Pith. sign in

REVIEW 4 major objections 6 minor 5 cited by

Medical World Model: Generative Simulation of Tumor Evolution for Treatment Planning

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read MeWM generates the post-treatment CT that each candidate TACE protocol would produce, scores the imagined outcome with survival analysis, and picks the lowest-risk plan; physicians using it raise their F1 agreement with the gold protocol…

desk verdict A genuinely novel action-conditioned tumor generator inside a world-model planning loop, but the 13% clinical gain is not yet credible: survival scores on synthetic CTs are never validated against real outcomes, and the test sets are too small for the reported deltas. read the letter →

arxiv 2506.02327 v1 pith:57RPXIOZ submitted 2025-06-02 cs.CV

Yijun Yang , Zhao-Yang Wang , Qiuping Liu , Shuwen Sun , Kang Wang , Rama Chellappa , Zongwei Zhou , Alan Yuille
show 3 more authors
Lei Zhu Yu-Dong Zhang Jieneng Chen
This is my paper · ORCID
classification cs.CV
keywords medicalworldmodeltumordynamicssimulationTACEtreatmentplanningdiffusiongenerativesurvivalanalysishepatocellularcarcinomavision-languagepolicyclinicaldecisionsupport
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 tries to establish that clinical treatment planning can be cast as a world-model problem: instead of asking a model to name the right treatment from a single image, MeWM first imagines what the tumor would look like after each candidate treatment, then scores the imagined outcome with a survival-risk model and searches for the protocol with the lowest risk. The application is Transarterial Chemoembolization (TACE) for hepatocellular carcinoma, where the action is a combination of chemotherapy drugs and embolization materials. If the claim holds, physicians gain a visually trackable second reader that improves their agreement with the gold TACE protocol by 13 percentage points in F1-score, and the same blueprint could apply to any treatment whose effect is visible on follow-up imaging. The paper also argues that its forward model produces post-treatment tumors realistic enough to fool radiologists in roughly a quarter of Turing-test presentations, and that its inverse dynamics model stratifies survival risk better than a radiomics-based Cox model.

What carries the argument

The central object is the closed loop of three modules: a vision-language policy model that proposes candidate TACE action combos (drugs plus embolisms) and prunes them with clinical rules; a dynamics model, an action-conditioned 3D latent diffusion model whose training uses Text-driven Morpho-Gaussian Attenuation to mimic embolization-induced shrinkage and a combo contrastive loss to make simulated outcomes depend on the specific drug and embolism keywords; and an inverse-dynamics model in which a finetuned segmentation assistant localizes the simulated tumor and a survival-analysis network — Cox proportional-hazards and overall-survival regression with cross-attention fusion of pre- and post-treatment features — emits a risk score. Beam search over the action space, replacing the worst-scoring beam and bounding drug and embolism counts, selects the optimal protocol.

What would settle it

Take patients who have both a real post-treatment CT and a MeWM-simulated post-treatment CT for the same case, run both through the survival heuristic, and compute their rank correlation; near-zero or negative correlation would show the search optimizes image artifacts rather than efficacy. Alternatively, a prospective randomized comparison of MeWM-selected versus physician-selected TACE protocols on overall survival would settle whether the 13% F1 gain translates into better outcomes.

Watch

Extended reading notes

Core claim

MeWM is presented as the first medical world model that visually predicts future disease states from clinical decisions: given a pre-treatment CT and a candidate TACE protocol, a 3D latent diffusion model synthesizes the post-treatment CT, and an inverse-dynamics heuristic assigns a survival risk score to that simulated outcome. The reported results: radiologists misclassified 21–29% of MeWM-synthesized tumors as real (specificity 79.17/70.83/75.00 across three readers), the model achieves the best FID (0.71) and LPIPS (0.612) among compared tumor synthesis methods, and the survival heuristic reaches c-index 0.752 versus 0.472 for a radiomics-based Cox baseline. In protocol exploration on the in-house dataset MeWM achieves F1 52.38 versus 41.97 for the best multimodal GPT baseline and 48.81 for physicians reading pre-treatment CT alone; when physicians consult MeWM their F1 rises to 61.51, and on the public HCC-TACE-Seg dataset the corresponding gain is from 71.43 to 80.00. The central claim is that this closed loop of generative forward simulation plus survival-scored inverse search, not the policy model alone, is what makes treatment selection improve.

Load-bearing premise

The whole optimization assumes that the survival-risk model, trained only on real pre/post CT pairs and survival times, judges synthetic post-treatment CT scans the same way it judges real ones, so that the beam search is optimizing a trustworthy measure of treatment benefit.

Editorial extensions

If this is right

  • Interventional physicians using MeWM as a second reader raise their F1-score for matching the gold TACE protocol from 48.81 to 61.51 on the in-house dataset and from 71.43 to 80.00 on the public cohort.
  • MeWM outperforms all evaluated multimodal and medical-specialized LLM baselines on F1, Jaccard, precision, and recall in both datasets.
  • Radiologists misclassify 21–29% of MeWM-synthesized post-treatment tumors as real, and MeWM attains the best FID and LPIPS among the compared synthesis methods.
  • Removing the segmentation assistant or the combo-contrastive loss reduces protocol F1 by 3.9 and 1.4 points respectively on the in-house dataset, showing both components are load-bearing.
  • The inverse-dynamics survival model separates high- from low-risk patients with c-index 0.752 and log-rank p = 6.74e-5, versus c-index 0.472 for the radiomics-based Cox baseline.

