pith. sign in

arxiv: 2511.17392 · v3 · pith:FATLBYQJnew · submitted 2025-11-21 · 💻 cs.CV

MorphSeek: Fine-grained Latent Representation-Level Policy Optimization for Deformable Image Registration

Runxun Zhang , Yizhou Liu , Li Dongrui , Bo XU , Jingwei Wei This is my paper

Pith reviewed 2026-05-21 18:14 UTC · model grok-4.3

classification 💻 cs.CV
keywords deformable image registrationlatent representationpolicy optimizationreinforcement learningmedical image analysis3D registrationGaussian policy
0
0 comments X

The pith

MorphSeek optimizes deformable image registration by learning stochastic policies over latent features rather than dense displacement fields.

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

The paper establishes that reformulating deformable image registration as a continuous optimization process in latent feature space, guided by a stochastic Gaussian policy head, can capture spatially variant deformations more effectively than prior reinforcement learning approaches that rely on coarse low-dimensional projections. This setup combines unsupervised warm-up with weakly supervised fine-tuning through group relative policy optimization, using multi-trajectory sampling to stabilize learning. The result is higher registration accuracy on 3D medical benchmarks while requiring fewer voxel-level labels and adding little computational overhead. A sympathetic reader would care because traditional methods struggle with the high dimensionality of deformation fields and the expense of dense supervision, so a representation-level policy could scale visual alignment to more clinical settings.

Core claim

MorphSeek introduces a fine-grained representation-level policy optimization paradigm that places a stochastic Gaussian policy head atop the encoder to model distributions over latent features, enabling efficient exploration and coarse-to-fine refinement in the deformation space; the framework demonstrates consistent Dice gains on OASIS brain MRI, LiTS liver CT, and Abdomen MR-CT benchmarks while preserving label efficiency, minimal parameter cost, and low step-level latency.

What carries the argument

The stochastic Gaussian policy head on the encoder, which models a distribution over latent features to support spatially continuous optimization and coarse-to-fine deformation refinement.

If this is right

  • Consistent Dice score gains appear across three 3D registration benchmarks with different modalities.
  • Label efficiency improves through unsupervised warm-up followed by weakly supervised fine-tuning.
  • The method adds only minimal parameters and keeps step-level latency low.
  • The representation-level policy approach is presented as backbone-agnostic and optimizer-agnostic.

Where Pith is reading between the lines

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

  • The same latent-policy structure could be tested on non-medical alignment tasks that also involve high-dimensional continuous spaces.
  • If the Gaussian policy successfully regularizes exploration, similar heads might reduce supervision needs in other dense prediction problems such as optical flow or scene flow.
  • The coarse-to-fine mechanism suggests a natural way to combine this optimizer with multi-scale encoders already common in medical imaging.

Load-bearing premise

That a stochastic Gaussian policy head on the encoder can model a distribution over latent features sufficient to capture spatially variant deformations and enable efficient coarse-to-fine refinement in the high-dimensional deformation space.

What would settle it

Running MorphSeek on the OASIS, LiTS, or Abdomen MR-CT benchmarks and finding no Dice improvement over strong baselines or a sharp rise in parameters or latency would falsify the central performance claims.

Figures

Figures reproduced from arXiv: 2511.17392 by Bo Xu, Jingwei Wei, Li Dongrui, Runxun Zhang, Yizhou Liu.

