Patient-Specific Autoregressive Models for Organ Motion Prediction in Radiotherapy

Jike Zhong; Vanessa Su; Xiaofeng Yang; Yuxiang Lai

REVIEW 5 major objections 5 minor 44 references

Patient-Specific Autoregressive Models for Organ Motion Prediction in Radiotherapy

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

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

Pith's one-line read This paper claims that organ motion in radiotherapy can be predicted as an autoregressive sequence of CT phases, and that the resulting model beats published DVF- and diffusion-based baselines on lung and heart motion.

desk verdict A sensible repackaging of autoregressive token prediction for 4D CT organ motion, but the headline accuracy numbers rest on an evaluation protocol that is not described. read the letter →

arxiv 2505.11832 v1 pith:MOBF3XOF submitted 2025-05-17 eess.IV cs.CV

Yuxiang Lai , Jike Zhong , Vanessa Su , Xiaofeng Yang This is my paper · ORCID

classification eess.IVcs.CV

keywords autoregressivemodelorganmotionpredictionradiotherapy4DCTVQGANtokenizationpatient-specificlungandheart

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that organ motion seen in radiotherapy can be predicted by treating a patient's 4D CT scan as a sequence and asking an autoregressive model to generate the next phase, the way language models generate the next token. The payoff would be practical: if future lung and heart positions can be forecast from the patient's own pre-treatment scans, radiation margins could be tightened and healthy tissue spared. The authors build Auto-RMP, a VQGAN-plus-causal-transformer pipeline, and report that it outperforms a DVF-based method (DAM) and a diffusion method (DiffuseRT) on both a public lung-cancer dataset and a private clinical dataset, reaching 90.75% IoU and 95.15% DSC for lung and 88.43% IoU and 93.85% DSC for heart on the public data. The same autoregressive setup also rolls out five future phases while keeping DSC above 85% on the last predicted phase.

What carries the argument

The load-bearing mechanism is the autoregressive factorization in Equation (1): the joint probability of all future phases is written as a product of next-phase conditional probabilities, so training and inference reduce to next-token prediction. Around that factorization, Auto-RMP wraps two components: VQGAN, which encodes each CT phase into a 16x16 grid of discrete tokens with a downsampling factor of 16 and a codebook size of 8192, and a unidirectional causal transformer, based on the LLaMA architecture with a 4096-token context, which predicts the token sequence of the next phase given all previous phase tokens; the VQGAN decoder then renders the predicted phase. As a result, no deformable image registration or DVF computation is required, and the model can feed its own predictions back into the context to generate multiple future phases.

What would settle it

Run leave-one-patient-out cross-validation on the public Hugo dataset, holding out all scans of each patient together, and compare Auto-RMP's lung and heart IoU and DSC against the same baselines; if the gap narrows to the level of DAM or DiffuseRT, the claim that the model learns patient-specific motion would be unsupported. A complementary check would be to re-score predictions using manually curated contours instead of the same nnUNet/TotalSegmentator masks used for training input.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that organ motion prediction is an autoregressive process over CT phases. Given the sequence of prior phases $X_0, \ldots, X_T$, the model maximizes the product of conditional likelihoods $p_\theta(X'_t \mid X_0, \ldots, X_{t-1})$, so that at inference the patient's earlier scans serve as a prefix from which future phases are generated one at a time. Auto-RMP implements this by tokenizing each 256x256 CT phase, concatenated with its organ mask, using VQGAN into 256 discrete tokens drawn from an 8192-entry codebook, then predicting the next phase's tokens with a causal LLaMA-style transformer and decoding them back into a CT image. Because conditioning is on the patient's own phase history rather than on a single image or a precomputed deformation vector field, the paper argues that the model captures patient-specific periodic motion and can be rolled forward for several phases without the error collapse it reports for single-step baselines. The reported evidence is the next-phase and multi-phase accuracy on the public and private datasets.

Load-bearing premise

The paper never states how the public or private 4D CT scans were divided into training and testing sets, so if scans from the same patient appear on both sides, the reported patient-specific accuracy could reflect memorized anatomy rather than generalization to a new patient.