Reading between the lines

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

  • Editorial: the reported F1 gains measure agreement with gold clinical actions, not patient survival; a prospective trial with overall or progression-free survival endpoints would test whether the simulated-image-driven plans actually improve outcomes.
  • Editorial: the architecture is not TACE-specific beyond its action base, so the same forward-simulation-plus-inverse-scoring loop should transfer to ablation, radiotherapy, or immunotherapy wherever follow-up imaging shows the lesion.
  • Editorial: the weakest link is the unvalidated transfer of the survival heuristic from real to synthetic post-treatment CTs; comparing risk scores on a patient's real and simulated post-treatment images would reveal whether the search optimizes a trustworthy objective.
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

4 major / 6 minor

Summary. This paper introduces the Medical World Model (MeWM), a generative decision-support system for transarterial chemoembolization (TACE) in hepatocellular carcinoma. The system couples three components: a vision-language policy model (GPT-4o with DeepSeek-R1 refinement) that proposes candidate drug and embolism combinations from the pre-treatment CT; an action-conditioned 3D latent diffusion dynamics model that synthesizes post-treatment CT volumes from the pre-treatment scan and a candidate action, using text-driven Morpho-Gaussian attenuation and combo contrastive learning; and an inverse dynamics model that segments the simulated tumor with a fine-tuned nnUNet, scores the simulated post-treatment scan with a 3D survival-analysis heuristic (CoxPH plus OS regression), and performs beam search over candidate protocols. Experiments are reported on an in-house 338-pair HCC-TACE dataset and the public HCC-TACE-Seg dataset. Evaluation includes radiologist Turing tests, FID and LPIPS, c-index and log-rank survival separation, and F1, Jaccard, precision, and recall of the predicted protocol against the clinician's gold action. The headline result is a 13-point F1 improvement for interventional physicians using MeWM (48.81 to 61.51) and F1 gains over medical GPT baselines.

Significance. MeWM is a genuinely integrative contribution: it connects generative tumor simulation to survival-based treatment planning, it validates on an external public dataset, and it includes human radiologist evaluation. The idea of scoring simulated post-treatment states with a survival heuristic is promising and clinically motivated, and the paper is clearly written with a complete algorithmic description. However, the central clinical claim currently rests on two unverified steps: the transfer of a survival heuristic trained on real scans to synthetic scans, and a physician-in-the-loop comparison whose protocol is not described. The reported gains also lack confidence intervals and significance tests at small test sizes. If the authors supply the missing validation and statistical rigor, the work could be a meaningful advance for medical world models and individualized treatment planning.

major comments (4)
  1. [3.3, Eq. (4)] The core of the protocol search is the assumption that H_surv, trained on real (pre-treatment CT, post-treatment CT, OS) tuples, yields valid risk scores when applied to f_DM-generated post-treatment CTs. This domain transfer is never validated. Because f_DM outputs are synthetic images with attenuation patterns determined by the training-time attenuation levels (Section 3.2), H_surv may rank synthetic images by artifacts rather than by survival. The paper should report, on held-out patients, the concordance between risk scores computed on real post-treatment CTs and risk scores computed on f_DM(x0,a) for the same patient, or calibrate and validate the heuristic on synthetic images against OS. Without this, Eq. (4) optimizes an unverified objective.
  2. [4.3, Tables 2 and 3] The test sets are small (approximately 34 in-house cases from a 9:1 split of 338, and 16 public cases from a 20% holdout of 78), yet no confidence intervals, bootstrap estimates, or significance tests are reported for the F1 differences. The ablation differences, such as 52.38 to 49.13 without the Assistant Model and 52.38 to 50.97 without CCL, are within the range one would expect from sampling noise at this sample size, so the component contributions and the 13-point headline gain are not established as robust. Please report per-patient confidence intervals and tests, and the variance across the T replicas used in Algorithm 1.
  3. [4.3, Table 2] The comparison 'Physician w/ Pre-CT' versus 'Physician w/ MeWM' is central to the abstract, but the protocol for this human-in-the-loop evaluation is not described. It is not stated how many physicians participated, what MeWM output they saw, whether they could override it, how a final protocol was obtained, or how disagreements were resolved. Without this protocol the 13% claim is not reproducible, and it is unclear whether the improvement is an artifact of the aggregation rule rather than of genuine decision support.
  4. [4.1, Table 1] Perceptual metrics (FID and LPIPS) are computed on whole CT volumes. Since the dynamics model is conditioned on the pre-treatment CT and only the tumor mask is attenuated in training (Section 3.2), most of the volume may be copied from the input, which would make the global FID artificially small regardless of whether the tumor region is realistic. Report region-restricted FID and LPIPS, for example restricted to the liver or tumor bounding box, and ideally a radiologist evaluation that isolates the tumor region, before claiming state-of-the-art generation realism.
minor comments (6)
  1. [3.3 and Table 3] There are typos: 'Assitant Model' appears in the Section 3.3 heading and in Table 3, and 'inverse dyanmics' appears in the Introduction; both should be corrected.
  2. [3.3 and Algorithm 1] The notation H_surv(x0, m0, xhat, m0, g) in Section 3.3 repeats m0; it should read H_surv(x0, m0, xhat, \hat{m}, g). In Algorithm 1, lines 8 and 16 use 'x' where 'plans[b]' appears to be intended; please clarify how the accumulated plan state is passed to f_DM.
  3. [Abstract and Table 2] The '+13%' in the abstract is actually a gain of 12.70 percentage points in F1 from 48.81 to 61.51, which is a relative increase of about 26%; please specify 'percentage points' to avoid ambiguity, and the same correction applies to the '+9%' on the public dataset.
  4. [3.2 and Appendix B] The attenuation levels l=1..4 and the contrastive temperature delta are introduced as important design choices, but no values or sensitivity analysis are reported; because the attenuation schedule encodes a strong biological assumption about treatment response, please report the chosen values and ablate them.
  5. [4.2] The survival-analysis evaluation (c-index, MSE, Kaplan-Meier curves) is reported without the sample size, censoring rate, and time horizon of the evaluation cohort; adding these details is necessary to judge comparability across models.
  6. [Algorithm 1] The hyperparameters D, E, T, B, H_d, and H_e are listed as inputs but no default values are given anywhere in the paper; please provide the values used in the experiments.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity: the planning objective is a genuine composite of separately trained generative and survival models, not a fit to the evaluation labels.

