Pith. sign in

REVIEW 3 major objections 3 minor 36 references

Towards 3D Semantic Image Synthesis for Medical Imaging

T0 review · 3 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read This paper claims that 3D medical volumes generated from semantic maps by a latent diffusion model are close enough to real scans to serve as privacy-preserving training data.

desk verdict Med-LSDM is a new 3D latent semantic synthesis pipeline, but its headline numbers rest on an unreliable FID estimate and the augmentation claim is untested. read the letter →

arxiv 2507.00206 v1 pith:KZU5DTHA submitted 2025-06-30 eess.IV cs.CV

classification eess.IVcs.CV
keywords 3DmedicalimagesynthesislatentdiffusionmodelsemanticVQ-GANdataaugmentationprivacypreservationdenoisingprobabilisticmodelsimaging
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 proposes Med-LSDM, a latent diffusion model that synthesizes full 3D medical volumes from semantic segmentation maps instead of generating 2D slices and stacking them. The authors aim to show that synthetic volumes generated this way lie close enough to real patient data to serve as privacy-preserving data augmentation. Their evidence is a 3D-FID of 0.0054 on the conditional Duke Breast MRI dataset and a downstream segmentation Dice score of 0.70964 on synthetic volumes versus 0.71496 on real test volumes. The practical promise is that institutions could share de-identified semantic maps and distribute synthetic images rather than raw patient scans, easing data scarcity and privacy restrictions.

What carries the argument

Med-LSDM couples a pre-trained 3D VQ-GAN, which compresses a 256x256x32 volume into a discrete latent code of size 64x64x8 with a 16384-entry codebook, with a semantic diffusion model that denoises Gaussian noise in that latent space conditioned on the semantic map. The conditioning is done by semantic diffusion decoder residual blocks that use SPADE-style spatially-adaptive normalization (modulating normalized feature maps with learned scale and shift from the semantic map). After denoising, the latent code is quantized against the frozen codebook and decoded to the voxel space, which is the step that carries the claim that 4x compression preserves clinically relevant structure.

What would settle it

Run the same segmentation network on real test volumes that have been passed through the VQ-GAN encoder and decoder without any diffusion; if the Dice score on these reconstructed volumes drops well below 0.715, then the 4x compression, not the diffusion model, is what limits anatomical fidelity, and the small-domain-gap result would not transfer to tasks needing finer detail.

Watch

Extended reading notes

Core claim

The paper's central claim is that running a denoising diffusion model inside the learned latent space of a 3D VQ-GAN, with the semantic map injected at multiple decoder stages via spatially-adaptive normalization, produces volumetric medical images whose distribution and anatomy match real data closely enough for downstream use. The authors report that segmentation performance on Med-LSDM-generated volumes (Dice 0.70964) is nearly identical to segmentation performance on real test volumes (Dice 0.71496), and far above the 2D slice-by-slice baseline (0.602). They position this as the first 3D semantic medical image synthesis in latent space, with the main remaining weakness being behavior in limited-data cross-modal settings like SynthRAD2023.

Load-bearing premise

The whole approach assumes that squeezing each volume down to a quarter of its original size keeps all clinically meaningful details; the paper shows only pictures to support this, so if fine structures get lost in that squeeze, the generated images will look smooth but cannot be trusted for clinical tasks.

Editorial extensions

If this is right

  • If correct, researchers can train segmentation networks on synthetic volumes and expect test-time behavior close to real-data training, making augmentation viable when real labels are scarce.
  • De-identified semantic maps become a shareable asset: only maps and synthetic volumes need leave an institution, reducing re-identification risk.
  • 3D consistency across slices is preserved, so volumetric analyses (coronal and sagittal views) avoid the inter-slice artifacts of stacked 2D generation.
  • On cross-modal tasks with small datasets, the method's advantage shrinks, so future work needs modality-aware conditioning or pretraining to close that gap.

