arxiv: 2604.07329 · v1 · submitted 2026-04-08 · 💻 cs.CV

Recognition: 2 theorem links

· Lean Theorem

Distilling Photon-Counting CT into Routine Chest CT through Clinically Validated Degradation Modeling

Junqi Liu , Xinze Zhou , Wenxuan Li , Scott Ye , Arkadiusz Sitek , Xiaofeng Yang , Yucheng Tang , Daguang Xu

show 5 more authors

Kai Ding Kang Wang Yang Yang Alan L. Yuille Zongwei Zhou

Authors on Pith no claims yet

Pith reviewed 2026-05-10 17:58 UTC · model grok-4.3

classification 💻 cs.CV

keywords photon-counting CTimage enhancementdegradation modelingchest CTlatent diffusion modelclinical validationlesion detectionenergy-integrating CT

0 comments

The pith

SUMI models realistic degradations from photon-counting CT to train enhancement of routine chest CT images without paired scans at scale.

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

The paper proposes SUMI to bridge the gap between limited high-quality photon-counting CT scans and widely available but lower-quality conventional energy-integrating CT by explicitly simulating acquisition degradations. It trains an enhancement model on these simulated pairs, validated for clinical realism by radiologists, to improve routine scans toward photon-counting quality. This yields better image similarity metrics than prior translation methods, higher radiologist-rated utility, and stronger performance on downstream lesion detection tasks. A reader would care because most clinical CT scanners are energy-integrating, so the approach could extend advanced imaging benefits without new hardware.

Core claim

SUMI transforms photon-counting CT into clinically plausible lower-quality energy-integrating counterparts via validated degradation simulation, then learns to invert the process with a latent diffusion model pre-trained on a large mixed CT dataset, producing enhanced routine CT images that improve SSIM by 15 percent, PSNR by 20 percent, clinical utility scores, and lesion detection sensitivity by up to 15 percent on external data.

What carries the argument

The clinically validated simulated degradation model that converts high-quality PCCT into realistic EICT-like images to create paired training data for the enhancement diffusion model.

If this is right

A latent diffusion model trained on 1,046 PCCT scans with an autoencoder pre-trained on 405,379 EICT scans from 145 hospitals extracts reusable general CT latent features.
A public dataset of over 17,316 enhanced EICT volumes is released with radiologist-validated annotations for airways, vessels, lungs, and lobes.
The method outperforms prior image translation approaches by 15 percent SSIM and 20 percent PSNR while raising lesion detection sensitivity up to 15 percent and F1 up to 10 percent.
Radiologist reader studies confirm improved clinical utility of the enhanced images.

Where Pith is reading between the lines

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

Hospitals could apply the released autoencoder features to other CT generative tasks without retraining from scratch.
Similar degradation modeling might extend to distilling benefits from other advanced modalities such as spectral CT or MRI into standard scanners.
If the degradation simulation generalizes across vendors, the approach could reduce the need for large paired acquisition studies in future imaging upgrades.

Load-bearing premise

The simulated degradations capture the full range of real-world acquisition artifacts and scanner variations in routine clinical EICT data.

What would settle it

A reader study or quantitative comparison on actual paired PCCT and EICT scans from the same patients where the SUMI-enhanced images show no measurable improvement in radiologist preference or lesion detection metrics over the original EICT.

Figures

Figures reproduced from arXiv: 2604.07329 by Alan L. Yuille, Arkadiusz Sitek, Daguang Xu, Junqi Liu, Kai Ding, Kang Wang, Scott Ye, Wenxuan Li, Xiaofeng Yang, Xinze Zhou, Yang Yang, Yucheng Tang, Zongwei Zhou.