full rationale

MeWM's derivation chain is not circular under the definitions required here. The dynamics model f_DM is trained with a diffusion objective (Eq. 2) on real (pre-treatment CT, action, post-treatment CT) pairs, while the heuristic function H_surv is trained on real pre/post CT pairs together with Overall Survival time via CoxPH and OS regression (Sec. 3.3). The TACE protocol search objective, Eq. (4), composes these two independently trained functions: argmin over actions of H_surv applied to f_DM-generated post-treatment CTs. This composite is not fitted to the gold actions used for evaluation in Table 2, so the reported F1 improvement is an empirical result of a genuine forward-simulation plus reward-scoring loop rather than a quantity that equals a training loss by construction. The radiologist Turing test (Table 1), FID/LPIPS metrics, the external HCC-TACE-Seg evaluation, and the c-index comparison (0.752 vs 0.472) provide external anchoring that is not sourced from the paper's own fitted values. The strongest skeptical concern—that H_surv, trained on real post-treatment CTs, is applied to synthetic post-treatment CTs without explicit validation of ranking fidelity across that domain shift—is a validity/correctness threat, not a circularity threat: the paper makes no claim that synthetic risk scores are identical by construction to real risk scores. Several implementation details cite prior work by overlapping authors (DiffTumor [14], Pixel2Cancer [42], CLIP-Driven Universal Model [49,50]), but these are architectural borrowings such as diffusion backbones, ellipsoid-based tumor shape pretraining, and organ pseudo-label post-processing; they are not load-bearing justifications of the central claim, and no uniqueness theorem or ansatz is imported from the authors' prior work to forbid alternatives. Accordingly, no specific circular step can be exhibited with a quoted reduction, and the appropriate finding is no significant circularity; the score of 1 reflects only the presence of non-load-bearing self-citations, not any circular reasoning.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The system rests on several domain assumptions: the historical TACE protocol is a valid gold standard, the survival model trained on real images transfers to synthetic images, and the radiologist Turing test is a valid proxy for clinical realism. The attenuation level mapping (l=1..4) is an ad hoc modeling choice. No new physical entities are introduced; 'concept embeddings' are learned parameters, not invented entities.

free parameters (3)
  • Attenuation level l = 1, 2, 3, 4
    Hand-crafted mapping from textual occlusion words to image attenuation level; controls erosion/dilation and Gaussian blur in the dynamics model (Section 3.2).
  • Contrastive temperature delta = not reported
    Temperature in the Combo Contrastive Learning loss (Eq. 3); no value given.
  • Search horizons and beams (D, E, T, B, H_d, H_e) = not reported
    Hyperparameters in Algorithm 1 controlling the tree search; values not specified in the paper.
assumptions (4)
  • domain assumption The historical TACE protocol (gold action) is a valid proxy for the optimal treatment.
    The treatment-planning evaluation in Table 2 computes F1 between predicted and gold actions, assuming the historical choice is the correct one; the paper does not validate against survival outcomes.
  • domain assumption Survival risk scores from the heuristic function generalize from real to synthetic post-treatment CT images.
    H_surv is trained on real pre/post pairs and then applied to synthetic x̂ in Eq. (4); no transfer validation is provided.
  • domain assumption The radiologist Turing test specificity is a valid measure of clinical realism.
    Section 4.1 interprets the lowest specificity as 'state-of-the-art' realism, equating human misclassification with clinical fidelity.
  • ad hoc to paper Text-driven Morpho-Gaussian Attenuation levels (l=1..4) faithfully represent biological response to TACE.
    The mapping from words like 'occluded' to discrete attenuation levels with hand-chosen kernels is a simplified, non-validated modeling choice in Section 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Medical World Model: Generative Simulation of Tumor Evolution for Treatment Planning." pith.science (2026). https://pith.science/paper/57RPXIOZ

@misc{pith2026250602327,
  author       = {Pith},
  title        = {Pith review of: Medical World Model: Generative Simulation of Tumor Evolution for Treatment Planning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/57RPXIOZ}},
  note         = {Machine review of arXiv:2506.02327}
}
read the original abstract

Providing effective treatment and making informed clinical decisions are essential goals of modern medicine and clinical care. We are interested in simulating disease dynamics for clinical decision-making, leveraging recent advances in large generative models. To this end, we introduce the Medical World Model (MeWM), the first world model in medicine that visually predicts future disease states based on clinical decisions. MeWM comprises (i) vision-language models to serve as policy models, and (ii) tumor generative models as dynamics models. The policy model generates action plans, such as clinical treatments, while the dynamics model simulates tumor progression or regression under given treatment conditions. Building on this, we propose the inverse dynamics model that applies survival analysis to the simulated post-treatment tumor, enabling the evaluation of treatment efficacy and the selection of the optimal clinical action plan. As a result, the proposed MeWM simulates disease dynamics by synthesizing post-treatment tumors, with state-of-the-art specificity in Turing tests evaluated by radiologists. Simultaneously, its inverse dynamics model outperforms medical-specialized GPTs in optimizing individualized treatment protocols across all metrics. Notably, MeWM improves clinical decision-making for interventional physicians, boosting F1-score in selecting the optimal TACE protocol by 13%, paving the way for future integration of medical world models as the second readers.

Figures

Figures reproduced from arXiv: 2506.02327 by the authors.

