REVIEW 2 major objections 4 minor 20 references
Ambient Denoising Diffusion Generative Adversarial Networks for Establishing Stochastic Object Models from Noisy Image Data
T0 review · 2 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash
Pith's one-line read The paper proposes ADDGAN, which trains a denoising diffusion GAN through the imaging operator, so stochastic object models can be learned from noisy CT and DBT data without clean ground truth.
desk verdict ADDGAN is a sensible combination of DDGAN and AmbientGAN with good empirical results, but the generator's omission of measurement noise in its simulated measurements is a load-bearing flaw that biases the learned object distribution. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is the ambient posterior-sampling identity in Eq. (5): pθ(y_{t-1} | y_t) = ∫ p(z) q(y_{t-1} | y_t, H(Gθ(x_t, z, t))) dz. It replaces the clean image used in standard DDGAN denoising with the simulated measurement of the generated object, enforcing consistency between the synthesized object and the observed noisy data. For CT, H includes the Radon transform and FBP reconstruction; for DBT in this study, H is additive Gaussian noise. This identity is what allows the generator to be trained without clean target images.
What would settle it
Train ADDGAN on CT data whose sinogram noise is Poisson-distributed rather than the i.i.d. Gaussian used in the paper, then measure the Hotelling-observer AUC on generated versus ground-truth patches; if the AUC gap grows or the discriminator can reliably tell real from simulated measurements, the ambient noise model is insufficient and the 'noise-independent SOM' claim fails for realistic noise.
Extended reading notes
Core claim
ADDGAN extends the DDGAN framework by inserting the imaging system into the denoising loop. At each reverse step the generator produces a candidate object x̂0, and the known measurement operator H (for CT, Radon transform followed by filtered back-projection) converts it into a simulated noisy image ŷ0 that conditions the posterior sampling. Training therefore uses only the measured noisy images y0, yet the final output is an object-space sample. In the reported experiments, ADDGAN achieves FID of 30.34 on CT and 41.13 on DBT, outperforming ProAmGAN (41.80) and Ambient StyleGAN3 (70.04), and its generated DBT patches produce SSIM and Hotelling-observer (a standard linear signal-detection template) statistics much closer to ground truth than the comparison models.
Load-bearing premise
The argument rests on assuming that the simulated measurements H(x̂0) generated inside the network are statistically indistinguishable, including their noise, from the real measured images y0; for CT the real noise enters the sinogram and is never sampled in the generator path, so if a discriminator can detect that difference, the learned object distribution is biased.
Editorial extensions
If this is right
- ADDGAN can establish SOMs from raw measured image data, so task-based image quality studies no longer require clean or denoised training images.
- The same architecture inherits DDGAN's fast sampling (four to eight diffusion steps in these experiments), avoiding the slow generation of standard diffusion models.
- Generated objects from ADDGAN are, by the paper's metrics, statistically closer to ground-truth anatomy than those from AmbientGAN models, particularly in high-resolution textures.
- The Hotelling-observer comparison indicates that ADDGAN-produced backgrounds support signal-detection tasks with performance similar to real backgrounds, a prerequisite for using the SOM in virtual clinical trials.
Reading between the lines
- The ambient loop should transfer to other computed imaging systems (PET, MRI, ultrasound) whenever the forward operator is differentiable and the noise process can be simulated; the paper only demonstrates CT and DBT.
- If real measurement noise is signal-dependent or correlated, the simple additive or Gaussian noise models used here may need to be replaced by a learned noise model, otherwise ADDGAN could memorize noise statistics rather than object statistics.
- A natural next test is to use ADDGAN-generated SOMs as backgrounds for virtual clinical trials and compare the resulting task-based image quality rankings with physical phantom measurements.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces ADDGAN, an ambient variant of the denoising diffusion GAN (DDGAN), intended to learn stochastic object models (SOMs) from noisy medical image measurements. The generator synthesizes an object, maps it through a known imaging operator (with reconstruction for CT) to produce simulated measurements, and uses these in the posterior sampling of the denoising diffusion process. Experiments on a stylized CT dataset (DeepLesion) and a DBT dataset (BCS-DBT) compare ADDGAN against DDPM, DDGAN, Ambient StyleGAN3, and ProAmGAN, using FID and Hotelling observer AUC. The paper reports that ADDGAN achieves the best FID and yields AUC closest to ground truth.
