly reflect interpolation toward the real test distribution; an out-of-distribution evaluation (generating from template radiomics) would better measure true synthesis quality.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes GCDM, a latent diffusion framework for mammogram synthesis that combines whole-image anatomical conditioning with localized lesion control. The model concatenates a Gaussian-blurred three-channel mask (background, breast, lesion) with the noisy latent, and additionally injects a gated conditioning vector built from PyRadiomics features of the lesion and CLIP embeddings of the mass mask, fused through a top-k gating mechanism. Experiments on VinDr-Mammo report FID 26.77, Mass IoU 86.30%, Breast IoU 97.63%, and PA 98.41%, outperforming SPADE, pSp, SR3, ControlNet, and Seg-Diff. Ablations show progressive gains from the lesion control branch, radiomics features, and gated fusion. The paper also reports a downstream classification experiment and an ablation with alternative segmentation tools.

Significance. The idea of conditioning mammogram synthesis on radiomic and geometric lesion descriptors through a gated fusion mechanism is interesting and potentially useful for controllable medical image generation, especially if it enables clinicians to specify lesion texture and shape. The code is promised publicly, and the ablations are internally consistent. However, the main state-of-the-art claim is currently not established because the evaluation gives GCDM test-time access to radiomic features of the exact real target image, an input the baselines do not receive; the reported gains therefore conflate the proposed gating design with an oracle conditioning signal. Should a fair comparison confirm the advantages, the contribution would be solid, but that confirmation is necessary before the quantitative claims can be accepted.

major comments (3)
  1. [Sec. 3.3, Eqs. (7)-(8); Sec. 4.1; Table 1] At inference, the conditioning vector c is built from radiomic features f extracted with PyRadiomics from the real test mammogram x0 and the corresponding lesion mask, as described in Sec. 3.3. This gives GCDM per-test-image access to 67 target-specific descriptors (histogram, GLCM, GLSZM, shape) that none of the baselines receive; the baselines are conditioned only on masks. The reported FID improvement (26.77 vs. 30.50, p=0.025) and Mass IoU gain (86.30% vs. 84.02%, p=0.039) therefore cannot be attributed to the gated fusion design alone. To support the central claim, the authors should either provide the same radiomic features to all comparators, or evaluate GCDM quantitatively in the Appendix A manual mode (where a template radiomics vector, not the real test image's radiomics, is used). Without such a protocol, Table 1 does not establish the state-of-the-art claim.
  2. [Sec. 4.1, Fig. 2; Sec. 3.3] Mass IoU is computed between the real mask produced by MedSAM from the bounding-box annotation and the MedSAM mask of the generated image, while the same MedSAM-derived mask is also used to extract the radiomic conditioning features. If MedSAM systematically deviates from true lesion boundaries, the model is rewarded for reproducing those systematic deviations rather than clinically accurate borders. Appendix C shows that replacing MedSAM with GroundedSAM2 or MedSAM2 at inference leaves performance similar, which mitigates the concern about tool-specific artifacts, but it does not establish agreement with expert or pathology-aligned ground truth. The authors should either report Mass IoU against radiologist-annotated lesion boundaries or explicitly state and discuss this limitation in the main text.
  3. [Sec. 4.2, Table 1] The p-values 0.025 and 0.039 are reported without any description of the statistical test, the number of samples or bootstrap/permutation repeats, whether the comparison is paired, or whether multiple-comparison correction was applied. No standard deviations or confidence intervals are given for any metric. Unless the significance-testing procedure is specified and repeated-seed variability is reported, the statement that the improvements are statistically significant cannot be verified.
minor comments (5)
  1. [Sec. 4.1] The phrase "cranial-caudal oblique (CC)" is incorrect: CC stands for craniocaudal, which is not an oblique view; the oblique view is MLO (mediolateral oblique).
  2. [Table 2] The header and rows of Table 2 are inconsistent: the first row uses "× — —" while subsequent rows use check and cross marks for LCB, RF, and GF; the intended layout should be made uniform so each row indicates the presence or absence of all three components.
  3. [Sec. 3.3] The paper calls the CLIP embedding of the mass mask "geometric features," but CLIP is a semantic vision-language model not designed for geometric shape encoding; a brief justification of this choice would help the reader understand why CLIP is appropriate for geometric conditioning.
  4. [Appendix A] The manual radiomics mode is only evaluated qualitatively in Fig. 5; providing quantitative metrics (e.g., FID or Mass IoU) for this mode would strengthen the claim of practical controllability and would also directly address the asymmetry concern raised in the major comments.
  5. [Sec. 4.4, Table 3] The Gaussian blur variance is selected after observing test-set FID and Mass IoU, and the same test set is later used for the headline comparison in Table 1; this test-set-based hyperparameter selection should be acknowledged, or a validation split should be used for the selection.