Reading between the lines

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

  • A direct test the paper leaves implicit: measure VQ-GAN reconstruction fidelity quantitatively (e.g., Dice between original and reconstructed real volumes); if that reconstruction Dice is much lower than 0.71, the latent space itself, not the diffusion, may be the bottleneck.
  • The reported 3D-FID of 0.0054 is unusually low compared with typical FID scales, suggesting the Med3D feature distribution is narrow; a reader should compare models on the same feature extractor before treating absolute numbers as meaningful.
  • The method can be extended to unconditional or weakly supervised settings by replacing the semantic map with a predicted or partial label, which would broaden the privacy-preserving claim beyond datasets that already have annotations.
  • Another testable extension is an augmentation experiment where a segmentation model is trained on real plus synthetic volumes and compared against real-only training; the paper demonstrates equivalence of synthetic to real but does not directly demonstrate a gain from adding synthetic data.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes Med-LSDM, a 3D latent semantic diffusion model for medical image synthesis. A 3D VQ-GAN is first trained to compress volumetric images into a discrete latent space; a semantic diffusion model is then trained in that latent space, using SPADE modules conditioned on semantic maps to guide denoising. The authors evaluate the model on AutoPET, SynthRAD2023, and Duke Breast datasets, comparing against a 2D segmentation-guided diffusion baseline (SegGuidedDiff). The central claims are that Med-LSDM achieves a very small distributional gap with real data (3D-FID 0.0054 on Duke Breast) and that the synthetic data are useful for data augmentation, based on a Dice similarity of 0.70964 between a real-trained segmentation network's output on synthetic images and the input masks, versus 0.71496 on real test images.

Significance. If the claims are substantiated, this would be a useful contribution to 3D medical image synthesis, offering a way to generate volumetric images from semantic maps in latent space, which could reduce computational cost and support privacy-preserving data sharing and augmentation. The architecture is clearly described, the two-stage training (VQ-GAN followed by latent diffusion) is sound in principle, and the paper includes a baseline comparison and qualitative volumetric visualizations. However, the central quantitative evidence for the 'small domain gap' and 'useful for augmentation' claims is not yet convincing: the 3D-FID estimate is prone to severe bias at the reported sample size, no actual augmentation experiment is performed, and the VQ-GAN reconstruction fidelity is only qualitatively assessed. These issues need to be addressed before the main conclusions can be accepted.

major comments (3)
  1. [Section 4.1 and Section 4.4, Table 3] The 3D-FID results, including the headline value of 0.0054 on Duke Breast, are not reliable as reported. The experimental setup states that the Duke Breast test set contains only 10 scans, and the synthetic images are generated from the semantic maps of that same test set. If the FID is computed on these 10 real and 10 synthetic volumes using high-dimensional Med3D features, the plug-in Fréchet distance estimate is strongly biased toward zero. The paper does not report the number of volumes used for the FID computation nor the feature dimensionality. The stark contrast between the 2D FID (193.620) and 3D-FID (0.0054) on the same data is consistent with this artifact. Please report the exact sample size, feature dimension, and ideally a bootstrap confidence interval or an unbiased estimator. Without this, the 'small domain gap' conclusion (abstract, Section 5) is not established.
  2. [Section 4.2 and Abstract] The abstract and Section 5 claim that the synthetic data are 'useful for data augmentation,' but no augmentation experiment is conducted. Table 2 reports Dice scores of a segmentation network trained on real Duke Breast data when evaluated on synthetic images. This is a mask-faithfulness check, not a test of whether adding synthetic images to a training set improves downstream task performance. To support the augmentation claim, the authors should train segmentation networks on (i) real data only and (ii) real plus synthetic data, and compare performance on a held-out real test set. This is a standard and necessary experiment for the stated practical utility.
  3. [Section 4.3] The assumption that the VQ-GAN latent space 'can effectively compress spatial information while preserving its correspondence with the pixel space' is load-bearing for the entire generation chain, but it is supported only by visual inspection (Figure 7). No quantitative reconstruction error is reported (e.g., PSNR/SSIM on held-out volumes, or segmentation Dice on reconstructed images). Given that a 4x compression is used, the authors should provide quantitative evidence that clinically relevant structures survive the compression step. Without this, the possibility that the diffusion model operates on an overly smoothed or anatomically altered latent space cannot be ruled out.
minor comments (3)
  1. [Throughout] There are numerous typos and grammatical errors, e.g., 'Trasnlation' (Section 2 heading), 'Econder' (Section 3.3), 'comparation' (Figure 6), 'advant a' (Section 4.4), 'Med-LSTM' (should be 'Med-LSDM', Section 2), and inconsistent capitalization. A careful proofread is needed.
  2. [Table 1] The Duke Breast row lists '3D Scans (Train/Test)' as '(822 + 90)/10'. The text explains that 822 unlabeled scans are used for VQ-GAN training and 90 labeled scans for the diffusion model, but this notation is confusing. Please separate the two training subsets explicitly.
  3. [Table 2 and Abstract] The abstract reports a Dice score of 0.71496 for real test data, while Table 2 reports 0.715. Please make the values consistent, and report them to the same number of significant digits in both places.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central numerical claims are measured on generated outputs against independent benchmarks; no fitted constant is renamed as a prediction.