Figure 1
Figure 1. Figure 1: Two Major Challenges Faced by DL-based DIR [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: MorphSeek Registration Framework Process [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The Performance of MorphSeek Across Three Different Tasks. Labels are overlaid only for the two abdomi [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Validation Dice on OASIS: Effect of Warm-up [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Impact of Warm-up and MorphSeek on GRPO Fine-tuning Performance with Limited Labeled Data (OA￾SIS dataset) 5.2 Policy & Label Efficiency Ablation The ablation on trajectory number and refinement steps on OASIS ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Deformable image registration (DIR) remains a fundamental yet challenging problem in medical image analysis, largely due to the prohibitively high-dimensional deformation space of dense displacement fields and the scarcity of voxel-level supervision. Existing reinforcement learning frameworks often project this space into coarse, low-dimensional representations, limiting their ability to capture spatially variant deformations. We propose MorphSeek, a fine-grained representation-level policy optimization paradigm that reformulates DIR as a spatially continuous optimization process in the latent feature space. MorphSeek introduces a stochastic Gaussian policy head atop the encoder to model a distribution over latent features, facilitating efficient exploration and coarse-to-fine refinement. The framework integrates unsupervised warm-up with weakly supervised fine-tuning through Group Relative Policy Optimization, where multi-trajectory sampling stabilizes training and improves label efficiency. Across three 3D registration benchmarks (OASIS brain MRI, LiTS liver CT, and Abdomen MR-CT), MorphSeek achieves consistent Dice improvements over competitive baselines while maintaining high label efficiency with minimal parameter cost and low step-level latency overhead. Beyond optimizer specifics, MorphSeek advances a representation-level policy learning paradigm that achieves spatially coherent and data-efficient deformation optimization, offering a principled, backbone-agnostic, and optimizer-agnostic solution for scalable visual alignment in high-dimensional settings.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes MorphSeek, a fine-grained latent representation-level policy optimization framework for deformable image registration (DIR). It reformulates the high-dimensional deformation space as optimization in latent feature space by adding a stochastic Gaussian policy head to the encoder, enabling exploration and coarse-to-fine refinement. The method combines unsupervised warm-up with weakly supervised fine-tuning via Group Relative Policy Optimization (GRPO) with multi-trajectory sampling. Evaluations on OASIS brain MRI, LiTS liver CT, and Abdomen MR-CT benchmarks report consistent Dice improvements over competitive baselines, along with high label efficiency, minimal added parameters, and low step-level latency.

Significance. If the empirical claims hold under scrutiny, the work offers a potentially useful shift toward representation-level policy learning for DIR, which could improve scalability and data efficiency in high-dimensional medical image alignment compared to direct displacement-field optimization. The emphasis on backbone- and optimizer-agnostic design, if substantiated, would be a constructive contribution to RL applications in visual registration tasks.

major comments (2)
  1. [§3.1] §3.1 (Policy Head Definition): The stochastic Gaussian policy is placed directly on encoder features without explicit spatial conditioning, convolutional structure, or position-aware parameterization in the latent distribution. This is load-bearing for the central claim that the approach captures spatially variant deformations; if the policy remains effectively global or channel-factorized, sampled trajectories cannot represent the local, high-frequency non-rigid displacements that distinguish DIR from coarse alignment, and reported Dice gains may instead stem from the unsupervised warm-up or GRPO components.
  2. [§4.3, Table 2] §4.3 and Table 2: Dice improvements are presented as 'consistent' across the three benchmarks, yet no per-run standard deviations, error bars, or statistical significance tests (e.g., paired t-tests or Wilcoxon) are reported. Without these, it is impossible to determine whether the observed gains exceed baseline variability, undermining the cross-benchmark superiority claim.
minor comments (2)
  1. [Abstract, §2.2] The abstract and §2.2 refer to 'minimal parameter cost' and 'low step-level latency overhead' without providing concrete numbers (e.g., additional parameters or ms per step relative to the base network).
  2. [§3.3] Notation for the GRPO objective in §3.3 could be clarified by explicitly distinguishing group-level versus trajectory-level quantities in the relative advantage computation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and have made revisions to strengthen the presentation of our method and results.

read point-by-point responses

  1. Referee: [§3.1] §3.1 (Policy Head Definition): The stochastic Gaussian policy is placed directly on encoder features without explicit spatial conditioning, convolutional structure, or position-aware parameterization in the latent distribution. This is load-bearing for the central claim that the approach captures spatially variant deformations; if the policy remains effectively global or channel-factorized, sampled trajectories cannot represent the local, high-frequency non-rigid displacements that distinguish DIR from coarse alignment, and reported Dice gains may instead stem from the unsupervised warm-up or GRPO components.