Circularity Check

1 steps flagged · score 6.0 of 10

Table 1 SOTA gains are partially forced by test-time oracle conditioning: the conditioning vector c is computed from the real test image x0, and the FID is then computed against that same x0.

  1. fitted input called prediction [Sec. 3.3 (Eqs. 7-8), Sec. 4.1 (Eq. 9), Table 1]
    "To emphasize the radiomic characteristics of the generated mass, we extract radiomic features f from the lesion channel M_mass and the corresponding image x0 using PyRadiomics. ... In the experiments, FID is calculated between all real and generated images in the test set."

    At inference, the conditioning vector c in Eq. (8) is built from radiomic features f extracted from the real test image x0 (Sec. 3.3), so each GCDM sample is generated conditional on 67 target-image descriptors. Eq. (9) then computes FID between all real and generated test images, i.e., between x0 and samples that were conditioned on descriptors of that same x0. The reported 12.2% FID improvement and p=0.025 therefore compare GCDM against baselines that receive no such target-derived information, so the gain is not attributable to the gated fusion design; it is in part an artifact of giving GCDM the test-set answer key. The ablation in Table 2 has the same confound, since adding the branch is the step that introduces this target-derived input.

full rationale

The core diffusion derivation (Eqs. 1-8) is self-contained and not mathematically circular: the training loss is a standard noise-prediction objective, and the gated fusion module is a learned network with no fitted parameter later renamed as a prediction. There is no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation; the self-citation [38] is used only to justify the 256x256 resolution choice. The circular element is confined to the evaluation protocol. Because c is a function of the real test mammogram x0 (radiomics extracted with PyRadiomics), and the headline FID is computed against those same real images, the Table 1 comparison conflates the method with an oracle input that the baselines do not receive. The MedSAM-mask concern is real but substantially mitigated by Appendix C, which shows similar Mass IoU with GroundedSAM2 and MedSAM2. Overall, the derivation chain is not circular, but the central quantitative claim is partially forced by the evaluation design, so the score is 6 rather than 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard diffusion machinery plus a gated conditioning branch. The main ad hoc design elements are the soft-mask blur, the gated fusion formula, and the top-k selection. The evaluation further depends on the domain assumption that MedSAM masks and PyRadiomics features are valid clinical proxies. No physically new entities are introduced.

free parameters (6)
  • Gaussian blur sigma for soft mask = 1.5
    Chosen by hand in Sec 4.4 based on the trade-off between FID and Mass IoU in Table 3; not derived. Affects the soft boundary between lesion and tissue, directly influencing the conditioning input.
  • Gated fusion dimensions m, n, k = 5, 5, 5
    Set in Implementation Details (Sec 4.1); control the number of geometric and radiomic candidate features and the top-k selection. No sensitivity analysis is provided.
  • CFG guidance scale = 7.5
    Set in Implementation Details for inference; standard classifier-free guidance scale.
  • CFG condition dropout probability = 0.1
    Set in Implementation Details; probability of randomly masking conditions during training.
  • Diffusion steps T / inference steps = 1000 / 50
    Standard DDPM/DDIM settings in Sec 4.1.
  • Number of training epochs, batch size, learning rate = 400, 32 per GPU, 1e-4
    Optimization settings in Sec 4.1; no sensitivity analysis.
assumptions (5)
  • standard math DDPM/DDIM forward and reverse processes (Eqs. 1-3) provide a valid generative model for the latent space.
    Adopted from Ho et al. (Sec 3.1) as the backbone of training and inference.
  • domain assumption The Stable Diffusion VAE encodes and decodes mammograms in a 32x32 latent space without significant information loss for the target task.
    The method relies on fine-tuning Stable Diffusion v1.5 on mammograms (Sec 3.2, 4.1); if the VAE cannot faithfully represent mammogram texture, the synthesis quality is capped regardless of the conditioning branch.
  • domain assumption Automatic MedSAM segmentations seeded by dataset bounding boxes are valid ground-truth lesion masks for both conditioning and evaluation.
    Mass IoU and conditioning use these masks (Sec 4.1, Fig 2); if they are systematically biased, the lesion-control claim is inflated.
  • domain assumption PyRadiomics features (shape, histogram, GLSZM, GLCM), after Min-Max normalization, capture clinically meaningful lesion characteristics that can be used as generation conditions.
    The entire mass control branch encodes these features as conditioning (Sec 3.3); no evidence is provided that these features are sufficient or faithful for controlling synthesis.
  • ad hoc to paper The gated fusion formulation (Eqs. 6-7) with AVG/MLP product and top-k selection improves conditioning without a formal justification.
    The design is introduced in Sec 3.3 based on intuition; its effectiveness is only validated empirically through the ablation study, not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Joint Holistic and Lesion Controllable Mammogram Synthesis via Gated Conditional Diffusion Model." pith.science (2026). https://pith.science/paper/4GJ67BNB