Editorial extensions

If this is right

Pre-treatment planning could use the predicted motion envelope to define smaller radiation margins, reducing dose to healthy tissue around the tumor.
Motion prediction no longer depends on deformable image registration: the model learns motion directly from CT phase sequences, removing a costly and error-prone preprocessing step.
Because the model conditions on the patient's own phase history, its accuracy is expected to track that patient's breathing pattern rather than a population average.
Multi-phase rollout means the same model can supply a full breathing cycle of future positions, not just the next phase, matching the temporal horizon needed for gated or adaptive delivery.

Reading between the lines

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

Editorial inference: the decisive test the paper does not report is a strict per-patient split on the longitudinal public dataset; if accuracy holds under leave-one-patient-out evaluation, the patient-specific claim is much stronger than if scans of the same patient appear in both training and testing.
Editorial inference: because the same nnUNet/TotalSegmentator pipeline produces both the auxiliary masks and the evaluation ground truth, an independent set of manually curated contours would clarify how much of the reported IoU and DSC reflects true anatomical prediction rather than shared segmentation bias.
Editorial inference: the same token-sequence framing could transfer to other periodic or quasi-periodic anatomies, such as liver motion in abdominal radiotherapy, and to other image modalities like cone-beam CT where longitudinal training data are beginning to exist.
Editorial inference: a head-to-head comparison with diffusion models that are themselves conditioned on multiple prior phases would isolate the contribution of the autoregressive conditioning from the choice of generative backbone.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

You should know two things about this paper. First, the core idea is genuinely useful: treating organ motion prediction as an autoregressive sequence problem over VQGAN tokens is a clean way to get multi-phase rollouts without precomputed DVFs. That is a real contribution, even though every component is borrowed. Second, the reported 8–10 point IoU gains over DAM and DiffuseRT are not yet credible, because the paper never tells you how the data were split into train and test.

The strength is the framing. The authors argue—correctly—that 4D CT is a temporal sequence, and that autoregressive modeling is a natural fit. The pipeline is standard (VQGAN tokenization, causal transformer, next-phase prediction), but the application is new and the logic is coherent. They also run a useful ablation on input modalities, showing that combining CT and masks helps. The limitations section honestly notes that the evaluation is single-session only, which is the right kind of humility.

The soft spot is the evaluation, and it is load-bearing. The public Hugo dataset is longitudinal: 20 patients, 80 scans, with multiple scans per patient. If the split is by scan rather than by patient, the model can memorize patient anatomy, which would directly inflate the numbers. The paper does not state the split. It also does not say how DAM and DiffuseRT were configured, whether they received the same inputs, or how metrics were pooled across phases. No error bars, no code. On top of that, the ground-truth masks come from the same TotalSegmentator model that provides auxiliary input, so part of the apparent accuracy may be mask propagation rather than true motion prediction. These are all fixable with additional reporting, but right now Table 1 is a claim without a verifiable protocol.

Who gets value from this? Researchers working on radiotherapy motion management and medical image sequence modeling. The paper is worth a serious referee, but the current version should be returned for major revision: clarify the split, describe baseline setups, add error bars and code, and discuss the mask-input circularity explicitly. If the numbers hold up after that, this becomes a solid application paper. As it stands, treat the quantitative claim as conditional.

Referee Report

5 major / 5 minor

Summary. The manuscript introduces Auto-RMP, an autoregressive model for predicting future 4D CT phases in lung and heart radiotherapy motion. After segmenting organs with TotalSegmentator/nnUNet, each CT phase is tokenized by a VQGAN into discrete tokens, and a LLaMA-style causal transformer predicts the next phase's tokens conditioned on previous phases. The method is evaluated on the public Hugo dataset (20 patients, 80 scans) and a private 50-patient dataset against DAM and DiffuseRT, reporting higher IoU/DSC and lower surface distances for next-phase and multi-phase prediction. An ablation compares CT-only, mask-only, and CT+mask inputs.