Figure 1
Figure 1. Formulation of Medical World Model. It integrates imaging observations with perception modules to form an initial state, which is then processed by a progression generative model to predict future states of disease under different treatment con￾ditions. Recovery-conditioned policies guide treatment decisions, creating a feedback loop for optimizing clinical interventions. tors including medical images and patient hi… view at source ↗
Figure 2
Figure 2. Overview of TACE Protocol Exploration by Medical World Model. (1) GPTs (Policy Model): construct the TACE action combos by the observation of pre-treatment CT, integrating clinical guidelines and policies. (2) Tumor Generative Model (Dynamics Model): simulates post-treatment tumor based on different TACE intervention protocols, predicting treatment outcomes. (3) Survival Analysis Model (Heuristic Function): assesses… view at source ↗
Figure 3
Figure 3. Dynamics Model based on Tumor Generative Model. The training framework consists of three parts: (a) Radiotherapy Report Extraction and Generation: GPT-4o and Deepseek-R1 extract key treatment details from radiotherapy reports and generate corresponding TACE surgical actions. (b) Post-Treatment Tumor Generation: An Action-driven 3D Diffusion Model is conditioned by fused action em￾beddings and attenuated CT features … view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Examples of Visual Turing Test. We present one real tumor alongside examples of synthetic tumors that were correctly and incorrectly identified. A red dot indicates the radiologist clas￾sified the post-treatment tumor as synthetic, while a green dot sig￾nifies it was i…
Figure 6
Figure 6. Figure 6: Kaplan-Meier Survival Curves: Radiomics-based Cox Model [67] vs. our deep model. The left shows the sur￾vival curves predicted by the Cox model based on Radiomics fea￾tures. The right presents the survival curves from our model based on deep features, which achieves a …
Figure 7
Figure 7. Figure 7: Example of MeWM intervention in clinical applica￾tions. The radiologist initially proposes a TACE protocol with Raltitrexed, Lobaplatin, Idarubicin, and embolization using Li￾piodol and Gelatin Sponge. MeWM simulates a protocol with Raltitrexed, Lobaplatin, and Lipiodo…
Figure 8
Figure 8. Figure 8: Example of HCC-TACE-Seg dataset. The first row shows HCC Pre-CT images, and the second row shows HCC Post-CT images. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Example of HCC-TACE dataset. The first row shows HCC Pre-CT images, and the second row shows HCC Post-CT images. The [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Policy model prompt template for our dataset. [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: VLM prompt template for our dataset [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: VLM prompt template for HCC-TACE-Seg [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. CLARITY: Medical World Model for Guiding Treatment Decisions by Modeling Context-Aware Disease Trajectories in Latent Space

    cs.LG 2025-12 conditional novelty 6.0 of 10

    A latent-space world model that forecasts treatment-conditioned tumor trajectories and selects therapies by iteratively minimizing its own predicted risk score.

  2. AuricularWorld: Hierarchical Action-Guided World Modeling for Fine-Grained Auricular Structure Segmentation from CT Scans

    cs.CV 2026-07 conditional novelty 5.5 of 10

    A latent RSSM with hierarchical anatomical add/remove actions cuts HD95 by ~43% versus nnU-Net on fine-grained nested auricular CT segmentation.

  3. A Definition and Roadmap for World Models

    cs.AI 2026-07 conditional novelty 5.0 of 10

    A perspective article defining world models as finite-resource compression of physical state transitions and outlining a roadmap toward physical AGI via unified representations and interactive simulators.

  4. ShapeKit

    eess.IV 2025-06 reject novelty 5.0 of 10

    ShapeKit, a rule-based post-processing toolkit, reports Dice score improvements of up to 8.8 percentage points on two CT datasets without retraining the segmentation model.

  5. The Path to Self-Evolving Clinical Systems: Scaling Medical Agents from Assistance to Autonomy

    cs.AI 2026-07 conditional novelty 4.5 of 10

    Medical agents should be scaled mainly by richer clinical environments and self-evolution loops, not parameter growth alone, under a three-level autonomy taxonomy.

Reference graph

Works this paper leans on

79 extracted references · 52 canonical work pages · cited by 5 Pith papers

  1. [2]

    Claude 3.5 sonnet model card addendum

    AI Anthropic. Claude 3.5 sonnet model card addendum. Claude-3.5 Model Card, 2024. 8

    work page 2024

  2. [3]

    Qwen2.5-vl technical report.arXiv preprint arXiv:2502.13923, 2025

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhao- hai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Cheng, Hang Zhang, Zhibo Yang, Haiyang Xu, and Jun- yang Lin. Qwen2.5-vl technical repor...

    arXiv 2025

  3. [4]

    Navigation world models

    Amir Bar, Gaoyue Zhou, Danny Tran, Trevor Darrell, and Yann LeCun. Navigation world models. InCVPR, 2025. 2

    work page 2025

  4. [5]

    The liver tumor segmentation benchmark (lits).Medical image analysis, 84:102680, 2023

    Patrick Bilic, Patrick Christ, Hongwei Bran Li, Eugene V orontsov, Avi Ben-Cohen, Georgios Kaissis, Adi Szeskin, Colin Jacobs, Gabriel Efrain Humpire Mamani, Gabriel Chartrand, et al. The liver tumor segmentation benchmark (lits).Medical image analysis, 84:102680, 2023. 2

    work page 2023

  5. [6]

    Synthseg: Segmenta- tion of brain mri scans of any contrast and resolution without retraining.Medical image analysis, 86:102789, 2023

    Benjamin Billot, Douglas N Greve, Oula Puonti, Axel Thielscher, Koen Van Leemput, Bruce Fischl, Adrian V Dalca, Juan Eugenio Iglesias, et al. Synthseg: Segmenta- tion of brain mri scans of any contrast and resolution without retraining.Medical image analysis, 86:102789, 2023. 2

    work page 2023

  6. [7]

    Stable video diffusion: Scaling latent video diffusion models to large datasets.arXiv preprint arXiv:2311.15127, 2023

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets.arXiv preprint arXiv:2311.15127, 2023. 2

    arXiv 2023

  7. [8]

    A vision–language foundation model for the generation of realistic chest x-ray images.Nature Biomedical Engineer- ing, pages 1–13, 2024

    Christian Bluethgen, Pierre Chambon, Jean-Benoit Del- brouck, Rogier van der Sluijs, Małgorzata Połacin, Juan Manuel Zambrano Chaves, Tanishq Mathew Abraham, Shivanshu Purohit, Curtis P Langlotz, and Akshay S Chaud- hari. A vision–language foundation model for the generation of realistic chest x-ray images.Nature Biomedical Engineer- ing, pages 1–13, 2024. 2

    work page 2024

  8. [9]

    On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021

    Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Alt- man, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al. On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021. 1

    arXiv 2021

Show all 79 references
  1. [10]

    Ge- nie: Generative interactive environments

    Jake Bruce, Michael D Dennis, Ashley Edwards, Jack Parker-Holder, Yuge Shi, Edward Hughes, Matthew Lai, Aditi Mavalankar, Richie Steigerwald, Chris Apps, et al. Ge- nie: Generative interactive environments. InForty-first Inter- national Conference on Machine Learning, 2024. 2

    2024

  2. [11]