@misc{pith2026250719201,
  author       = {Pith},
  title        = {Pith review of: Joint Holistic and Lesion Controllable Mammogram Synthesis via Gated Conditional Diffusion Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GJ67BNB}},
  note         = {Machine review of arXiv:2507.19201}
}
read the original abstract

Mammography is the most commonly used imaging modality for breast cancer screening, driving an increasing demand for deep-learning techniques to support large-scale analysis. However, the development of accurate and robust methods is often limited by insufficient data availability and a lack of diversity in lesion characteristics. While generative models offer a promising solution for data synthesis, current approaches often fail to adequately emphasize lesion-specific features and their relationships with surrounding tissues. In this paper, we propose Gated Conditional Diffusion Model (GCDM), a novel framework designed to jointly synthesize holistic mammogram images and localized lesions. GCDM is built upon a latent denoising diffusion framework, where the noised latent image is concatenated with a soft mask embedding that represents breast, lesion, and their transitional regions, ensuring anatomical coherence between them during the denoising process. To further emphasize lesion-specific features, GCDM incorporates a gated conditioning branch that guides the denoising process by dynamically selecting and fusing the most relevant radiomic and geometric properties of lesions, effectively capturing their interplay. Experimental results demonstrate that GCDM achieves precise control over small lesion areas while enhancing the realism and diversity of synthesized mammograms. These advancements position GCDM as a promising tool for clinical applications in mammogram synthesis. Our code is available at https://github.com/lixinHUST/Gated-Conditional-Diffusion-Model/

Figures

Figures reproduced from arXiv: 2507.19201 by the authors.

Figure 1
Figure 1. An overview of the proposed Gated Conditional Diffusion Model (GCDM) for precise and effective mammogram [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The flowchart of quantitative evaluation. We con [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Visual results on the VinDr-Mammo dataset. We compare our GCDM with several SOTA methods, [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualization of breast mask softness variations. Each column demonstrates progressively increasing degrees of [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Controllable mammogram synthesis results with [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 18 canonical work pages

  1. [1]

    Wenting Chen, Pengyu Wang, Hui Ren, Lichao Sun, Quanzheng Li, Yixuan Yuan, and Xiang Li. 2024. Medical image synthesis via fine-grained image-text Mammogram Synthesis via Gated Conditional Diffusion Model MM ’25, October 27–31, 2025, Dublin, Ireland alignment and anatomy-pathology prompting. In International Conference on Medical Image Computing and Compu...

  2. [2]

    Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems 34 (2021), 8780–8794

  3. [3]

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative adversarial nets. Advances in neural information processing systems 27 (2014)

  4. [4]

    Anees Ur Rehman Hashmi, Ibrahim Almakky, Mohammad Areeb Qazi, Santosh Sanjeev, Vijay Ram Papineni, Jagalpathy Jagdish, and Mohammad Yaqub. 2024. Xreal: Realistic anatomy and pathology-aware x-ray generation via controllable diffusion model. arXiv preprint arXiv:2403.09240 (2024)

  5. [5]

    Yang Heng, Ma Yinghua, Fiaz Gul Khan, Ahmad Khan, Farman Ali, Ahmad Ali AlZubi, and Zeng Hui. 2024. Survey: application and analysis of generative adversarial networks in medical images. Artificial Intelligence Review 58, 2 (2024), 39