full rationale

The paper's derivation chain is empirical and self-contained. The VQ-GAN is trained by reconstruction (Eqs. 3-6), and the latent diffusion model is trained with the standard noise-prediction loss (Eq. 13); neither loss contains the evaluation quantities (3D-FID, Dice). The reported 3D-FID is computed with Med3D features pretrained on an external 3DSeg-8 dataset, and the Dice evaluation uses a segmentation network trained only on real Duke Breast training images and then applied to synthetic images generated from test-set semantic maps. Thus the target numbers are not built into the model's objective or into the conditioning masks by construction. The self-citations to the authors' earlier works (MedGAN [15], USIS [20]) appear only in the related-work discussion and are not load-bearing for the architecture or the evaluation. Concerns that n=10 test volumes make the 3D-FID estimate statistically unreliable, and that the data-augmentation utility claim is not directly tested by an augmentation experiment, are validity and support-strength issues rather than circularity, so they do not change the circularity score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard diffusion mathematics plus several domain assumptions about latent-space fidelity and evaluation proxies; no quantitative test of the core latent-preservation assumption is given.

free parameters (3)
  • compression factor t = 4
    Chosen in Section 4.3 as a hand-selected compromise between latent compactness and semantic preservation; the paper notes that 2x or 4x preserves structure while 8x loses detail, but no quantitative optimization is reported.
  • latent codebook size K and code dimension nz = K=16384, nz=8
    Set in Section 4.1 for the 3D VQ-GAN; the discrete latent capacity directly determines what information survives into the diffusion stage.
  • number of slices per volume = 32
    All volumes are cropped to (256,256,32,1) due to GPU memory limits in Section 4, so the method is tested only on shallow volumes, a scope limitation the authors acknowledge.
assumptions (5)
  • standard math Standard DDPM forward and reverse Gaussian noise assumptions and the simplified L2 noise-prediction objective (Equations 10-13).
    Borrowed from Ho et al. [7]; no new derivation is provided, so the framework inherits these assumptions.
  • domain assumption The VQ-GAN latent space preserves enough spatial and semantic information for the diffusion process to be meaningful.
    This is explicitly called an assumption in Section 4.3 and is only visually supported.
  • domain assumption A segmentation network trained on real Duke training volumes is a valid probe for distributional similarity of synthetic volumes.
    Section 4.2 uses Dice parity between real and synthetic segmentations as evidence of small domain gap; this assumes the segmenter is sensitive to domain shift and that mask agreement is not inflated by conditioning.
  • domain assumption TotalSegmentator-generated CT semantic maps are correct and useful for cross-modal training on SynthRAD2023.
    The paper uses automatic segmentations as ground-truth conditioning in Section 4, and the cross-modal failure is partly attributed to semantic mismatch.
  • domain assumption Med3D-pretrained features provide a meaningful and stable 3D-FID measure on 10 test volumes.
    3D-FID is computed with a Med3D feature extractor in Section 4.1; no confidence intervals or sample-size analysis are provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards 3D Semantic Image Synthesis for Medical Imaging." pith.science (2026). https://pith.science/paper/KZU5DTHA