**Figure 1.** Figure 1: Overview of SUMI. (1.) A continual learning autoencoder is pre-trained on over 400,000 CT volumes from 145 hospitals across 19 countries under pixel-wise loss Lpp and adversarial loss Ldis, making it the largest and most geographically diverse open-source medical CT autoencoder to date, overpassing prior methods (LDM [27]: 0, MedDiff [35]: 2, MONAI [5]: 7, Med3D [7]: 8, MAISI [39]: 10). (2.) A clinical ver… view at source ↗

**Figure 2.** Figure 2: SUMI preserves anatomical structure and tissue density. Pearson correlation (r) between ground truth and enhanced CT measurements across all organs shows that SUMI maintains anatomical accuracy and HU consistency. Organ masks are obtained with VISTA3D [16]. Baseline and implementation. We select representative methods from four major categories: traditional filtering (NLM [4]), Vision Transformers (Swin2S… view at source ↗

**Figure 3.** Figure 3: SUMI demonstrates superior generalization and enhancement quality. For each example, the left image shows the CT slice and the right shows the airway tree segmentation. (a) Cross-dataset evaluation: compared with autoencoders trained with limited medical data scale (LDM, Med3D, MAISI), SUMI better preserves fine airway topology under domain shift. (b) PCCT enhancement with ground truth: compared with stro… view at source ↗

**Figure 4.** Figure 4: SUMI improves chest CT quality across datasets. Quality score distributions (by a pretrained scorer, 0–1, higher is better) on Luna16 [29], LNDb19 [25], DSB17 [18], and a private PCCT dataset. On public datasets, enhanced scans consistently shift toward higher scores, indicating robust quality gains. On PCCT, a controlled degradation-to-enhancement benchmark shows that SUMI narrows the gap to real high-… view at source ↗

read the original abstract

Photon-counting CT (PCCT) provides superior image quality with higher spatial resolution and lower noise compared to conventional energy-integrating CT (EICT), but its limited clinical availability restricts large-scale research and clinical deployment. To bridge this gap, we propose SUMI, a simulated degradation-to-enhancement method that learns to reverse realistic acquisition artifacts in low-quality EICT by leveraging high-quality PCCT as reference. Our central insight is to explicitly model realistic acquisition degradations, transforming PCCT into clinically plausible lower-quality counterparts and learning to invert this process. The simulated degradations were validated for clinical realism by board-certified radiologists, enabling faithful supervision without requiring paired acquisitions at scale. As outcomes of this technical contribution, we: (1) train a latent diffusion model on 1,046 PCCTs, using an autoencoder first pre-trained on both these PCCTs and 405,379 EICTs from 145 hospitals to extract general CT latent features that we release for reuse in other generative medical imaging tasks; (2) construct a large-scale dataset of over 17,316 publicly available EICTs enhanced to PCCT-like quality, with radiologist-validated voxel-wise annotations of airway trees, arteries, veins, lungs, and lobes; and (3) demonstrate substantial improvements: across external data, SUMI outperforms state-of-the-art image translation methods by 15% in SSIM and 20% in PSNR, improves radiologist-rated clinical utility in reader studies, and enhances downstream top-ranking lesion detection performance, increasing sensitivity by up to 15% and F1 score by up to 10%. Our results suggest that emerging imaging advances can be systematically distilled into routine EICT using limited high-quality scans as reference.

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

The paper gives a workable way to turn limited PCCT scans into training data for enhancing real-world EICT by simulating and validating degradations, then releasing the resulting annotated dataset.

read the letter

The main point is that they start with high-quality PCCT images, apply a degradation model tuned to look like routine energy-integrating CT, get radiologists to confirm the simulated images are plausible, and train a latent diffusion model to undo the process. They also pre-train an autoencoder on a large mix of PCCT and EICT data and put out over 17k enhanced EICT volumes with voxel labels for airways, vessels, and lungs. That dataset release is the most immediately usable part of the work. The reported gains on external scans—15% SSIM and 20% PSNR over other translation baselines, plus higher radiologist utility scores and better lesion detection sensitivity—come from training on the simulated pairs and testing on real EICT. The downstream task lift is a reasonable check that the enhancement is not just cosmetic. The approach is a direct extension of image-to-image translation ideas but adds the clinical validation step and the scale of the released data. The weakest link is whether the simulated degradations actually match the joint distribution of artifacts across 145 hospitals. Radiologist review on a limited set is better than nothing, but it does not quantify coverage of scanner-specific noise textures, motion, or contrast timing differences. If those are under-modeled, the external gains could shrink on truly unseen data. The abstract does not give error bars or full details on how the reader study was blinded or how exclusions were handled. This is aimed at groups working on CT image quality or medical dataset curation. The concrete outputs—the pre-trained encoder, the annotated volumes, and the downstream numbers—make it worth sending to referees even if the simulation realism needs tighter testing in revision.