    Authors: We appreciate the referee's careful reading of §3.1. The encoder backbone is a convolutional network that outputs spatially structured feature maps with explicit height, width, and depth dimensions. The stochastic Gaussian policy head is applied directly to these feature maps, so the modeled distribution inherits the spatial resolution of the latent representation rather than operating globally or in a purely channel-wise manner. Sampled trajectories therefore correspond to spatially localized adjustments in the deformation field, supporting the coarse-to-fine refinement described in the paper. We have revised §3.1 to make this spatial inheritance explicit and to clarify that no additional position-aware parameterization was introduced because the convolutional encoder already supplies the necessary locality. revision: yes

  2. Referee: [§4.3, Table 2] §4.3 and Table 2: Dice improvements are presented as 'consistent' across the three benchmarks, yet no per-run standard deviations, error bars, or statistical significance tests (e.g., paired t-tests or Wilcoxon) are reported. Without these, it is impossible to determine whether the observed gains exceed baseline variability, undermining the cross-benchmark superiority claim.

    Authors: We agree that the absence of variability measures and statistical tests limits the strength of the empirical claims. In the revised version we have recomputed all Dice scores over five independent runs, added standard deviations to Table 2, and included paired t-test p-values comparing MorphSeek against each baseline on every benchmark. These additions are now reported in §4.3 and the updated table. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on empirical benchmarks

full rationale

The paper proposes MorphSeek as a representation-level policy optimization method for DIR, introducing a stochastic Gaussian policy head on encoder features and integrating unsupervised warm-up with GRPO. All central claims of Dice improvements and label efficiency are presented as outcomes of experiments on OASIS, LiTS, and Abdomen MR-CT benchmarks rather than derived from equations that reduce to fitted inputs or self-citations by construction. No load-bearing derivations, uniqueness theorems, or ansatzes are visible that would collapse the method to its own inputs; the framework is described as backbone-agnostic and validated through external performance metrics.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only view supplies no explicit free parameters, axioms, or invented entities; the approach implicitly assumes standard encoder-decoder architectures and RL policy gradients but does not detail any ad-hoc choices.

pith-pipeline@v0.9.0 · 5763 in / 1125 out tokens · 53220 ms · 2026-05-21T18:14:50.082455+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

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

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

64 extracted references · 64 canonical work pages · 6 internal anchors

  1. [1]

    Avants, C.L

    B.B. Avants, C.L. Epstein, M. Grossman, and J.C. Gee. Symmetric diffeomorphic image registration with cross-correlation: Evaluating automated label- ing of elderly and neurodegenerative brain. Medical Image Analysis , 12(1):26–41, 2008. Special Issue on The Third International Workshop on Biomedi- cal Image Registration WBIR 2006

    work page 2008

  2. [2]

    Multiresolution elastic matching

    Ruzena Bajcsy and Stane Kovai. Multiresolution elastic matching. Computer Vision, Graphics, and Image Processing, 46(1):1–21, 1989

    work page 1989

  3. [3]

    Sabuncu, John Guttag, and Adrian V

    Guha Balakrishnan, Amy Zhao, Mert R. Sabuncu, John Guttag, and Adrian V . Dalca. V oxelmorph: A learning framework for deformable medical image registration. IEEE Transactions on Medical Imag- ing, 38(8):17881800, August 2019

    work page 2019

  4. [4]

    The liver tumor segmentation benchmark (lits)

    Patrick Bilic, Patrick Christ, Hongwei Bran Li, Eu- gene 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 Anal- ysis, 84:102680, 2023

    work page 2023

  5. [5]

    Deep learning based inter-modality image registration supervised by intra-modality similarity, 2018

    Xiaohuan Cao, Jianhua Y ang, Li Wang, Zhong Xue, Qian Wang, and Dinggang Shen. Deep learning based inter-modality image registration supervised by intra-modality similarity, 2018

    work page 2018

  6. [6]

    Frey, Y ufan He, William P

    Junyu Chen, Eric C. Frey, Y ufan He, William P . Segars, Y e Li, and Y ong Du. Transmorph: Trans- former for unsupervised medical image registra- tion. Medical Image Analysis , 82:102615, Novem- ber 2022

    work page 2022

  7. [7]

    Pretraining deformable image regis- tration networks with random images, 2025

    Junyu Chen, Shuwen Wei, Yihao Liu, Aaron Carass, and Y ong Du. Pretraining deformable image regis- tration networks with random images, 2025

    work page 2025

  8. [8]

    Zeyuan Chen, Y uanjie Zheng, and James C. Gee. Transmatch: A transformer-based multilevel dual- stream feature matching network for unsupervised deformable image registration. IEEE Transactions on Medical Imaging, 43(1):15–27, 2024

    work page 2024

  9. [9]