@misc{pith2026250700206,
  author       = {Pith},
  title        = {Pith review of: Towards 3D Semantic Image Synthesis for Medical Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KZU5DTHA}},
  note         = {Machine review of arXiv:2507.00206}
}
read the original abstract

In the medical domain, acquiring large datasets is challenging due to both accessibility issues and stringent privacy regulations. Consequently, data availability and privacy protection are major obstacles to applying machine learning in medical imaging. To address this, our study proposes the Med-LSDM (Latent Semantic Diffusion Model), which operates directly in the 3D domain and leverages de-identified semantic maps to generate synthetic data as a method of privacy preservation and data augmentation. Unlike many existing methods that focus on generating 2D slices, Med-LSDM is designed specifically for 3D semantic image synthesis, making it well-suited for applications requiring full volumetric data. Med-LSDM incorporates a guiding mechanism that controls the 3D image generation process by applying a diffusion model within the latent space of a pre-trained VQ-GAN. By operating in the compressed latent space, the model significantly reduces computational complexity while still preserving critical 3D spatial details. Our approach demonstrates strong performance in 3D semantic medical image synthesis, achieving a 3D-FID score of 0.0054 on the conditional Duke Breast dataset and similar Dice scores (0.70964) to those of real images (0.71496). These results demonstrate that the synthetic data from our model have a small domain gap with real data and are useful for data augmentation.

Figures

Figures reproduced from arXiv: 2507.00206 by the authors.

Figure 1
Figure 1. Comparison of anonymous and non-anonymous data sharing [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Examples of semantic map (first row), original 2D MR image slices (second row) and 2D synthetic image [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the proposed Med-LSDM Framework. (a) Training of the VQ-GAN for mapping between voxel [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Encoder and Decoder Resblock of denosing U-Net [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: One 3D CT scan (top rows) and one MRT scan (bottom rows), which are generated by our model from the [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Comparation of segmentation results. The first column is the Ground Truth of semantic maps, the second [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: The corresponding 2D slices of reconstructed images [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: The examples of slices from coronal and sagittal planes from our model (top row) and SegGuidedDiff Model [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Conditional AutoPET progressive generation [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 9
Figure 9. Figure 9: Duke Breast (top columns in last page), AutoPET (bottoms columns in last page) and SynthRAD2023 (top [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 29 canonical work pages

  1. [1]

    Balasubramaniam .S, Prasanth Ap, Satheesh Kumar, and V . Kavitha. Medical image analysis based on deep learning approach for early diagnosis of diseases, 03 2024

  2. [2]

    van der Laak, Bram van Ginneken, and Clara I

    Geert Litjens, Thijs Kooi, Babak Ehteshami Bejnordi, Arnaud Arindra Adiyoso Setio, Francesco Ciompi, Mohsen Ghafoorian, Jeroen A.W.M. van der Laak, Bram van Ginneken, and Clara I. Sánchez. A survey on deep learning in medical image analysis. Medical Image Analysis, 42:60–88, December 2017

  3. [3]

    Transformers in medical image analysis: A review, 2022

    Kelei He, Chen Gan, Zhuoyuan Li, Islem Rekik, Zihao Yin, Wen Ji, Yang Gao, Qian Wang, Junfeng Zhang, and Dinggang Shen. Transformers in medical image analysis: A review, 2022

  4. [4]

    Deep learning techniques for medical image segmentation: Achievements and challenges

    Pim Moeskops, Max A Viergever, Juan M Iñesta, and Josien P Pluim. Deep learning techniques for medical image segmentation: Achievements and challenges. Journal of Imaging Informatics in Medicine, 10:123–139, 2022

  5. [5]

    Impact of quality, type and volume of data used by deep learning models in the analysis of medical images

    Andreea Roxana Luca, Tudor Florin Ursuleanu, and Liliana Gheorghe. Impact of quality, type and volume of data used by deep learning models in the analysis of medical images. Informatics in Medicine Unlocked, 29:100911, 2022

  6. [6]

    Challenges related to artificial intelligence research in medical imaging and the importance of image analysis competitions