    Multilingual feasibility of gpt-4o for automated voice-to-text ct and mri report transcription.Eu- ropean Journal of Radiology, 182:111827, 2025

    Felix Busch, Philipp Prucker, Alexander Komenda, Sebas- tian Ziegelmayer, Marcus R Makowski, Keno K Bressem, and Lisa C Adams. Multilingual feasibility of gpt-4o for automated voice-to-text ct and mri report transcription.Eu- ropean Journal of Radiology, 182:111827, 2025. 3

    2025

  3. [12]

    Monai: An open-source framework for deep learning in healthcare.arXiv preprint arXiv:2211.02701, 2022

    M Jorge Cardoso, Wenqi Li, Richard Brown, Nic Ma, Eric Kerfoot, Yiheng Wang, Benjamin Murrey, Andriy Myro- nenko, Can Zhao, Dong Yang, et al. Monai: An open-source framework for deep learning in healthcare.arXiv preprint arXiv:2211.02701, 2022. 5

    2022 arXiv

  4. [13]

    Huatuogpt-vision, to- wards injecting medical visual knowledge into multimodal llms at scale.arXiv preprint arXiv:2406.19280, 2024

    Junying Chen, Chi Gui, Ruyi Ouyang, Anningzhe Gao, Shu- nian Chen, Guiming Hardy Chen, Xidong Wang, Ruifei Zhang, Zhenyang Cai, Ke Ji, et al. Huatuogpt-vision, to- wards injecting medical visual knowledge into multimodal llms at scale.arXiv preprint arXiv:2406.19280, 2024. 3

    2024 arXiv

  5. [14]

    Towards generaliz- able tumor synthesis

    Qi Chen, Xiaoxi Chen, Haorui Song, Zhiwei Xiong, Alan Yuille, Chen Wei, and Zongwei Zhou. Towards generaliz- able tumor synthesis. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 11147–11158, 2024. 2, 4, 5, 6, 7

    2024

  6. [15]

    Surgsora: Decoupled rgbd-flow diffusion model for controllable surgical video generation

    Tong Chen, Shuya Yang, Junyi Wang, Long Bai, Hongliang Ren, and Luping Zhou. Surgsora: Decoupled rgbd-flow diffusion model for controllable surgical video generation. arXiv preprint arXiv:2412.14018, 2024. 2

    2024 arXiv

  7. [16]

    Meditron-70b: Scaling medical pretraining for large language models.arXiv preprint arXiv:2311.16079, 2023

    Zeming Chen, Alejandro Hern ´andez Cano, Angelika Ro- manou, Antoine Bonnet, Kyle Matoba, Francesco Salvi, Matteo Pagliardini, Simin Fan, Andreas K ¨opf, Amirkeivan Mohtashami, et al. Meditron-70b: Scaling medical pretraining for large language models.arXiv preprint arXiv:2311...

    2023 arXiv

  8. [17]

    Empirical comparisons between kaplan- meier and nelson-aalen survival function estimators.Journal of Statistical Computation and Simulation, 72(4):299–308,

    Enrico Colosimo, Fla´ vio Ferreira, Maristela Oliveira, and Cleide Sousa. Empirical comparisons between kaplan- meier and nelson-aalen survival function estimators.Journal of Statistical Computation and Simulation, 72(4):299–308,

  9. [18]

    End-to-end adversarial retinal image syn- thesis.IEEE transactions on medical imaging, 37(3):781– 791, 2017

    Pedro Costa, Adrian Galdran, Maria Ines Meyer, Meindert Niemeijer, Michael Abr `amoff, Ana Maria Mendonc ¸a, and Aur´elio Campilho. End-to-end adversarial retinal image syn- thesis.IEEE transactions on medical imaging, 37(3):781– 791, 2017. 2

    2017

  10. [19]

    Medgpt, 2025

    Michael D Moor. Medgpt, 2025. Accessed: March 7, 2025. 8

    2025

  11. [20]

    Video language plan- ning.arXiv preprint arXiv:2310.10625, 2023

    Yilun Du, Mengjiao Yang, Pete Florence, Fei Xia, Ayzaan Wahid, Brian Ichter, Pierre Sermanet, Tianhe Yu, Pieter Abbeel, Joshua B Tenenbaum, et al. Video language plan- ning.arXiv preprint arXiv:2310.10625, 2023. 2

    2023 arXiv

  12. [21]

    Video language plan- ning

    Yilun Du, Mengjiao Yang, Pete Florence, Fei Xia, Ayzaan Wahid, Brian Ichter, Pierre Sermanet, Tianhe Yu, Pieter Abbeel, Joshua B Tenenbaum, et al. Video language plan- ning. InICML, 2024. 2

    2024

  13. [22]

    Taming transformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image synthesis. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12873–12883, 2021. 4

    2021

  14. [23]

    Cox proportional-hazards regression for survival data.An R and S-PLUS companion to applied regression, 2002, 2002

    John Fox and Sanford Weisberg. Cox proportional-hazards regression for survival data.An R and S-PLUS companion to applied regression, 2002, 2002. 7

    2002

  15. [24]

    Emu video: Factoriz- ing text-to-video generation by explicit image conditioning

    Rohit Girdhar, Mannat Singh, Andrew Brown, Quentin Du- val, Samaneh Azadi, Sai Saketh Rambhatla, Akbar Shah, Xi Yin, Devi Parikh, and Ishan Misra. Emu video: Factoriz- ing text-to-video generation by explicit image conditioning. arXiv preprint arXiv:2311.10709, 2023. 2

    2023 arXiv

  16. [25]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025. 3, 5

    2025 arXiv

  17. [26]

    World models.arXiv preprint arXiv:1803.10122, 2018

    David Ha and J ¨urgen Schmidhuber. World models.arXiv preprint arXiv:1803.10122, 2018. 2

    2018 arXiv

  18. [27]

    Mastering diverse domains through world models

    Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, and Timothy Lillicrap. Mastering diverse domains through world models. arXiv preprint arXiv:2301.04104, 2023. 2

    2023 arXiv

  19. [28]

    Deep learning for cancer prognosis predic- tion using portrait photos by stylegan embedding

    Amr Hagag, Ahmed Gomaa, Dominik Kornek, Andreas Maier, Rainer Fietkau, Christoph Bert, Yixing Huang, and Florian Putz. Deep learning for cancer prognosis predic- tion using portrait photos by stylegan embedding. InIn- ternational Conference on Medical Image Computing and Compu...