Significance. The core idea, casting organ motion as autoregressive sequence generation, is a clean reformulation that avoids DVF dependence and enables multi-phase roll-out. If the reported numbers survive a correctly specified evaluation, this would be a practically relevant advance for pre-treatment motion management. The paper motivates the work well, and the token-level formulation is simple enough to reproduce. The main contribution is currently empirical, and the empirical claims are not yet verifiable because evaluation protocol details are missing.

major comments (5)

[§3.1 Dataset / §3.2 Table 1] The manuscript never states how the 80 public and 50 private 4D CT scans were split into training, validation, and test sets. The public Hugo dataset is longitudinal (20 patients, 80 scans), so if scans from the same patient appear in both training and testing, the reported patient-specific accuracy may reflect memorization of that patient's anatomy rather than generalization. Please report patient-level splits, the number of patients and scans in each split, and per-patient test metrics (or confidence intervals) to rule out leakage.
[§2.1 Pre-processing / §3.1 Evaluation] The same TotalSegmentator nnUNet model is described as providing ground-truth labels for organ shape and position, and those masks are also concatenated with CT as input to the model. This creates a circular evaluation: high IoU/DSC may partly measure how well the model propagates the input mask rather than true motion prediction. Please clarify whether ground-truth masks for evaluation are independent of the input masks (e.g., manual contours or a different tool), or quantify the sensitivity of the reported metrics to segmentation noise.
[§3.2 Table 1] No configuration details are given for the DAM and DiffuseRT baselines: architecture, hyperparameters, training data, input format (CT only versus CT+mask), context length, or compute budget. Without these, 'significantly outperforms' cannot be verified; the comparison may be unfair. Please specify baseline setups, report the same splits and metrics for all methods, and make code and checkpoints available.
[§3.2 Long-term prediction] Long-term motion prediction is claimed as a key advantage, but the evidence is a qualitative figure and a single sentence stating that Auto-RMP maintains over 85% DSC in the final predicted phase. No per-phase quantitative table, no error bars, and no baseline comparison are provided. Please report per-phase IoU/DSC/SD/HD for Auto-RMP and all baselines in the five-phase roll-out setting.
[Table 1 / Table 2] All reported numbers are point estimates without standard deviations, confidence intervals, or statistical tests across patients and phases. The caption of Table 1 uses 'significantly outperforms,' which is not supported by any significance test. Please include per-patient variability and paired tests (e.g., Wilcoxon signed-rank) for the main comparisons.

minor comments (5)

[Table 1 caption] The caption refers to 'NSD' values, but the table contains no NSD column; please correct this.
[§2.2 Implementation Details] The sentence 'We adopt pre-trained parameters from Yutong et al. [1]' should cite Bai et al., matching reference [1].
[Figure 2 caption] The caption contains a grammatical error: 'We first arranges each phase' should be 'We first arrange each phase.'
[Table 2] The ablation uses a 'Mixed' set (20 public + 20 private scans) but does not describe which private scans were selected or how they were split; please specify this.
[Equation (1)] Equation (1) has typographical spacing issues in the product limit notation, and the notation for the final phase X_T should be made consistent throughout.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Auto-RMP is an empirical autoregressive pipeline whose prediction accuracy is not derived from its inputs by construction.

full rationale

The central claim is an empirical performance comparison, not a first-principles derivation. Equation (1) is the standard chain-rule factorization of a joint distribution over CT phases; it is a modeling choice rather than a result obtained from fitted parameters. The model is trained on CT sequences and evaluated on held-out phases, and the reported advantages over DAM and DiffuseRT are external benchmark comparisons, not consequences of the paper's own definitions. The many self-citations in the autoregressive-methods list are not load-bearing: the formulation is standard and is also supported by external references (Bai et al., El-Nouby et al., Yu et al.). No uniqueness theorem or fitted parameter is imported from the authors' prior work. The only self-referential concern is that the ground-truth masks and the evaluation masks both come from the same TotalSegmentator nnUNet (Sections 2.1 and 3.1); this is a measurement-validity caveat and could affect absolute accuracy and fair comparison if predicted masks are generated by the same segmenter, but it does not make the predicted future CT phases equivalent to the input by construction. The paper's own limitation statement (single-session evaluation) is a scope restriction, not a circular step. Under the stated rules demanding a specific reduction, no circular step can be exhibited.