Referee Report

2 major / 2 minor

Summary. The paper proposes SUMI, a simulated degradation-to-enhancement framework that distills photon-counting CT (PCCT) quality into routine energy-integrating CT (EICT) scans. It explicitly models realistic acquisition degradations on 1,046 PCCT volumes (validated by board-certified radiologists), pre-trains an autoencoder on 405,379 EICT scans from 145 hospitals, trains a latent diffusion model to invert the degradations, and releases a dataset of 17,316 enhanced EICT volumes with voxel-wise annotations of airways, vessels, lungs, and lobes. Reported outcomes include 15% SSIM and 20% PSNR gains over SOTA image translation methods on external data, improved radiologist-rated clinical utility, and up to 15% higher sensitivity / 10% higher F1 in downstream lesion detection.

Significance. If the central transfer claim holds, the work provides a practical route to upgrade widely available EICT without new hardware, using limited PCCT references. The release of the pre-trained autoencoder and the large annotated enhanced-EICT dataset constitutes a concrete community resource for generative medical imaging. The combination of radiologist-validated simulation, quantitative metrics, reader studies, and downstream task evaluation strengthens the clinical relevance.

major comments (2)

[§3] §3 (Degradation Modeling): The claim that radiologist validation on a limited set establishes clinical realism for the simulated degradations is load-bearing for the external-data performance numbers. The manuscript does not report quantitative coverage metrics (e.g., distribution distances on noise texture, beam-hardening, or motion artifacts) across the 145-hospital EICT corpus or scanner-specific variations; without this, the 15% SSIM / 20% PSNR gains and lesion-detection uplift on external data risk being overstated.
[§5.2] §5.2 (Reader Study): The description of the reader study does not specify blinding procedures, handling of post-hoc exclusions, or inter-reader agreement statistics. These details are required to support the claim of improved radiologist-rated clinical utility and to rule out bias in the validation of the simulated degradations.

minor comments (2)

[Methods] The abstract states the autoencoder is pre-trained on both PCCT and EICT data, but the exact loss weighting and training schedule for the joint pre-training are not stated in the methods; adding these would improve reproducibility.
[Table 2] Table 2 (quantitative comparison) reports SSIM and PSNR but omits standard deviations or confidence intervals across the external test cases; including them would strengthen the statistical interpretation of the 15%/20% gains.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which highlight important aspects of clinical validation and methodological transparency. We address each major point below and will incorporate revisions to strengthen the manuscript.

read point-by-point responses

Referee: [§3] §3 (Degradation Modeling): The claim that radiologist validation on a limited set establishes clinical realism for the simulated degradations is load-bearing for the external-data performance numbers. The manuscript does not report quantitative coverage metrics (e.g., distribution distances on noise texture, beam-hardening, or motion artifacts) across the 145-hospital EICT corpus or scanner-specific variations; without this, the 15% SSIM / 20% PSNR gains and lesion-detection uplift on external data risk being overstated.

Authors: We agree that quantitative coverage metrics would provide additional support for the generalizability of the degradation model across the heterogeneous 145-hospital EICT corpus. In the revised manuscript, we will add analyses computing distribution distances (including FID on noise texture patches and normalized power spectrum differences for beam-hardening and motion effects) between the simulated degradations and real EICT scans, with stratification by major scanner vendors where metadata permits. This will directly quantify coverage and scanner-specific fidelity. At the same time, we note that board-certified radiologist validation remains the most direct assessment of clinical plausibility, as perceptual and diagnostic realism in CT cannot be fully captured by distribution metrics alone; the added quantitative results will complement rather than replace this validation. revision: yes
Referee: [§5.2] §5.2 (Reader Study): The description of the reader study does not specify blinding procedures, handling of post-hoc exclusions, or inter-reader agreement statistics. These details are required to support the claim of improved radiologist-rated clinical utility and to rule out bias in the validation of the simulated degradations.