Significance. If correct, the paper would provide a fast, high-quality method for estimating object statistics from noisy clinical data, an important goal for task-based image quality assessment. The use of a diffusion GAN as the backbone is well-motivated, and the empirical results are promising. However, the central modeling assumption—that the generator's simulated measurement properly represents the real noisy measurement—is not established, so the significance is conditional on resolving that issue.
major comments (2)
- [Section 2.3 (Eq. 5) and Section 3.1 (Eq. 6)] The generator's simulated measurement ŷ0 = H(x̂0) omits the additive measurement noise n that defines the real y0. In the CT study, real y0 = FBP(Rf + n), while simulated ŷ0 = FBP(Rx̂0); in the DBT study, real y0 = x0 + n, while simulated ŷ0 = x̂0. Because the discriminator is trained on real noisy y0 and noiseless simulated ŷ0, it can distinguish them based solely on the presence of measurement noise. The generator may then encode the missing noise into the object x̂0, biasing the learned SOM. This is a load-bearing flaw for the paper's claim of learning SOMs from noisy data. Please either modify the forward simulation to sample measurement noise from the known noise model, or provide a theoretical argument or ablation study showing the omission is harmless.
- [Section 3.2 (Tables 1, 2 and Fig. 5)] The evaluation measures generated objects against ground-truth clean objects, which is appropriate for SOM assessment. However, because the generator's simulated measurements are noiseless, the reported FID improvement may stem from the generator producing noise-free objects rather than correctly inverting the stochastic measurement process. To substantiate the ambient claim, please report a distributional comparison between real y0 and model-simulated y0 (e.g., Wasserstein distance in sinogram/image space), not only x0-based metrics.
minor comments (4)
- [Eq. (5)] In the last expression of Eq. (5), the argument of G is written as Gθ(xt, z, t), but in the ambient setting it should be yt to match the preceding definitions; please correct this typo.
- [Section 3.1] For the CT study, the sinogram noise standard deviation is set to 1 while the object intensities are normalized to [0,1]; please clarify how this noise level was chosen and report the resulting noise level in the reconstructed image domain.
- [Figures 2 and 3] The captions contain a typo, 'Denoisng' instead of 'Denoising'; please also make the caption labels consistent with the subfigures.
- [Conclusions and broader context] The paper assumes the imaging operator H (and reconstruction operator O) is exactly known and differentiable; a brief discussion of robustness to operator mismatch would be valuable given the method's reliance on this assumption.
Circularity Check
No circularity found: ADDGAN is validated against independent ground-truth images and external task-based metrics, and the self-citations are background rather than load-bearing.
full rationale
The ADDGAN derivation is self-contained. The generator parametrizes the reverse denoising distribution in Eq. (5) by posterior sampling conditioned on H(x̂0), which is the standard ambient-model inversion of the measurement operator. The central claims are evaluated by FID computed against held-out real CT and DBT images, SSIM patch statistics, and a Hotelling observer signal-detection task using ground-truth and model-produced patches. These benchmarks are independent of the training loss and of any fitted parameter, so the reported superiority over DDPM, DDGAN, ProAmGAN, and Ambient StyleGAN3 is an empirical result rather than a construction. The paper cites the same group's earlier AmbientGAN work (refs. 2-4 and 9-11) for background and baseline methods, but those citations are not used to justify the ADDGAN formulation or to forbid alternatives; the architecture is an extension of the external DDGAN framework (ref. 13). The failure to sample measurement noise in the generator path, noted in the CT and DBT studies, is a potential correctness limitation concerning statistical mismatch between the simulated and true measurement operators, not a circularity: no equation in the paper reduces to its own input, and no fitted quantity is renamed as a prediction.
Assumptions & free parameters
free parameters (4)
- Number of diffusion steps =
4 for CT, 8 for DBT
- Adam learning rates =
2e-4 (CT); DBT generator 1.6e-5, discriminator 1e-5
- Sinogram Gaussian noise standard deviation (CT) =
1
- DBT Gaussian noise standard deviation =
0.06
assumptions (4)
- ad hoc to paper The generator can learn the object distribution from the denoising objective even though the simulated y0 = H(x0) in Eq. (5) does not include the additive measurement noise n present in Eq. (6).