    The cancer imaging archive (tcia): maintaining and operating a public informa- tion repository

    K Clark, B V endt, K Smith, J Freymann, J Kirby, P Koppel, S Moore, S Phillips, D Maffitt, M Pringle, L Tarbox, and F Prior. The cancer imaging archive (tcia): maintaining and operating a public informa- tion repository. J Digit Imaging , 26(6):1045–1057, Dec 2013

    work page 2013

  10. [10]

    Unsupervised Learning for Fast Probabilistic Diffeomorphic Registration

    Adrian V . Dalca, Guha Balakrishnan, John V . Gut- tag, and Mert R. Sabuncu. Unsupervised learning for 9 arXiv Template A P REPRINT fast probabilistic diffeomorphic registration. CoRR, abs/1805.04605, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  11. [11]

    DeepSeek-AI, Daya Guo, Dejian Y ang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xi- aokang Zhang, Xingkai Y u, Y u Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, Da...

    work page 2025

  12. [12]

    Lee R. Dice. Measures of the amount of ecologic as- sociation between species. Ecology, 26(3):297–302, 1945

    work page 1945

  13. [13]

    Weakly supervised learning of metric aggregations for deformable image registration

    Enzo Ferrante, Puneet Kumar Dokania, Rafael Marini Silva, and Nikos Paragios. Weakly supervised learning of metric aggregations for deformable image registration. IEEE Journal of Biomedical and Health Informatics , 23(4):1374– 1384, 2019

    work page 2019

  14. [14]

    Deep learning in medi- cal image registration: a review

    Y abo Fu, Y ang Lei, Tonghe Wang, Walter J Curran, Tian Liu, and Xiaofeng Y ang. Deep learning in medi- cal image registration: a review. Physics in Medicine & Biology, 65(20):20TR01, October 2020

    work page 2020

  15. [15]

    Learning deep similarity met- ric for 3d mr-trus image registration

    G Haskins, J Kruecker, U Kruger, S Xu, PA Pinto, BJ Wood, and P Y an. Learning deep similarity met- ric for 3d mr-trus image registration. Int J Comput Assist Radiol Surg, 14(3):417–425, Mar 2019

    work page 2019

  16. [16]

    Mind: modality independent neighbourhood descriptor for multi-modal deformable registration

    MP Heinrich, M Jenkinson, M Bhushan, T Matin, FV Gleeson, SM Brady, and JA Schnabel. Mind: modality independent neighbourhood descriptor for multi-modal deformable registration. Med Image Anal, 16(7):1423–1435, Oct 2012

    work page 2012

  17. [17]

    Learn2reg: comprehen- sive multi-task medical image registration challenge, dataset and evaluation in the era of deep learning

    Alessa Hering, Lasse Hansen, Tony CW Mok, Al- bert CS Chung, et al. Learn2reg: comprehen- sive multi-task medical image registration challenge, dataset and evaluation in the era of deep learning. IEEE Transactions on Medical Imaging , 42(3):697– 712, 2022

    work page 2022

  18. [18]

    Synthmorph: Learning contrast- invariant registration without acquired images

    M Hoffmann, B Billot, DN Greve, JE Iglesias, B Fis- chl, and A V Dalca. Synthmorph: Learning contrast- invariant registration without acquired images. IEEE Trans Med Imaging, 41(3):543–558, Mar 2022

    work page 2022

  19. [19]

    End-to-end multimodal image registration via reinforcement learning

    Jing Hu, Ziwei Luo, Xin Wang, Shanhui Sun, Y oub- ing Yin, Kunlin Cao, Qi Song, Siwei Lyu, and Xi Wu. End-to-end multimodal image registration via reinforcement learning. Medical Image Analysis, 68:101878, 2021

    work page 2021

  20. [20]

    Moore, Mark Emberton, J

    Yipeng Hu, Marc Modat, Eli Gibson, Nooshin Ghavami, Ester Bonmati, Caroline M. Moore, Mark Emberton, J. Alison Noble, Dean C. Barratt, and Tom V ercauteren. Label-driven weakly-supervised learning for multimodal deformable image registra- tion. In 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018) . IEEE, April 2018

    work page 2018

  21. [21]