Authors: We concur that these details are necessary for full transparency and to substantiate the reader-study claims. The revised §5.2 will explicitly describe: the blinding protocol (readers were blinded to image origin, enhancement method, and whether scans were original or simulated); that no post-hoc exclusions were applied after initial scoring; and inter-reader agreement statistics (Cohen’s kappa for categorical utility ratings and intraclass correlation coefficient for continuous scores). These additions will allow readers to assess potential bias and strengthen the evidence for improved clinical utility. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper derives SUMI by explicitly modeling acquisition degradations from independent high-quality PCCT references (1,046 scans), validating clinical realism via board-certified radiologist review, and training a latent diffusion model (with autoencoder pre-trained on combined PCCT + 405k EICT) to invert the process. All reported outcomes—15% SSIM / 20% PSNR gains, reader-study utility, and downstream lesion-detection uplift—are measured on external EICT data using standard independent metrics and tasks. No step reduces a prediction to a fitted input by construction, renames a known result, or relies on load-bearing self-citation chains; the central claim rests on the external reference data and explicit simulation rather than tautological definitions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that a latent diffusion model trained on simulated degradations can generalize to real clinical EICT variability; no free parameters are explicitly named in the abstract, but the diffusion process and autoencoder pre-training involve standard learned weights.

axioms (1)

domain assumption Latent diffusion models can learn to invert realistic CT acquisition degradations when trained on paired simulated data
Invoked in the description of SUMI training on 1,046 PCCTs with simulated lower-quality counterparts.

pith-pipeline@v0.9.0 · 5667 in / 1322 out tokens · 39598 ms · 2026-05-10T17:58:20.704133+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We apply three degradation strategies. Sparse View reduces projection numbers in Radon space... Low Dose decreases photon counts and models signal-dependent Poisson noise... Conventional Degradation applies spatial downsampling with Gaussian and Poisson noise injection
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

train a latent diffusion model on 1,046 PCCTs, using an autoencoder first pre-trained on both these PCCTs and 405,379 EICTs

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

39 extracted references · 11 canonical work pages · 1 internal anchor

[1]

Medical physics38(2), 915–931 (2011)

Armato III, S.G., McLennan, G., Bidaut, L., McNitt-Gray, M.F., Meyer, C.R., Reeves, A.P., Zhao, B., Aberle, D.R., Henschke, C.I., Hoffman, E.A., et al.: The lungimagedatabaseconsortium(lidc)andimagedatabaseresourceinitiative(idri): a completed reference database of lung nodules on ct scans. Medical physics38(2), 915–931 (2011)

2011
[2]

The Insight Journal (2006)

Bartholmai, B., Karwoski, R., Zavaletta, V., Robb, R., Holmes, D.: The lung tissue research consortium: An extensive open database containing histological, clinical, and radiological data to study chronic lung disease. The Insight Journal (2006)

2006
[3]

European Journal of Radiology190, 112189 (2025)

van der Bie, J., van der Laan, T., van Straten, M., Booij, R., Bos, D., Dijkshoorn, M.L., Hirsch, A., Oei, E.H., Budde, R.P.: Photon-counting ct: an updated review of clinical results. European Journal of Radiology190, 112189 (2025)

2025
[4]

In: 2005IEEEcomputersocietyconferenceoncomputervisionandpatternrecognition (CVPR’05)

Buades, A., Coll, B., Morel, J.M.: A non-local algorithm for image denoising. In: 2005IEEEcomputersocietyconferenceoncomputervisionandpatternrecognition (CVPR’05). vol. 2, pp. 60–65. Ieee (2005)

2005
[5]

MONAI: An open-source framework for deep learning in healthcare

Cardoso, M.J., Li, W., Brown, R., Ma, N., Kerfoot, E., Wang, Y., Murrey, B., Myronenko, A., Zhao, C., Yang, D., et al.: Monai: An open-source framework for deep learning in healthcare. arXiv preprint arXiv:2211.02701 (2022)

work page internal anchor Pith review arXiv 2022
[6]