    Luciano M Prevedello, Safwan S Halabi, George Shih, Carol C Wu, Marc D Kohli, Falgun H Chokshi, Bradley J Erickson, Jayashree Kalpathy-Cramer, Katherine P Andriole, and Adam E Flanders. Challenges related to artificial intelligence research in medical imaging and the importance of image analysis competitions. Radiology: Artificial Intelligence, 1(1):e180031, 2019

  7. [7]

    Denoising diffusion probabilistic models, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models, 2020

  8. [8]

    Shifting machine learning for healthcare from development to deployment and from models to data

    Angela Zhang, Lei Xing, James Zou, and Joseph C Wu. Shifting machine learning for healthcare from development to deployment and from models to data. Nature, 580(252-256), 2021

Show all 36 references
  1. [9]

    X. Chen, Y . Zhang, Y . Li, and C. Sun. The use of generative adversarial networks in medical image augmentation: A review. Neural Computing and Applications, 33:15227–15246, 2021

  2. [10]

    Health insurance portability and accountability act of 1996, August 21, 1996

    National Archives Office of the Federal Register and Records Administration. Health insurance portability and accountability act of 1996, August 21, 1996

  3. [11]

    Segment anything in medical images

    Jun Ma, Yuting He, Feifei Li, Lin Han, Chenyu You, and Bo Wang. Segment anything in medical images. Nature Communications, 15(1), 2024

  4. [12]

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A. Efros. Image-to-image translation with conditional adversarial networks, 2018

  5. [13]

    High-resolution image synthesis and semantic manipulation with conditional gans

    Ting-Chun Wang, Ming-Yu Liu, Jun-Yan Zhu, Andrew Tao, Jan Kautz, and Bryan Catanzaro. High-resolution image synthesis and semantic manipulation with conditional gans. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018

  6. [14]

    Taming transformers for high-resolution image synthesis, 2021

    Patrick Esser, Robin Rombach, and Björn Ommer. Taming transformers for high-resolution image synthesis, 2021

  7. [15]

    Medgan: Medical image translation using gans

    Karim Armanious, Chenming Jiang, Marc Fischer, Thomas Küstner, Tobias Hepp, Konstantin Nikolaou, Sergios Gatidis, and Bin Yang. Medgan: Medical image translation using gans. Computerized Medical Imaging and Graphics, 79:101684, January 2020

  8. [16]

    Dual diffusion implicit bridges for image-to-image translation, 2023

    Xuan Su, Jiaming Song, Chenlin Meng, and Stefano Ermon. Dual diffusion implicit bridges for image-to-image translation, 2023

  9. [17]

    Semantic image synthesis with spatially-adaptive normalization

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. Semantic image synthesis with spatially-adaptive normalization. CoRR, abs/1903.07291, 2019

  10. [18]

    Sean: Image synthesis with semantic region-adaptive normalization

    Peihao Zhu, Rameen Abdal, Yipeng Qin, and Peter Wonka. Sean: Image synthesis with semantic region-adaptive normalization. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, June 2020

  11. [19]

    Analyzing and improving the image quality of stylegan, 2020

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of stylegan, 2020. 14 A PREPRINT - S EPTEMBER 15, 2025

  12. [20]

    Usis: Unsupervised semantic image synthesis, 2021

    George Eskandar, Mohamed Abdelsamad, Karim Armanious, and Bin Yang. Usis: Unsupervised semantic image synthesis, 2021

  13. [21]

    Semantic image synthesis via diffusion models, 2022

    Weilun Wang, Jianmin Bao, Wengang Zhou, Dongdong Chen, Dong Chen, Lu Yuan, and Houqiang Li. Semantic image synthesis via diffusion models, 2022

  14. [22]

    High-resolution image synthesis with latent diffusion models, 2022

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models, 2022

  15. [23]

    Generative adversarial nets

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

  16. [24]

    Generative adversarial network in medical imaging: A review

    Xin Yi, Ekta Walia, and Paul Babyn. Generative adversarial network in medical imaging: A review. Medical Image Analysis, 58:101552, December 2019

  17. [25]

    Red-gan: Attacking class imbalance via conditioned generation

    Ahmad B Qasim, Ivan Ezhov, Suprosanna Shit, Oliver Schoppe, Johannes C Paetzold, Anjany Sekuboyina, Florian Kofler, Jana Lipkova, Hongwei Li, and Bjoern Menze. Red-gan: Attacking class imbalance via conditioned generation. yet another perspective on medical image synthesis for...