- domain assumption The imaging operator H and reconstruction operator O (Radon transform and FBP for CT) are known, differentiable, and accurate enough for posterior sampling.
- domain assumption The DeepLesion CT slices and BCS-DBT images, after resizing and normalization, are representative ensembles of to-be-imaged objects.
- standard math Standard adversarial training converges to a meaningful equilibrium in the GAN and DDGAN settings.
Cite this review
Pith. "Pith review of Ambient Denoising Diffusion Generative Adversarial Networks for Establishing Stochastic Object Models from Noisy Image Data." pith.science (2026). https://pith.science/paper/NOYVHOPV
@misc{pith2026250119094,
author = {Pith},
title = {Pith review of: Ambient Denoising Diffusion Generative Adversarial Networks for Establishing Stochastic Object Models from Noisy Image Data},
year = {2026},
howpublished = {\url{https://pith.science/paper/NOYVHOPV}},
note = {Machine review of arXiv:2501.19094}
}
read the original abstract
It is widely accepted that medical imaging systems should be objectively assessed via task-based image quality (IQ) measures that ideally account for all sources of randomness in the measured image data, including the variation in the ensemble of objects to be imaged. Stochastic object models (SOMs) that can randomly draw samples from the object distribution can be employed to characterize object variability. To establish realistic SOMs for task-based IQ analysis, it is desirable to employ experimental image data. However, experimental image data acquired from medical imaging systems are subject to measurement noise. Previous work investigated the ability of deep generative models (DGMs) that employ an augmented generative adversarial network (GAN), AmbientGAN, for establishing SOMs from noisy measured image data. Recently, denoising diffusion models (DDMs) have emerged as a leading DGM for image synthesis and can produce superior image quality than GANs. However, original DDMs possess a slow image-generation process because of the Gaussian assumption in the denoising steps. More recently, denoising diffusion GAN (DDGAN) was proposed to permit fast image generation while maintain high generated image quality that is comparable to the original DDMs. In this work, we propose an augmented DDGAN architecture, Ambient DDGAN (ADDGAN), for learning SOMs from noisy image data. Numerical studies that consider clinical computed tomography (CT) images and digital breast tomosynthesis (DBT) images are conducted. The ability of the proposed ADDGAN to learn realistic SOMs from noisy image data is demonstrated. It has been shown that the ADDGAN significantly outperforms the advanced AmbientGAN models for synthesizing high resolution medical images with complex textures.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
Barrett, H. H. and Myers, K. J., [Foundations of image science ], John Wiley & Sons (2013)
work page 2013
-
[2]
Zhou, W., Li, H., and Anastasio, M. A., “Approximating the Ideal Observer and Hotelling Observer for binary signal detection tasks by use of supervised learning methods,”IEEE Transactions on Medical Imag- ing 38(10), 2456–2468 (2019)
work page 2019
-
[3]
Zhou, W., Li, H., and Anastasio, M. A., “Approximating the ideal observer for joint signal detection and localization tasks by use of supervised learning methods,”IEEE transactions on medical imaging 39(12), 3992–4000 (2020)
work page 2020
-
[4]
Ideal observer computation by use of markov-chain monte carlo with generative adversarial networks,
Zhou, W., Villa, U., and Anastasio, M. A., “Ideal observer computation by use of markov-chain monte carlo with generative adversarial networks,”IEEE transactions on medical imaging (2023)
work page 2023
-
[5]
An ideal observer for a model of x-ray imaging in breast parenchymal tissue,
Abbey, C. K. and Boone, J. M., “An ideal observer for a model of x-ray imaging in breast parenchymal tissue,” in [Digital Mammography: 9th International Workshop, IWDM 2008 Tucson, AZ, USA, July 20-23, 2008 Proceedings 9], 393–400, Springer (2008)
work page 2008
-
[6]
Effect of random background inhomogeneity on observer detection performance,