    Weakly-Supervised Convolutional Neural Networks for Multimodal Image Registration

    Yipeng Hu, Marc Modat, Eli Gibson, Wenqi Li, Nooshin Ghavami, Ester Bonmati, Guotai Wang, Steven Bandula, Caroline M. Moore, Mark Ember- ton, Sébastien Ourselin, J. Alison Noble, Dean C. Barratt, and Tom V ercauteren. Weakly-supervised convolutional neural networks for multimodal image registration. CoRR, abs/1807.03361, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  22. [22]

    Styles, Hyung Jin Chang, Ales Leonardis, Antonio de Marvao, Declan P

    Xi Jia, Alexander Thorley, Wei Chen, Huaqi Qiu, Linlin Shen, Iain B. Styles, Hyung Jin Chang, Ales Leonardis, Antonio de Marvao, Declan P . O’Regan, Daniel Rueckert, and Jinming Duan. Learning a model-driven variational network for deformable im- age registration. CoRR, abs/2105.12227, 2021. 10 arXiv Template A P REPRINT

    work page arXiv 2021

  23. [23]

    Emre Kavur, N

    A. Emre Kavur, N. Sinem Gezer, Mustafa Bar, Sinem Aslan, Pierre-Henri Conze, Vladimir Groza, Duc Duy Pham, Soumick Chatterjee, Philipp Ernst, Sava Özkan, Bora Baydar, Dmitry Lachinov, Shuo Han, Josef Pauli, Fabian Isensee, Matthias Perkonigg, Rachana Sathish, Ronnie Rajan, Deb- doot Sheet, Gurbandurdy Dovletov, Oliver Speck, Andreas Nürnberger, Klaus H. M...

    work page 2021

  24. [24]

    Unsupervised Deformable Image Registration Using Cycle-Consistent CNN

    Boah Kim, Jieun Kim, June-Goo Lee, Dong Hwan Kim, Seong Ho Park, and Jong Chul Y e. Unsu- pervised deformable image registration using cycle- consistent CNN. CoRR, abs/1907.01319, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  25. [25]

    Ghesu, Shun Miao, Andreas K

    Julian Krebs, Tommaso Mansi, Hervé Delingette, Li Zhang, Florin C. Ghesu, Shun Miao, Andreas K. Maier, Nicholas Ayache, Rui Liao, and Ali Kamen. Robust non-rigid registration through agent-based action learning. In Maxime Descoteaux, Lena Maier- Hein, Alfred M. Franz, Pierre Jannin, D. Louis Collins, and Simon Duchesne, editors, Medical Im- age Computing ...

    work page 2017

  26. [26]

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. Imagenet classification with deep convolu- tional neural networks. Commun. ACM, 60(6):8490, May 2017

    work page 2017

  27. [27]

    On Reducing Negative Jacobian Determinant of the Deformation Predicted by Deep Registration Networks

    Dongyang Kuang. On reducing negative jacobian de- terminant of the deformation predicted by deep reg- istration networks. CoRR, abs/1907.00068, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  28. [28]

    Yiwen Li, Y unguan Fu, Iani J. M. B. Gayo, Qianye Y ang, Zhe Min, Shaheer U. Saeed, Wen Y an, Yipei Wang, J. Alison Noble, Mark Emberton, Matthew J. Clarkson, Dean C. Barratt, Victor A. Prisacariu, and Yipeng Hu. Semi-weakly-supervised neural network training for medical image registration, 2024

    work page 2024

  29. [29]

    An Artificial Agent for Robust Image Registration

    Rui Liao, Shun Miao, Pierre de Tournemire, Sasa Grbic, Ali Kamen, Tommaso Mansi, and Dorin Co- maniciu. An artificial agent for robust image regis- tration. CoRR, abs/1611.10336, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  30. [30]

    Stochastic planner-actor-critic for unsupervised de- formable image registration

    Ziwei Luo, Jing Hu, Xin Wang, Shu Hu, Bin Kong, Y oubing Yin, Qi Song, Xi Wu, and Siwei Lyu. Stochastic planner-actor-critic for unsupervised de- formable image registration. In Proceedings of the AAAI Conference on Artificial Intelligence , vol- ume 36, pages 1917–1925, 2022

    work page 1917

  31. [31]