  18. [26]

    Mazurowski

    Nicholas Konz, Yuwen Chen, Haoyu Dong, and Maciej A. Mazurowski. Anatomically-controllable medical image generation with segmentation-guided diffusion models, 2024

  19. [27]

    Medical diffusion: Denoising diffusion probabilistic models for 3d medical image generation, 2023

    Firas Khader, Gustav Mueller-Franzes, Soroosh Tayebi Arasteh, Tianyu Han, Christoph Haarburger, Maximilian Schulze-Hagen, Philipp Schad, Sandy Engelhardt, Bettina Baessler, Sebastian Foersch, Johannes Stegmaier, Christiane Kuhl, Sven Nebelung, Jakob Nikolas Kather, and Daniel ...

  20. [28]

    Conditional diffusion models for semantic 3d brain mri synthesis

    Zolnamar Dorjsembe, Hsing-Kuo Pao, Sodtavilan Odonchimed, and Furen Xiao. Conditional diffusion models for semantic 3d brain mri synthesis. IEEE Journal of Biomedical and Health Informatics, 28(7):4084–4093, July 2024

  21. [29]

    Prajit Ramachandran, Barret Zoph, and Quoc V . Le. Searching for activation functions, 2017

  22. [30]

    Rectified linear units improve restricted boltzmann machines

    Vinod Nair and Geoffrey E Hinton. Rectified linear units improve restricted boltzmann machines. In Proceedings of the 27th international conference on machine learning (ICML-10), pages 807–814, 2010

  23. [31]

    A whole-body fdg-pet/ct dataset with manually annotated tumor lesions

    Sergios Gatidis, Tobias Hepp, Marcel Früh, Christian La Fougère, Konstantin Nikolaou, Christina Pfannenberg, Bernhard Schölkopf, Thomas Küstner, Clemens Cyran, and Daniel Rubin. A whole-body fdg-pet/ct dataset with manually annotated tumor lesions. Scientific Data, 9(1):601, 2022

  24. [32]

    Adrian Thummerer, Erik van der Bijl, Arthur Galapon, Joost J. C. Verhoeff, Johannes A. Langendijk, Stefan Both, Cornelis (Nico) A. T. van den Berg, and Matteo Maspero. Synthrad2023 grand challenge dataset: Generating synthetic ct for radiotherapy. Medical Physics, 50(7):4664–4...

  25. [33]

    Saha, M.R

    A. Saha, M.R. Harowicz, and L.J. Grimm. A machine learning approach to radiogenomics of breast cancer: a study of 922 subjects and 529 dce-mri features. Journal of Magnetic Resonance Imaging, 49:508–516, 2018

  26. [34]

    Meyer, Maurice Pradella, Daniel Hinck, Alexander W

    Jakob Wasserthal, Hanns-Christian Breit, Manfred T. Meyer, Maurice Pradella, Daniel Hinck, Alexander W. Sauter, Tobias Heye, Daniel T. Boll, Joshy Cyriac, Shan Yang, Michael Bach, and Martin Segeroth. Totalsegmentator: Robust segmentation of 104 anatomic structures in ct image...

  27. [35]

    Hierarchical amortized gan for 3d high resolution medical image synthesis

    Li Sun, Junxiang Chen, Yanwu Xu, Mingming Gong, Ke Yu, and Kayhan Batmanghelich. Hierarchical amortized gan for 3d high resolution medical image synthesis. IEEE Journal of Biomedical and Health Informatics , 26(8):3966–3975, 2022

  28. [36]

    Med3d: Transfer learning for 3d medical image analysis, 2019

    Sihong Chen, Kai Ma, and Yefeng Zheng. Med3d: Transfer learning for 3d medical image analysis, 2019. 15 A PREPRINT - S EPTEMBER 15, 2025 Appendix A Number of model parameters of our architecture Table 4: Trainable Parameters of VQ-GAN Components Name Encoder Decoder 2D-Dis 3D-...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.