Rolland, J. P. and Barrett, H. H., “Effect of random background inhomogeneity on observer detection performance,” JOSA A 9(5), 649–658 (1992)
work page 1992
-
[7]
Experimental determination of object statistics from noisy images,
Kupinski, M. A., Clarkson, E., Hoppin, J. W., Chen, L., and Barrett, H. H., “Experimental determination of object statistics from noisy images,”JOSA A 20(3), 421–429 (2003)
work page 2003
-
[8]
Ambientgan: Generative models from lossy measurements,
Bora, A., Price, E., and Dimakis, A. G., “Ambientgan: Generative models from lossy measurements,” in [International conference on learning representations ], (2018)
work page 2018
Show all 20 references
-
[9]
Learning stochastic object models from medical imaging measurements by use of advanced ambient generative adversarial networks,
Zhou, W., Bhadra, S., Brooks, F. J., Li, H., and Anastasio, M. A., “Learning stochastic object models from medical imaging measurements by use of advanced ambient generative adversarial networks,”Journal of Medical Imaging 9(1), 015503–015503 (2022)
2022
-
[10]
Ambientcyclegan for establishing interpretable stochastic object models based on mathematical phantoms and medical imaging measurements,
Xu, X., Chen, W., and Zhou, W., “Ambientcyclegan for establishing interpretable stochastic object models based on mathematical phantoms and medical imaging measurements,” in [Medical Imaging 2024: Image Perception, Observer Performance, and Technology Assessment ], 12929, 234–...
2024
-
[11]
Ambient-pix2pixgan for translating medical images from noisy data,
Chen, W., Xu, X., Luo, J., and Zhou, W., “Ambient-pix2pixgan for translating medical images from noisy data,” in [Medical Imaging 2024: Image Perception, Observer Performance, and Technology Assessment ], 12929, 83–89, SPIE (2024)
2024
-
[12]
Assessing the capacity of a denoising diffusion probabilistic model to reproduce spatial context,
Deshpande, R., Özbey, M., Li, H., Anastasio, M. A., and Brooks, F. J., “Assessing the capacity of a denoising diffusion probabilistic model to reproduce spatial context,”IEEE Transactions on Medical Imaging (2024)
2024
-
[13]
Tackling the generative learning trilemma with denoising diffusion gans,
Xiao, Z., Kreis, K., and Vahdat, A., “Tackling the generative learning trilemma with denoising diffusion gans,” arXiv preprint arXiv:2112.07804 (2021)
2021 arXiv
-
[14]
Denoising diffusion probabilistic models,
Ho, J., Jain, A., and Abbeel, P., “Denoising diffusion probabilistic models,”Advances in neural information processing systems 33, 6840–6851 (2020)
2020
-
[15]
Books and publications:
Kak, A. C., Slaney, M., and Wang, G., “Books and publications:" principles of computerized tomographic imaging",” Medical Physics 29(1), 107 (2002)
2002
-
[16]
Deeplesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning,
Yan, K., Wang, X., Lu, L., and Summers, R. M., “Deeplesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning,”Journal of medical imaging 5(3), 036501– 036501 (2018)
2018
-
[17]
Torchradon: Fast differentiable routines for computed tomography,
Ronchetti, M., “Torchradon: Fast differentiable routines for computed tomography,” arXiv preprint arXiv:2009.14788 (2020)
2020 arXiv
-
[18]
Detection of masses and architectural distortions in digital breast tomosynthesis: a publicly available dataset of 5,060 patients and a deep learning model,
Buda, M., Saha, A., Walsh, R., Ghate, S., Li, N., Święcicki, A., Lo, J. Y., and Mazurowski, M. A., “Detection of masses and architectural distortions in digital breast tomosynthesis: a publicly available dataset of 5,060 patients and a deep learning model,”arXiv preprint arXiv...
2020 arXiv
-
[19]
Adam: A method for stochastic optimization,
Kingma, D. P. and Ba, J., “Adam: A method for stochastic optimization,”arXiv preprint arXiv:1412.6980 (2014)
2014 arXiv
-
[20]
Alias-freegenerative adversarial networks,
Karras, T., Aittala, M., Laine, S., Härkönen, E., Hellsten, J., Lehtinen, J., andAila, T., “Alias-freegenerative adversarial networks,” Advances in neural information processing systems 34, 852–863 (2021)
2021
Reviewed August 9, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.