  20. [29]

    Evaluation and mitigation of the limitations of large lan- guage models in clinical decision-making.Nature medicine, 30(9):2613–2622, 2024

    Paul Hager, Friederike Jungmann, Robbie Holland, Kunal Bhagat, Inga Hubrecht, Manuel Knauer, Jakob Vielhauer, Marcus Makowski, Rickmer Braren, Georgios Kaissis, et al. Evaluation and mitigation of the limitations of large lan- guage models in clinical decision-making.Nature me...

    2024

  21. [30]

    Ct2rep: Automated radiology report generation for 3d medi- cal imaging

    Ibrahim Ethem Hamamci, Sezgin Er, and Bjoern Menze. Ct2rep: Automated radiology report generation for 3d medi- cal imaging. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 476–

  22. [31]

    Parametric re- sponse mapping of contrast-enhanced biphasic ct for evalu- ating tumour viability of hepatocellular carcinoma after tace

    Jan B Hinrichs, Hoen-Oh Shin, Daniel Kaercher, Davut Has- demir, Tim Murray, Till Kaireit, Carolin Lutat, Arndt V ogel, Bernhard C Meyer, Frank K Wacker, et al. Parametric re- sponse mapping of contrast-enhanced biphasic ct for evalu- ating tumour viability of hepatocellular c...

    2016

  23. [32]

    Denoising diffu- sion probabilistic models.Advances in Neural Information Processing Systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models.Advances in Neural Information Processing Systems, 33:6840–6851, 2020. 2, 5

    2020

  24. [33]

    Gaia-1: A generative world model for au- tonomous driving.arXiv preprint arXiv:2309.17080, 2023

    Anthony Hu, Lloyd Russell, Hudson Yeo, Zak Murez, George Fedoseev, Alex Kendall, Jamie Shotton, and Gian- luca Corrado. Gaia-1: A generative world model for au- tonomous driving.arXiv preprint arXiv:2309.17080, 2023. 2

    2023 arXiv

  25. [34]

    Label-free liver tumor segmentation

    Qixin Hu, Yixiong Chen, Junfei Xiao, Shuwen Sun, Jieneng Chen, Alan L Yuille, and Zongwei Zhou. Label-free liver tumor segmentation. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 7422–7432, 2023. 2, 6, 7

    2023

  26. [35]

    Common feature learning for brain tumor mri synthesis by context-aware generative adversarial network

    Pu Huang, Dengwang Li, Zhicheng Jiao, Dongming Wei, Bing Cao, Zhanhao Mo, Qian Wang, Han Zhang, and Ding- gang Shen. Common feature learning for brain tumor mri synthesis by context-aware generative adversarial network. Medical Image Analysis, 79:102472, 2022. 2

    2022

  27. [36]

    Ct-based deep learning predicts prognosis in esophageal squamous cell cancer patients receiving immunotherapy combined with chemotherapy.Academic Radiology, 2025

    Xiaoyu Huang, Yong Huang, Ping Li, and Kai Xu. Ct-based deep learning predicts prognosis in esophageal squamous cell cancer patients receiving immunotherapy combined with chemotherapy.Academic Radiology, 2025. 3

    2025

  28. [37]

    Gpt-4o system card

    Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perel- man, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Weli- hinda, Alan Hayes, Alec Radford, et al. Gpt-4o system card. arXiv preprint arXiv:2410.21276, 2024. 1, 8

    2024 arXiv

  29. [38]

    nnu-net: a self-configuring method for deep learning-based biomedical image segmen- tation.Nature methods, 18(2):203–211, 2021

    Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Pe- tersen, and Klaus H Maier-Hein. nnu-net: a self-configuring method for deep learning-based biomedical image segmen- tation.Nature methods, 18(2):203–211, 2021. 5, 2

    2021

  30. [39]

    Deep- surv: personalized treatment recommender system using a cox proportional hazards deep neural network.BMC medi- cal research methodology, 18:1–12, 2018

    Jared L Katzman, Uri Shaham, Alexander Cloninger, Jonathan Bates, Tingting Jiang, and Yuval Kluger. Deep- surv: personalized treatment recommender system using a cox proportional hazards deep neural network.BMC medi- cal research methodology, 18:1–12, 2018. 5

    2018

  31. [40]

    Mdagents: An adaptive collaboration of llms for medical decision- making.Advances in Neural Information Processing Sys- tems, 37:79410–79452, 2024

    Yubin Kim, Chanwoo Park, Hyewon Jeong, Yik Siu Chan, Xuhai Xu, Daniel McDuff, Hyeonhoon Lee, Marzyeh Ghas- semi, Cynthia Breazeal, Hae Park, et al. Mdagents: An adaptive collaboration of llms for medical decision- making.Advances in Neural Information Processing Sys- tems, 37:...

    2024

  32. [41]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. InProceedings of the IEEE/CVF international confer- ence on computer vision, pages 4015–4026, 2023. 5

    2023

  33. [42]

    From pixel to cancer: Cellular automata in computed tomography

    Yuxiang Lai, Xiaoxi Chen, Angtian Wang, Alan Yuille, and Zongwei Zhou. From pixel to cancer: Cellular automata in computed tomography. InInternational Conference on Med- ical Image Computing and Computer-Assisted Intervention, pages 36–46. Springer, 2024. 2, 6, 7

    2024

  34. [43]

    A path towards autonomous machine intelli- gence version 0.9

    Yann LeCun. A path towards autonomous machine intelli- gence version 0.9. 2, 2022-06-27.Open Review, 62(1):1–62,

    2022

  35. [44]

    Kyung Hwa Lee, Jungwook Lee, Gwang Hyeon Choi, Ji- hye Yun, Jiseon Kang, Jonggi Choi, Kang Mo Kim, and Namkug Kim. Deep learning-based prediction of post- treatment survival in hepatocellular carcinoma patients using pre-treatment ct images and clinical data.Journal of Imaging...