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

The paper introduces no new particles, forces, dimensions, or conserved quantities. It is an empirical machine learning study whose load-bearing choices are the pretrained VQGAN tokenizer, hand-chosen hyperparameters such as codebook size and context length, and the use of TotalSegmentator labels as ground truth for evaluation.

free parameters (4)

VQGAN codebook size = 8192
Chosen without ablation; quantization granularity affects tokenization fidelity and thus the upper bound on prediction quality.
VQGAN downsampling factor = 16
Chosen to produce a 16x16 token grid per 256x256 image; no analysis of other downsampling rates.
Context length = 4096 tokens, up to 16 images
Chosen to fit 16 tokenized images; constrains how many CT phases can condition the prediction.
Number of conditioning phases for long-term prediction = 5
The long-term experiment feeds the first five phases; no sensitivity analysis for this choice.

assumptions (4)

standard math Autoregressive factorization P(X1,...,XT) = product over t of p_theta(X'_t | X0,...,X_{t-1}) is a valid characterization of 4D CT phase sequences (Eq. 1).
Standard chain rule of probability; used to define the training objective.
domain assumption TotalSegmentator nnUNet segmentations of lungs and heart are accurate enough to serve both as auxiliary input and as ground truth for evaluation.
Section 2.1 states these segmentations 'provide ground-truth labels'; if the segmentation has systematic errors, the reported IoU and DSC numbers are affected.
ad hoc to paper A VQGAN pretrained on natural images transfers to CT images without further domain adaptation.
Section 2.2 says the model adopts pretrained parameters from Yutong et al., but does not describe fine-tuning on CT; CT intensities and textures differ substantially from natural images.
domain assumption The causal transformer can capture respiratory motion from flattened token sequences of CT phases.
The paper assumes the VQGAN token grid flattening preserves the spatial and temporal structure needed for motion; no analysis of token ordering is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Patient-Specific Autoregressive Models for Organ Motion Prediction in Radiotherapy." pith.science (2026). https://pith.science/paper/MOBF3XOF

@misc{pith2026250511832,
  author       = {Pith},
  title        = {Pith review of: Patient-Specific Autoregressive Models for Organ Motion Prediction in Radiotherapy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MOBF3XOF}},
  note         = {Machine review of arXiv:2505.11832}
}

read the original abstract

Radiotherapy often involves a prolonged treatment period. During this time, patients may experience organ motion due to breathing and other physiological factors. Predicting and modeling this motion before treatment is crucial for ensuring precise radiation delivery. However, existing pre-treatment organ motion prediction methods primarily rely on deformation analysis using principal component analysis (PCA), which is highly dependent on registration quality and struggles to capture periodic temporal dynamics for motion modeling.In this paper, we observe that organ motion prediction closely resembles an autoregressive process, a technique widely used in natural language processing (NLP). Autoregressive models predict the next token based on previous inputs, naturally aligning with our objective of predicting future organ motion phases. Building on this insight, we reformulate organ motion prediction as an autoregressive process to better capture patient-specific motion patterns. Specifically, we acquire 4D CT scans for each patient before treatment, with each sequence comprising multiple 3D CT phases. These phases are fed into the autoregressive model to predict future phases based on prior phase motion patterns. We evaluate our method on a real-world test set of 4D CT scans from 50 patients who underwent radiotherapy at our institution and a public dataset containing 4D CT scans from 20 patients (some with multiple scans), totaling over 1,300 3D CT phases. The performance in predicting the motion of the lung and heart surpasses existing benchmarks, demonstrating its effectiveness in capturing motion dynamics from CT images. These results highlight the potential of our method to improve pre-treatment planning in radiotherapy, enabling more precise and adaptive radiation delivery.