Clinical Imaging104, 110008 (2023)

Chamberlin, J.H., Smith, C.D., Maisuria, D., Parrish, J., van Swol, E., Mah, E., Emrich, T., Schoepf, U.J., Varga-Szemes, A., O’Doherty, J., et al.: Ultra-high- resolution photon-counting detector computed tomography of the lungs: Phantom and clinical assessment of radiation dose and image quality. Clinical Imaging104, 110008 (2023)

2023
[7]

Med3d: Transfer learning for 3d medical image analysis

Chen, S., Ma, K., Zheng, Y.: Med3d: Transfer learning for 3d medical image anal- ysis. arXiv preprint arXiv:1904.00625 (2019)

work page arXiv 1904
[8]

Radiology: Artificial Intelligence3(2), e200254 (2021)

Colak, E., Kitamura, F.C., Hobbs, S.B., Wu, C.C., Lungren, M.P., Prevedello, L.M., Kalpathy-Cramer, J., Ball, R.L., Shih, G., Stein, A., et al.: The rsna pul- monary embolism ct dataset. Radiology: Artificial Intelligence3(2), e200254 (2021)

2021
[9]

In: European conference on computer vision

Conde, M.V., Choi, U.J., Burchi, M., Timofte, R.: Swin2sr: Swinv2 transformer for compressed image super-resolution and restoration. In: European conference on computer vision. pp. 669–687. Springer (2022)

2022
[10]

Medical physics51(12), 8776–8788 (2024)

Eulig, E., Ommer, B., Kachelrieß, M.: Benchmarking deep learning-based low-dose ct image denoising algorithms. Medical physics51(12), 8776–8788 (2024)

2024
[11]

Medical image analysis105, 103710 (2025)

Gao, Q., Chen, Z., Zeng, D., Zhang, J., Ma, J., Shan, H.: Noise-inspired diffusion model for generalizable low-dose ct reconstruction. Medical image analysis105, 103710 (2025)

2025
[12]

IEEE Transactions on Medical Imaging43(2), 745–759 (2023)

Gao, Q., Li, Z., Zhang, J., Zhang, Y., Shan, H.: Corediff: Contextual error- modulated generalized diffusion model for low-dose ct denoising and generalization. IEEE Transactions on Medical Imaging43(2), 745–759 (2023)

2023
[13]

Text2ct: Towards 3d ct volume generation from free-text descriptions using diffusion model.arXiv preprint arXiv:2505.04522,

Guo, P., Zhao, C., Yang, D., He, Y., Nath, V., Xu, Z., Bassi, P.R., Zhou, Z., Simon, B.D., Harmon, S.A., Syed, A.B., Roth, H., Xu, D.: Text2ct: Towards 3d ct volume generation from free-text descriptions using diffusion model. arXiv preprint arXiv:2505.04522 (2025)

work page arXiv 2025
[14]

Developing generalist foundation models from a multimodal dataset for 3d computed tomography.arXiv preprint arXiv:2403.17834, 2024

Hamamci, I.E., Er, S., Wang, C., Almas, F., Simsek, A.G., Esirgun, S.N., Do- gan, I., Durugol, O.F., Hou, B., Shit, S., et al.: Developing generalist foundation models from a multimodal dataset for 3d computed tomography. arXiv preprint arXiv:2403.17834 (2024)

work page arXiv 2024
[15]

He et al

He, Y., Guo, P., Tang, Y., Myronenko, A., Nath, V., Xu, Z., Yang, D., Zhao, C., Si- mon, B., Belue, M., et al.: Vista3d: Versatile imaging segmentation and annotation model for 3d computed tomography. arXiv preprint arXiv:2406.05285 (2024) SUMI11

work page arXiv 2024
[16]

In: Proceedings of the Computer Vision and Pattern Recognition Conference

He, Y., Guo, P., Tang, Y., Myronenko, A., Nath, V., Xu, Z., Yang, D., Zhao, C., Si- mon, B., Belue, M., et al.: Vista3d: A unified segmentation foundation model for 3d medical imaging. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 20863–20873 (2025)