    2024

  36. [45]

    Endora: Video generation models as endoscopy simu- lators

    Chenxin Li, Hengyu Liu, Yifan Liu, Brandon Y Feng, Wuyang Li, Xinyu Liu, Zhen Chen, Jing Shao, and Yixuan Yuan. Endora: Video generation models as endoscopy simu- lators. InInternational Conference on Medical Image Com- puting and Computer-Assisted Intervention, pages 230–240....

    2024

  37. [46]

    Large language models-powered clinical decision support: enhanc- ing or replacing human expertise?, 2025

    Jia Li, Zichun Zhou, Han Lyu, and Zhenchang Wang. Large language models-powered clinical decision support: enhanc- ing or replacing human expertise?, 2025. 3

    2025

  38. [47]

    Text-driven tumor synthesis.arXiv preprint arXiv:2412.18589, 2024

    Xinran Li, Yi Shuai, Chen Liu, Qi Chen, Qilong Wu, Pengfei Guo, Dong Yang, Can Zhao, Pedro RAS Bassi, Daguang Xu, et al. Text-driven tumor synthesis.arXiv preprint arXiv:2412.18589, 2024. 2, 6, 7, 5

    2024 arXiv

  39. [48]

    Deep learning supported discovery of biomarkers for clinical prognosis of liver cancer.Nature Machine Intelli- gence, 5(4):408–420, 2023

    Junhao Liang, Weisheng Zhang, Jianghui Yang, Meilong Wu, Qionghai Dai, Hongfang Yin, Ying Xiao, and Lingjie Kong. Deep learning supported discovery of biomarkers for clinical prognosis of liver cancer.Nature Machine Intelli- gence, 5(4):408–420, 2023. 3

    2023

  40. [49]

    Clip-driven universal model for organ segmentation and tumor detection

    Jie Liu, Yixiao Zhang, Jie-Neng Chen, Junfei Xiao, Yongyi Lu, Bennett A Landman, Yixuan Yuan, Alan Yuille, Yucheng Tang, and Zongwei Zhou. Clip-driven universal model for organ segmentation and tumor detection. InProceedings of the IEEE/CVF International Conference on Computer...

    2023

  41. [50]

    Universal and extensible language-vision models for organ segmentation and tumor detection from abdominal computed tomography.Medical image analysis, 97:103226, 2024

    Jie Liu, Yixiao Zhang, Kang Wang, Mehmet Can Yavuz, Xiaoxi Chen, Yixuan Yuan, Haoliang Li, Yang Yang, Alan Yuille, Yucheng Tang, et al. Universal and extensible language-vision models for organ segmentation and tumor detection from abdominal computed tomography.Medical image a...

    2024

  42. [51]

    Generative world explorer.arXiv preprint arXiv:2411.11844, 2024

    Taiming Lu, Tianmin Shu, Alan Yuille, Daniel Khashabi, and Jieneng Chen. Generative world explorer.arXiv preprint arXiv:2411.11844, 2024. 2

    2024 arXiv

  43. [52]

    Grounding video models to ac- tions through goal conditioned exploration.arXiv preprint arXiv:2411.07223, 2024

    Yunhao Luo and Yilun Du. Grounding video models to ac- tions through goal conditioned exploration.arXiv preprint arXiv:2411.07223, 2024. 2

    2024 arXiv

  44. [53]

    Pseudo-label guided image synthesis for semi-supervised covid-19 pneumonia in- fection segmentation.IEEE Transactions on Medical Imag- ing, 42(3):797–809, 2022

    Fei Lyu, Mang Ye, Jonathan Frederik Carlsen, Kenny Er- leben, Sune Darkner, and Pong C Yuen. Pseudo-label guided image synthesis for semi-supervised covid-19 pneumonia in- fection segmentation.IEEE Transactions on Medical Imag- ing, 42(3):797–809, 2022. 2

    2022

  45. [54]

    Foundation models for generalist medi- cal artificial intelligence.Nature, 616(7956):259–265, 2023

    Michael Moor, Oishi Banerjee, Zahra Shakeri Hossein Abad, Harlan M Krumholz, Jure Leskovec, Eric J Topol, and Pranav Rajpurkar. Foundation models for generalist medi- cal artificial intelligence.Nature, 616(7956):259–265, 2023. 8

    2023

  46. [55]

    A ma- chine learning model to predict hepatocellular carcinoma re- sponse to transcatheter arterial chemoembolization.Radiol- ogy: Artificial Intelligence, 1(5):e180021, 2019

    Ali Morshid, Khaled M Elsayes, Ahmed M Khalaf, Mo- hab M Elmohr, Justin Yu, Ahmed O Kaseb, Manal Hassan, Armeen Mahvash, Zhihui Wang, John D Hazle, et al. A ma- chine learning model to predict hepatocellular carcinoma re- sponse to transcatheter arterial chemoembolization.Radi...

    2019

  47. [56]

    Improved denoising diffusion probabilistic models

    Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. InInternational Conference on Machine Learning, pages 8162–8171. PMLR,

  48. [57]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. InInternational conference on machine learning, p...

    2021

  49. [58]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2, 4

    2022

  50. [59]

    Masked world models for visual control

    Younggyo Seo, Danijar Hafner, Hao Liu, Fangchen Liu, Stephen James, Kimin Lee, and Pieter Abbeel. Masked world models for visual control. InConference on Robot Learning, pages 1332–1344. PMLR, 2023. 2

    2023

  51. [60]

    Transmil: Transformer based correlated multiple instance learning for whole slide image classification.Advances in neural information processing systems, 34:2136–2147, 2021

    Zhuchen Shao, Hao Bian, Yang Chen, Yifeng Wang, Jian Zhang, Xiangyang Ji, et al. Transmil: Transformer based correlated multiple instance learning for whole slide image classification.Advances in neural information processing systems, 34:2136–2147, 2021. 6

    2021

  52. [61]

    Large language models encode clinical knowledge.Nature, 620(7972):172–180, 2023

    Karan Singhal, Shekoofeh Azizi, Tao Tu, S Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tan- wani, Heather Cole-Lewis, Stephen Pfohl, et al. Large language models encode clinical knowledge.Nature, 620(7972):172–180, 2023. 1