Figures

Figures reproduced from arXiv: 2505.11832 by the authors.

**Figure 1.** Throughout the extended duration of radiotherapy, patients’ natural breathing motion results in the continuous rhythmic expansion and contraction of the lungs. Consequently, the tumor moves up and down in the coronal view (dashed curve). However, current treatment plans often designate the target with a large margin for dose delivery (red rectangle), which accounts for tumor motion but results in unintended radiat… view at source ↗

**Figure 2.** Auto-RMP. We first arranges each phase of a 4D CT scan into an input image sequence (X0 to XT ). Then, VQGAN serves as the vision encoder to convert CT images into discrete tokens. The output tokens are arranged into a 1D sequence, which is fed into the Autoregressive Transformer Prediction model. During the autoregressive process, the model predicts the next phase tokens based on the previous phase. Finally, the pr… view at source ↗

**Figure 3.** Long-term motion prediction. We evaluate the ability of models to perform long-term motion prediction. In this setting, the model is given the first five phases of the 4D CT scan and must predict the next five phases to assess its capability for learning long-term motion patterns. Auto-RMP demonstrates strong robustness in long-term motion prediction, generating smooth and consistent motion predictions [PITH_FULL_I… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 33 canonical work pages

[1]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Bai, Y., Geng, X., Mangalam, K., Bar, A., Yuille, A.L., Darrell, T., Malik, J., Efros, A.A.: Sequential modeling enables scalable learning for large vision models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22861–22872 (2024) 3, 6

work page 2024
[2]

Advances in neural information processing systems33, 1877–1901 (2020) 5

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language models are few-shot learners. Advances in neural information processing systems33, 1877–1901 (2020) 5

work page 2020
[3]

The Lancet398(10295), 171–184 (2021) 1

Chandra, R.A., Keane, F.K., Voncken, F.E., Thomas, C.R.: Contemporary radio- therapy: present and future. The Lancet398(10295), 171–184 (2021) 1

work page 2021
[4]

Federated Learning of Shareable Bases for Personalization-Friendly Image Classification

Chen*, H.Y., Zhong*, J., Zhang, M., Jia, X., Qi, H., Gong, B., Chao, W.L., Zhang, L.: Federated learning of shareable bases for personalization-friendly image classi- fication. arXiv preprint arXiv:2304.07882 (2023) 2

work page Pith review arXiv 2023
[5]

Generative Machine Learning Models in Medical Image Computing p

Chen, Q., Lai, Y., Chen, X., Hu, Q., Yuille, A., Zhou, Z.: Analyzing tumors by synthesis. Generative Machine Learning Models in Medical Image Computing p. 85 (2024) 5

work page 2024
[6]

Clinical oncology26(2), 67–80 (2014) 2

Cole, A., Hanna, G., Jain, S., O’Sullivan, J.: Motion management for radical ra- diotherapy in non-small cell lung cancer. Clinical oncology26(2), 67–80 (2014) 2

work page 2014
[7]

Nature reviews Disease primers5(1), 13 (2019) 2

De Ruysscher, D., Niedermann, G., Burnet, N.G., Siva, S., Lee, A.W., Hegi- Johnson, F.: Radiotherapy toxicity. Nature reviews Disease primers5(1), 13 (2019) 2

work page 2019
[8]

International Conference on Learning Representations (2020) 5

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An image is worth 16x16 words: Transformers for image recognition at scale. International Conference on Learning Representations (2020) 5

work page 2020

Show all 44 references

[9]

arXiv preprint arXiv:2401.08541 (2024) 3

El-Nouby, A., Klein, M., Zhai, S., Bautista, M.A., Toshev, A., Shankar, V., Susskind, J.M., Joulin, A.: Scalable pre-training of large autoregressive image mod- els. arXiv preprint arXiv:2401.08541 (2024) 3

2024 arXiv
[10]

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

Esser, P., Rombach, R., Ommer, B.: Taming transformers for high-resolution image synthesis. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 12873–12883 (2021) 5 10 Y. Lai et al

2021
[11]