  6. [6]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33 (2020), 6840–6851

  7. [7]

    Hedayatollah Hosseini, Milan MS Obradović, Martin Hoffmann, Kathryn L Harper, Maria Soledad Sosa, Melanie Werner-Klein, Lahiri Kanth Nanduri, Chris- tian Werno, Carolin Ehrl, Matthias Maneck, et al. 2016. Early dissemination seeds metastasis in breast cancer. Nature 540, 7634 (2016), 552–558

  8. [8]

    Jie Hu, Li Shen, and Gang Sun. 2018. Squeeze-and-excitation networks. InProceed- ings of the IEEE conference on computer vision and pattern recognition . 7132–7141

Show all 43 references
  1. [9]

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. 2017. Image-to- image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition . 1125–1134

  2. [10]

    Annie Julie Joseph, Priyansh Dwivedi, Jiffy Joseph, Seenia Francis, Pournami PN, Jayaraj PB, Ashna V Shamsu, and Praveen Sankaran. 2024. Prior-guided generative adversarial network for mammogram synthesis. Biomedical Signal Processing and Control 87 (2024), 105456

  3. [11]

    Sharif Amit Kamran, Khondker Fariha Hossain, Alireza Tavakkoli, Stewart Lee Zuckerbrod, and Salah A Baker. 2021. Vtgan: Semi-supervised retinal image synthesis and disease prediction using vision transformers. In Proceedings of the IEEE/CVF international conference on computer...

  4. [12]

    Tero Karras. 2019. A Style-Based Generator Architecture for Generative Adver- sarial Networks. arXiv preprint arXiv:1812.04948 (2019)

  5. [13]

    Diederik P Kingma. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

  6. [14]

    Nicholas Konz, Yuwen Chen, Haoyu Dong, and Maciej A Mazurowski. 2024. Anatomically-controllable medical image generation with segmentation-guided diffusion models. In International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 88–98

  7. [15]

    Ruoshi Liu, Rundi Wu, Basile Van Hoorick, Pavel Tokmakov, Sergey Zakharov, and Carl Vondrick. 2023. Zero-1-to-3: Zero-shot One Image to 3D Object. arXiv:2303.11328 [cs.CV]

  8. [16]

    Chia-Feng Lu, Fei-Ting Hsu, Kevin Li-Chun Hsieh, Yu-Chieh Jill Kao, Sho-Jen Cheng, Justin Bo-Kai Hsu, Ping-Huei Tsai, Ray-Jade Chen, Chao-Ching Huang, Yun Yen, et al. 2018. Machine learning–based radiomics for molecular subtyping of gliomas. Clinical Cancer Research 24, 18 (20...

  9. [17]

    Ricardo Montoya-del Angel, Karla Sam-Millan, Joan C Vilanova, and Robert Martí. 2024. MAM-E: Mammographic synthetic image generation with diffusion models. Sensors 24, 7 (2024), 2076

  10. [18]

    SM Moss, Lennarth Nyström, Hakan Jonsson, E Paci, E Lynge, S Njor, and M Broeders. 2012. The impact of mammographic screening on breast cancer mortal- ity in Europe: a review of trend studies. Journal of medical screening 19, 1_suppl (2012), 26–32

  11. [19]

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. 2024. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 4296–4304

  12. [20]

    Inye Na, Jonghun Kim, Eun Sook Ko, and Hyunjin Park. 2024. RadiomicsFill- Mammo: Synthetic Mammogram Mass Manipulation with Radiomics Features. In International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 723–733

  13. [21]

    Hieu T Nguyen, Ha Q Nguyen, Hieu H Pham, Khanh Lam, Linh T Le, Minh Dao, and Van Vu. 2023. VinDr-Mammo: A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography. Scientific Data 10, 1 (2023), 277

  14. [22]

    Junzhi Ning, Cheng Tang, Kaijin Zhou, Diping Song, Lihao Liu, Ming Hu, Wei Li, Yanzhou Su, Tianbing Li, Jiyao Liu, et al. 2025. RetinaLogos: Fine-Grained Synthesis of High-Resolution Retinal Images Through Captions. arXiv preprint arXiv:2505.12887 (2025)

  15. [23]

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. 2019. Semantic image synthesis with spatially-adaptive normalization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2337–2346

  16. [24]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learni...