    2023

  53. [62]

    Toward expert-level med- ical question answering with large language models.Nature Medicine, pages 1–8, 2025

    Karan Singhal, Tao Tu, Juraj Gottweis, Rory Sayres, Ellery Wulczyn, Mohamed Amin, Le Hou, Kevin Clark, Stephen R Pfohl, Heather Cole-Lewis, et al. Toward expert-level med- ical question answering with large language models.Nature Medicine, pages 1–8, 2025. 1

    2025

  54. [63]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. InInternational confer- ence on machine learning, pages 2256–2265. PMLR, 2015. 2

    2015

  55. [64]

    Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023

    Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean- Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023. 1

    2023 arXiv

  56. [65]

    A closer look at spatiotemporal convolutions for action recognition

    Du Tran, Heng Wang, Lorenzo Torresani, Jamie Ray, Yann LeCun, and Manohar Paluri. A closer look at spatiotemporal convolutions for action recognition. InProceedings of the IEEE conference on Computer Vision and Pattern Recogni- tion, pages 6450–6459, 2018. 5

    2018

  57. [66]

    Mocogan: Decomposing motion and content for video generation

    Sergey Tulyakov, Ming-Yu Liu, Xiaodong Yang, and Jan Kautz. Mocogan: Decomposing motion and content for video generation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1526–1535,

  58. [67]

    Joost van Griethuysen et. al. Radiomics-based cox model,

  59. [68]

    Towards real-world-driven world models for autonomous driving.arXiv preprint arXiv:2309.09777, 2023

    X Wang, Z Zhu, G Huang, X Chen, and J Drivedreamer Lu. Towards real-world-driven world models for autonomous driving.arXiv preprint arXiv:2309.09777, 2023. 2

    2023 arXiv

  60. [69]

    Drivedreamer: Towards real-world- drive world models for autonomous driving

    Xiaofeng Wang, Zheng Zhu, Guan Huang, Xinze Chen, Jia- gang Zhu, and Jiwen Lu. Drivedreamer: Towards real-world- drive world models for autonomous driving. InEuropean Conference on Computer Vision, pages 55–72, 2024. 2

    2024

  61. [70]

    Daydreamer: World models for physical robot learning

    Philipp Wu, Alejandro Escontrela, Danijar Hafner, Pieter Abbeel, and Ken Goldberg. Daydreamer: World models for physical robot learning. InConference on robot learning, pages 2226–2240. PMLR, 2023. 2

    2023

  62. [71]

    Cross-conditioned diffu- sion model for medical image to image translation

    Zhaohu Xing, Sicheng Yang, Sixiang Chen, Tian Ye, Yi- jun Yang, Jing Qin, and Lei Zhu. Cross-conditioned diffu- sion model for medical image to image translation. InIn- ternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 201–211. Springer,

  63. [72]

    Jiawen Yao, Yu Shi, Kai Cao, Le Lu, Jianping Lu, Qike Song, Gang Jin, Jing Xiao, Yang Hou, and Ling Zhang. Deepprognosis: Preoperative prediction of pancreatic can- cer survival and surgical margin via comprehensive un- derstanding of dynamic contrast-enhanced ct imaging and t...

    2021

  64. [73]

    Deepprognosis: Preoperative prediction of pancreatic cancer survival and surgical margin via contrast-enhanced ct imag- ing

    Jiawen Yao, Yu Shi, Le Lu, Jing Xiao, and Ling Zhang. Deepprognosis: Preoperative prediction of pancreatic cancer survival and surgical margin via contrast-enhanced ct imag- ing. InInternational Conference on Medical Image Com- puting and Computer-Assisted Intervention, pages ...

    2020

  65. [74]

    Label-free segmentation of covid-19 lesions in lung ct.IEEE transactions on medical imaging, 40(10):2808–2819, 2021

    Qingsong Yao, Li Xiao, Peihang Liu, and S Kevin Zhou. Label-free segmentation of covid-19 lesions in lung ct.IEEE transactions on medical imaging, 40(10):2808–2819, 2021. 2

    2021

  66. [75]

    Addressing asynchronicity in clinical multimodal fusion via individualized chest x-ray generation.Advances in Neural Information Processing Systems, 37:29001–29028,

    Wenfang Yao, Chen Liu, Kejing Yin, William Cheung, and Jing Qin. Addressing asynchronicity in clinical multimodal fusion via individualized chest x-ray generation.Advances in Neural Information Processing Systems, 37:29001–29028,

  67. [76]

    Huatuogpt, towards taming language model to be a doctor.arXiv preprint arXiv:2305.15075, 2023

    Hongbo Zhang, Junying Chen, Feng Jiang, Fei Yu, Zhi- hong Chen, Jianquan Li, Guiming Chen, Xiangbo Wu, Zhiyi Zhang, Qingying Xiao, et al. Huatuogpt, towards taming language model to be a doctor.arXiv preprint arXiv:2305.15075, 2023. 8

    2023 arXiv

  68. [77]

    Large language models for medicine: a survey.International Journal of Machine Learn- ing and Cybernetics, pages 1–26, 2024

    Yanxin Zheng, Wensheng Gan, Zefeng Chen, Zhenlian Qi, Qian Liang, and Philip S Yu. Large language models for medicine: a survey.International Journal of Machine Learn- ing and Cybernetics, pages 1–26, 2024. 3

    2024

  69. [78]

    High-resolution diabetic retinopathy im- age synthesis manipulated by grading and lesions

    Yi Zhou, Xiaodong He, Shanshan Cui, Fan Zhu, Li Liu, and Ling Shao. High-resolution diabetic retinopathy im- age synthesis manipulated by grading and lesions. InIn- ternational conference on medical image computing and computer-assisted intervention, pages 505–513. Springer,

  70. [2019]

    image":{image_property}

    2 Appendix Table of Contents A . Data Preprocessing 2 A.1 . HCC-TACE-Seg dataset preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 A.2 . HCC-TACE dataset preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....

  71. [2025]

    Accessed: March 7, 2025. 7

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.