Physics in Medicine & Biology65(20), 20TR01 (2020) 2

Fu, Y.,Lei, Y., Wang, T., Curran,W.J., Liu, T., Yang, X.:Deep learning in medical image registration: a review. Physics in Medicine & Biology65(20), 20TR01 (2020) 2

2020
[12]

Advances in Neural Information Processing Systems33, 6840–6851 (2020) 2, 3

Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems33, 6840–6851 (2020) 2, 3

2020
[13]

Plastic and reconstructive surgery 133(1), 49e–56e (2014) 2

Hubenak, J.R., Zhang, Q., Branch, C.D., Kronowitz, S.J.: Mechanisms of injury to normal tissue after radiotherapy: a review. Plastic and reconstructive surgery 133(1), 49e–56e (2014) 2

2014
[14]

Medical physics44(2), 762–771 (2017) 3, 4, 6

Hugo, G.D., Weiss, E., Sleeman, W.C., Balik, S., Keall, P.J., Lu, J., Williamson, J.F.: A longitudinal four-dimensional computed tomography and cone beam com- puted tomography dataset for image-guided radiation therapy research in lung cancer. Medical physics44(2), 762–771 (20...

2017
[15]

Nature Methods 18(2), 203–211 (2021) 5

Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H.: nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods 18(2), 203–211 (2021) 5

2021
[16]

Physics in Medicine & Biology 57(23), R161 (2012) 2

Korreman, S.S.: Motion in radiotherapy: photon therapy. Physics in Medicine & Biology 57(23), R161 (2012) 2

2012
[17]

Journal of medical imaging and radiation oncology59(5), 545–554 (2015) 3, 4

Kwong, Y., Mel, A.O., Wheeler, G., Troupis, J.M.: F our-dimensional computed tomography (4dct): a review of the current status and applications. Journal of medical imaging and radiation oncology59(5), 545–554 (2015) 3, 4

2015
[18]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Lai, Y., Chen, X., Wang, A., Yuille, A., Zhou, Z.: From pixel to cancer: Cellular automata in computed tomography. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 36–46. Springer (2024) 5

2024
[19]

arXiv preprint arXiv:2503.13939 (2025) 3

Lai, Y., Zhong, J., Li, M., Zhao, S., Yang, X.: Med-r1: Reinforcement learn- ing for generalizable medical reasoning in vision-language models. arXiv preprint arXiv:2503.13939 (2025) 3

2025
[20]

arXiv preprint arXiv:2310.05453 (2023) 2, 5

Lai, Y., Zhou, Y., Liu, X., Zhou, T.: Memory-assisted sub-prototype mining for universal domain adaptation. arXiv preprint arXiv:2310.05453 (2023) 2, 5

2023
[22]

arXiv preprint arXiv:2411.01492 (2024) 3

Li, M., Zhong, J., Chen, T., Lai, Y., Psounis, K.: Eee-bench: A comprehen- sive multimodal electrical and electronics engineering benchmark. arXiv preprint arXiv:2411.01492 (2024) 3

2024 arXiv
[23]

EMNLP 2024 (2024) 3

Li, M., Zhong, J., Li, C., Li, L., Lin, N., Sugiyama, M.: Vision-language model fine-tuning via simple parameter-efficient modification. EMNLP 2024 (2024) 3

2024
[24]

arXiv preprint arXiv:2503.16188 (2025) 3

Li, M., Zhong, J., Zhao, S., Lai, Y., Zhang, K.: Think or not think: A study of explicit thinking in rule-based visual reinforcement fine-tuning. arXiv preprint arXiv:2503.16188 (2025) 3

2025
[25]

Medical Image Analysis 97, 103285 (2024) 5

Li, W., Qu, C., Chen, X., Bassi, P.R., Shi, Y., Lai, Y., Yu, Q., Xue, H., Chen, Y., Lin, X., et al.: Abdomenatlas: A large-scale, detailed-annotated, & multi-center dataset for efficient transfer learning and open algorithmic benchmarking. Medical Image Analysis 97, 103285 (2024) 5