2025
[17]

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

Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with condi- tional adversarial networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1125–1134 (2017)

2017
[18]

arXiv preprint arXiv:1705.09435 (2017)

Kuan, K., Ravaut, M., Manek, G., Chen, H., Lin, J., Nazir, B., Chen, C., Howe, T.C., Zeng, Z., Chandrasekhar, V.: Deep learning for lung cancer detection: tack- ling the kaggle data science bowl 2017 challenge. arXiv preprint arXiv:1705.09435 (2017)

work page arXiv 2017
[19]

In: CVPR

Ledig, C., Theis, L., Huszár, F., Caballero, J., Cunningham, A., Acosta, A., Aitken, A.P., Tejani, A., Totz, J., Wang, Z., et al.: Photo-realistic single image super- resolution using a generative adversarial network. In: CVPR. vol. 2, p. 4 (2017)

2017
[20]

IEEE transactions on neural networks and learning systems30(11), 3484–3495 (2019)

Liao, F., Liang, M., Li, Z., Hu, X., Song, S.: Evaluate the malignancy of pulmonary nodules using the 3-d deep leaky noisy-or network. IEEE transactions on neural networks and learning systems30(11), 3484–3495 (2019)

2019
[21]

Lin, T., Li, X., Zhuang, C., Chen, Q., Cai, Y., Ding, K., Yuille, A.L., Zhou, Z.: Are pixel-wise metrics reliable for sparse-view computed tomography reconstruction? arXiv preprint arXiv:2506.02093 (2025),https://github.com/MrGiovanni/CARE

work page arXiv 2025
[22]

arXiv preprint arXiv:2512.07251 (2025),https://github

Liu, J., Wu, Z., Bassi, P.R., Zhou, X., Li, W., Hamamci, I.E., Er, S., Lin, T., Luo, Y., Menze, B., et al.: See more, change less: Anatomy-aware diffusion for contrast enhancement. arXiv preprint arXiv:2512.07251 (2025),https://github. com/MrGiovanni/SMILE

work page arXiv 2025
[23]

Medical Physics52(1), 329–345 (2025)

Liu, X., Xie, Y., Liu, C., Cheng, J., Diao, S., Tan, S., Liang, X.: Diffusion prob- abilistic priors for zero-shot low-dose ct image denoising. Medical Physics52(1), 329–345 (2025)

2025
[24]

arXiv preprint arXiv:2504.06897 (2025),https://github.com/jwmao1/MedSegFactory

Mao, J., Wang, Y., Tang, Y., Xu, D., Wang, K., Yang, Y., Zhou, Z., Zhou, Y.: Med- segfactory: Text-guided generation of medical image-mask pairs. arXiv preprint arXiv:2504.06897 (2025),https://github.com/jwmao1/MedSegFactory

work page arXiv 2025
[25]

arXiv preprint arXiv:1911.08434 (2019)

Pedrosa, J., Aresta, G., Ferreira, C., Rodrigues, M., Leitão, P., Carvalho, A.S., Rebelo, J., Negrão, E., Ramos, I., Cunha, A., et al.: Lndb: a lung nodule database on computed tomography. arXiv preprint arXiv:1911.08434 (2019)

work page arXiv 1911
[26]

Nature communications12(1), 5678 (2021)

Perkonigg, M., Hofmanninger, J., Herold, C.J., Brink, J.A., Pianykh, O., Prosch, H., Langs, G.: Dynamic memory to alleviate catastrophic forgetting in continual learning with medical imaging. Nature communications12(1), 5678 (2021)

2021
[27]

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
[28]

IEEE transactions on pattern analysis and ma- chine intelligence45(4), 4713–4726 (2022)

Saharia, C., Ho, J., Chan, W., Salimans, T., Fleet, D.J., Norouzi, M.: Image super- resolution via iterative refinement. IEEE transactions on pattern analysis and ma- chine intelligence45(4), 4713–4726 (2022)

2022
[29]

Medical image analysis 42, 1–13 (2017)