  17. [25]

    Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kunchang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, et al. 2024. Grounded sam: Assembling open-world models for diverse visual tasks. arXiv preprint arXiv:2401.14159 (2024)

  18. [26]

    Elad Richardson, Yuval Alaluf, Or Patashnik, Yotam Nitzan, Yaniv Azar, Stav Shapiro, and Daniel Cohen-Or. 2021. Encoding in style: a stylegan encoder for image-to-image translation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2287–2296

  19. [27]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 10684–10695

  20. [28]

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. 2022. Image super-resolution via iterative refinement. IEEE transactions on pattern analysis and machine intelligence 45, 4 (2022), 4713–4726

  21. [29]

    Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020)

  22. [30]

    Zizhao Sun, Huiqin Jiang, Ling Ma, Zhan Yu, and Hongwei Xu. 2022. Transformer based multi-view network for mammographic image classification. In Interna- tional conference on medical image computing and computer-assisted intervention . Springer, 46–54

  23. [31]

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. 2016. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition . 2818–2826

  24. [32]

    Joost JM Van Griethuysen, Andriy Fedorov, Chintan Parmar, Ahmed Hosny, Nicole Aucoin, Vivek Narayan, Regina GH Beets-Tan, Jean-Christophe Fillion- Robin, Steve Pieper, and Hugo JWL Aerts. 2017. Computational radiomics system to decode the radiographic phenotype. Cancer researc...

  25. [33]

    A Vaswani. 2017. Attention is all you need. Advances in Neural Information Processing Systems (2017)

  26. [34]

    Guotai Wang, Xiangde Luo, Ran Gu, Shuojue Yang, Yijie Qu, Shuwei Zhai, Qianfei Zhao, Kang Li, and Shaoting Zhang. 2023. PyMIC: A deep learning toolkit for annotation-efficient medical image segmentation. Computer Methods and Programs in Biomedicine 231 (2023), 107398

  27. [35]

    Shanshan Wang, Cheng Li, Rongpin Wang, Zaiyi Liu, Meiyun Wang, Hongna Tan, Yaping Wu, Xinfeng Liu, Hui Sun, Rui Yang, et al. 2021. Annotation-efficient deep learning for automatic medical image segmentation.Nature communications 12, 1 (2021), 5915

  28. [36]

    Ting-Chun Wang, Ming-Yu Liu, Jun-Yan Zhu, Andrew Tao, Jan Kautz, and Bryan Catanzaro. 2018. High-resolution image synthesis and semantic manipulation with conditional gans. In Proceedings of the IEEE conference on computer vision and pattern recognition. 8798–8807

  29. [37]

    Weilun Wang, Jianmin Bao, Wengang Zhou, Dongdong Chen, Dong Chen, Lu Yuan, and Houqiang Li. 2022. Semantic image synthesis via diffusion models. arXiv preprint arXiv:2207.00050 (2022)

  30. [38]

    Zhiwei Wang, Junlin Xian, Kangyi Liu, Xin Li, Qiang Li, and Xin Yang. 2023. Dual-view correlation hybrid attention network for robust holistic mammogram classification. arXiv preprint arXiv:2306.10676 (2023)

  31. [39]

    McKell Woodland, John Wood, Brian M Anderson, Suprateek Kundu, Ethan Lin, Eugene Koay, Bruno Odisio, Caroline Chung, Hyunseon Christine Kang, Aradhana M Venkatesan, et al. 2022. Evaluating the performance of StyleGAN2- ADA on medical images. In International Workshop on Simula...

  32. [40]

    Zhaohu Xing, Sicheng Yang, Sixiang Chen, Tian Ye, Yijun Yang, Jing Qin, and Lei Zhu. 2024. Cross-conditioned diffusion model for medical image to im- age translation. In International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 201–211

  33. [41]

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. 2023. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721 (2023)

  34. [42]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding conditional con- trol to text-to-image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision . 3836–3847

  35. [43]

    Jiayuan Zhu, Abdullah Hamdi, Yunli Qi, Yueming Jin, and Junde Wu. 2024. Medi- cal sam 2: Segment medical images as video via segment anything model 2.arXiv preprint arXiv:2408.00874 (2024). MM ’25, October 27–31, 2025, Dublin, Ireland Xin Li, Kaixiang Yang, Qiang Li, Zhiwei Wa...

Pith tools

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