2024
[26]

arXiv preprint arXiv:2503.06030 (2025) 3

Li, Y., Lai, Y., Thor, M., Marshall, D., Buchwald, Z., Yu, D.S., Yang, X.: Towards universal text-driven ct image segmentation. arXiv preprint arXiv:2503.06030 (2025) 3

2025 arXiv
[27]

Medical physics40(4), 041911 (2013) 2, 3

Nie, K., Chuang, C., Kirby, N., Braunstein, S., Pouliot, J.: Site-specific deformable imaging registration algorithm selection using patient-based simulated deforma- tions. Medical physics40(4), 041911 (2013) 2, 3

2013
[28]

Radiation oncology journal 35(2), 101 (2017) 2 Title Suppressed Due to Excessive Length 11

Oh, S., Kim, S.: Deformable image registration in radiation therapy. Radiation oncology journal 35(2), 101 (2017) 2 Title Suppressed Due to Excessive Length 11

2017
[29]

arXiv preprint arXiv:1711.00937 (2017) 2, 3

Oord, A.v.d., Vinyals, O., Kavukcuoglu, K.: Neural discrete representation learn- ing. arXiv preprint arXiv:1711.00937 (2017) 2, 3

2017 arXiv
[30]

Pastor-Serrano, O., Habraken, S., Hoogeman, M., Lathouwers, D., Schaart, D., No- mura, Y., Xing, L., Perkó, Z.: A probabilistic deep learning model of inter-fraction anatomicalvariationsinradiotherapy.PhysicsinMedicine&Biology 68(8),085018 (2023) 2, 3, 6, 7

2023
[31]

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) 2

2022
[32]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedi- cal image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 234–241. Springer (2015) 5

2015
[33]

Nature reviews Clinical oncology12(9), 527–540 (2015) 1

Schaue, D., McBride, W.H.: Opportunities and challenges of radiotherapy for treat- ing cancer. Nature reviews Clinical oncology12(9), 527–540 (2015) 1

2015
[34]

Physics in Medicine & Bi- ology 69(15), 155016 (2024) 2, 3, 6, 7

Smolders, A., Rivetti, L., Vatterodt, N., Korreman, S., Lomax, A., Sharma, M., Studen,A.,Weber,D.C.,Jeraj,R.,Albetini,F.:Diffusert:predictinglikelyanatom- ical deformations of patients undergoing radiotherapy. Physics in Medicine & Bi- ology 69(15), 155016 (2024) 2, 3, 6, 7

2024
[35]

arXiv preprint arXiv:2302.13971 (2023) 6

Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al.: Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023) 6

2023 arXiv
[36]

NeurIPS 2023 (2023) 2

Tu, C.H., Chen, H.Y., Mai, Z., Zhong, J., Pahuja, V., Berger-Wolf, T., Gao, S., Stewart, C., Su, Y., Chao, W.L.: Holistic transfer: Towards non-disruptive fine- tuning with partial target data. NeurIPS 2023 (2023) 2

2023
[37]

International Journal of Radiation Oncology* Biology* Physics 63(1), 253–260 (2005) 3

Underberg, R.W., Lagerwaard, F.J., Slotman, B.J., Cuijpers, J.P., Senan, S.: Use of maximum intensity projections (mip) for target volume generation in 4dct scans for lung cancer. International Journal of Radiation Oncology* Biology* Physics 63(1), 253–260 (2005) 3

2005
[38]

Advances in neural information processing systems30 (2017) 2, 5

Van Den Oord, A., Vinyals, O., et al.: Neural discrete representation learning. Advances in neural information processing systems30 (2017) 2, 5

2017
[39]

Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, L.,Polosukhin,I.:Attentionisallyouneed.arXivpreprintarXiv:1706.03762(2017) 5

2017 arXiv
[40]

Virginia Commonwealth University (2015) 2, 3

Vile, D.J.: Statistical modeling of interfractional tissue deformation and its appli- cation in radiation therapy planning. Virginia Commonwealth University (2015) 2, 3

2015
[41]