Setio, A.A.A., Traverso, A., De Bel, T., Berens, M.S., Van Den Bogaard, C., Cerello, P., Chen, H., Dou, Q., Fantacci, M.E., Geurts, B., et al.: Validation, com- parison, and combination of algorithms for automatic detection of pulmonary nod- ules in computed tomography images: the luna16 challenge. Medical image analysis 42, 1–13 (2017)

2017
[30]

Physics in Medicine & Biology70(10), 10TR01 (2025) 12 J

Shah, K.D., Zhou, J., Roper, J., Dhabaan, A., Al-Hallaq, H., Pourmorteza, A., Yang, X.: Photon-counting ct in cancer radiotherapy: technological advances and clinical benefits. Physics in Medicine & Biology70(10), 10TR01 (2025) 12 J. Liu et al

2025
[31]

Radiology 317(2), e242961 (2025)

Tavakoli, N., Shakeri, Z., Gowda, V., Samsel, K., Bedayat, A., Ghasemiesfe, A., Bagci, U., Hsiao, A., Leiner, T., Carr, J., et al.: Generative ai and foundation mod- els in radiology: Applications, opportunities, and potential challenges. Radiology 317(2), e242961 (2025)

2025
[32]

Radiology258(1), 243–253 (2011)

Team, N.L.S.T.R.: The national lung screening trial: overview and study design. Radiology258(1), 243–253 (2011)

2011
[33]

American Journal of Neuroradiology45(10), 1450–1457 (2024)

Tóth, A., Chetta, J.A., Yazdani, M., Matheus, M.G., O‘Doherty, J., Tipnis, S.V., Spampinato, M.V.: Neurovascular imaging with ultra-high-resolution photon- counting ct: preliminary findings on image-quality evaluation. American Journal of Neuroradiology45(10), 1450–1457 (2024)

2024
[34]

European Radiology Experimental9(1), 38 (2025)

Varga-Szemes, A., Emrich, T.: Photon-counting detector ct: a disrupting innova- tion in medical imaging. European Radiology Experimental9(1), 38 (2025)

2025
[35]

IEEE Transactions on Medical Imaging (2025)

Wang, H., Liu, Z., Sun, K., Wang, X., Shen, D., Cui, Z.: 3d meddiffusion: A 3d medical latent diffusion model for controllable and high-quality medical image generation. IEEE Transactions on Medical Imaging (2025)

2025
[36]

Journal of Medical Imaging10(6), 061105– 061105 (2023)

Whitney, H.M., Baughan, N., Myers, K.J., Drukker, K., Gichoya, J., Bower, B., Chen, W., Gruszauskas, N., Kalpathy-Cramer, J., Koyejo, S., et al.: Longitudinal assessment of demographic representativeness in the medical imaging and data resource center open data commons. Journal of Medical Imaging10(6), 061105– 061105 (2023)

2023
[37]

Medical world model: Generative simulation of tumor evolution for treatment planning.arXiv preprint arXiv:2506.02327, 2025

Yang, Y., Wang, Z.Y., Liu, Q., Sun, S., Wang, K., Chellappa, R., Zhou, Z., Yuille, A., Zhu, L., Zhang, Y.D., Chen, J.: Medical world model: Generative simulation of tumor evolution for treatment planning. arXiv preprint arXiv:2506.02327 (2025), https://github.com/scott-yjyang/MeWM

work page arXiv 2025
[38]

IEEE transactions on medical imaging39(7), 2531–2540 (2020)

Zhang, L., Wang, X., Yang, D., Sanford, T., Harmon, S., Turkbey, B., Wood, B.J., Roth, H., Myronenko, A., Xu, D., et al.: Generalizing deep learning for medical image segmentation to unseen domains via deep stacked transformation. IEEE transactions on medical imaging39(7), 2531–2540 (2020)

2020
[39]

In: Proceedings of the AAAI Conference on Artificial Intelligence

Zhao, C., Guo, P., Yang, D., He, Y., Tang, Y., Simon, B., Belue, M., Harmon, S., Turkbey, B., Xu, D.: Maisi-v2: Accelerated 3d high-resolution medical image synthesis with rectified flow and region-specific contrastive loss. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 40, pp. 13088–13098 (2026)

2026