Radiology: Artificial Intelligence 5(5) (2023) 5

Wasserthal, J., Breit, H.C., Meyer, M.T., Pradella, M., Hinck, D., Sauter, A.W., Heye, T., Boll, D.T., Cyriac, J., Yang, S., et al.: Totalsegmentator: robust segmen- tation of 104 anatomic structures in ct images. Radiology: Artificial Intelligence 5(5) (2023) 5

2023
[42]

arXiv preprint arXiv:2110.04627 (2021) 3

Yu, J., Li, X., Koh, J.Y., Zhang, H., Pang, R., Qin, J., Ku, A., Xu, Y., Baldridge, J., Wu, Y.: Vector-quantized image modeling with improved vqgan. arXiv preprint arXiv:2110.04627 (2021) 3

2021 arXiv
[43]

Journal of Hepatology 78, S286–S287 (2023) 2

Yu, Q., Zhou, Y., Lai, Y., Qi, X., Ju, S.: A multimodal deep learning network for non-invasive prediction of the hepatic decompensation risk in compensated cirrhotic people: a multicentre cohort study (chess1701). Journal of Hepatology 78, S286–S287 (2023) 2

2023
[44]

In: ECCV 2022, pp

Zhang, C., Pan, T.Y., Chen, T., Zhong, J., Fu, W., Chao, W.L.: Learning with free object segments for long-tailed instance segmentation. In: ECCV 2022, pp. 655–672. Springer Nature Switzerland (2022) 2 12 Y. Lai et al

2022
[45]

FL@NeurIPS 2023 (2023) 2

Zhong*, J., Chen*, H.Y., Chao, W.L.: Making batch normalization great in feder- ated deep learning. FL@NeurIPS 2023 (2023) 2

2023

Pith tools

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

[1] [1]

In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Bai, Y., Geng, X., Mangalam, K., Bar, A., Yuille, A.L., Darrell, T., Malik, J., Efros, A.A.: Sequential modeling enables scalable learning for large vision models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22861–22872 (2024) 3, 6

work page 2024

[2] [2]

Advances in neural information processing systems33, 1877–1901 (2020) 5

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language models are few-shot learners. Advances in neural information processing systems33, 1877–1901 (2020) 5

work page 2020

[3] [3]

The Lancet398(10295), 171–184 (2021) 1

Chandra, R.A., Keane, F.K., Voncken, F.E., Thomas, C.R.: Contemporary radio- therapy: present and future. The Lancet398(10295), 171–184 (2021) 1

work page 2021

[4] [4]

Federated Learning of Shareable Bases for Personalization-Friendly Image Classification

Chen*, H.Y., Zhong*, J., Zhang, M., Jia, X., Qi, H., Gong, B., Chao, W.L., Zhang, L.: Federated learning of shareable bases for personalization-friendly image classi- fication. arXiv preprint arXiv:2304.07882 (2023) 2

work page Pith review arXiv 2023

[5] [5]

Generative Machine Learning Models in Medical Image Computing p

Chen, Q., Lai, Y., Chen, X., Hu, Q., Yuille, A., Zhou, Z.: Analyzing tumors by synthesis. Generative Machine Learning Models in Medical Image Computing p. 85 (2024) 5

work page 2024

[6] [6]

Clinical oncology26(2), 67–80 (2014) 2

Cole, A., Hanna, G., Jain, S., O’Sullivan, J.: Motion management for radical ra- diotherapy in non-small cell lung cancer. Clinical oncology26(2), 67–80 (2014) 2

work page 2014

[7] [7]

Nature reviews Disease primers5(1), 13 (2019) 2

De Ruysscher, D., Niedermann, G., Burnet, N.G., Siva, S., Lee, A.W., Hegi- Johnson, F.: Radiotherapy toxicity. Nature reviews Disease primers5(1), 13 (2019) 2

work page 2019

[8] [8]

International Conference on Learning Representations (2020) 5

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An image is worth 16x16 words: Transformers for image recognition at scale. International Conference on Learning Representations (2020) 5

work page 2020