    A spatiotemporal agent for robust multimodal registration

    Ziwei Luo, Xin Wang, Xi Wu, Y oubing Yin, Kun- lin Cao, Qi Song, and Jing Hu. A spatiotemporal agent for robust multimodal registration. IEEE Ac- cess, 8:75347–75358, 2020

    work page 2020

  32. [32]

    Multimodal image registration with deep context reinforcement learning

    Kai Ma, Jiangping Wang, Vivek Singh, Birgi Tamer- soy, Y ao-Jen Chang, Andreas Wimmer, and Ter- rence Chen. Multimodal image registration with deep context reinforcement learning. In Maxime Descoteaux, Lena Maier-Hein, Alfred Franz, Pierre Jannin, D. Louis Collins, and Simon Duchesne, edi- tors, Medical Image Computing and Computer As- sisted Intervention...

    work page 2017

  33. [34]

    Pivit: Large deformation image registration with˘apyramid-iterative vision transformer

    Tai Ma, Xinru Dai, Suwei Zhang, and Ying Wen. Pivit: Large deformation image registration with˘apyramid-iterative vision transformer. In Hayit Greenspan, Anant Madabhushi, Parvin Mousavi, Septimiu Salcudean, James Duncan, Tanveer Syeda- Mahmood, and Russell Taylor, editors, Medical Im- age Computing and Computer Assisted Interven- tion – MICCAI 2023 , pag...

    work page 2023

  34. [35]

    Viergever

    J.B.Antoine Maintz and Max A. Viergever. A sur- vey of medical image registration. Medical Image Analysis, 2(1):1–36, 1998

    work page 1998

  35. [36]

    Milone, and Enzo Fer- rante

    Lucas Mansilla, Diego H. Milone, and Enzo Fer- rante. Learning deformable registration of medi- cal images with anatomical constraints. Neural Net- works, 124:269–279, 2020

    work page 2020

  36. [37]

    Open access series of imaging studies (oasis): cross-sectional mri data in young, middle aged, nondemented, and demented older adults

    DS Marcus, TH Wang, J Parker, JG Csernansky, JC Morris, and RL Buckner. Open access series of imaging studies (oasis): cross-sectional mri data in young, middle aged, nondemented, and demented older adults. J Cogn Neurosci , 19(9):1498–1507, Sep 2007

    work page 2007

  37. [38]

    Non-iterative Coarse-to-Fine Registra- tion Based on Single-Pass Deep Cumulative Learn- ing, page 8897

    Mingyuan Meng, Lei Bi, Dagan Feng, and Jin- man Kim. Non-iterative Coarse-to-Fine Registra- tion Based on Single-Pass Deep Cumulative Learn- ing, page 8897. Springer Nature Switzerland, 2022

    work page 2022

  38. [39]

    Non-iterative Coarse-to- Fine Transformer Networks for Joint Affine and De- formable Image Registration, page 750760

    Mingyuan Meng, Lei Bi, Michael Fulham, Dagan Feng, and Jinman Kim. Non-iterative Coarse-to- Fine Transformer Networks for Joint Affine and De- formable Image Registration, page 750760. Springer Nature Switzerland, 2023

    work page 2023

  39. [40]

    Correlation-aware coarse-to-fine mlps for de- formable medical image registration

    Mingyuan Meng, Dagan Feng, Lei Bi, and Jinman Kim. Correlation-aware coarse-to-fine mlps for de- formable medical image registration. In IEEE/CVF Conference on Computer Vision and Pattern Recog- nition (CVPR), pages 9645–9654, 2024

    work page 2024

  40. [41]

    Agent-Based Methods for Medical Image Registration , pages 323–345

    Shun Miao and Rui Liao. Agent-Based Methods for Medical Image Registration , pages 323–345. Springer International Publishing, Cham, 2019

    work page 2019

  41. [42]

    Ridgway, Zeike A

    Marc Modat, Gerard R. Ridgway, Zeike A. Tay- lor, Manja Lehmann, Josephine Barnes, David J. Hawkes, Nick C. Fox, and Sébastien Ourselin. Fast free-form deformation using graphics process- ing units. Computer Methods and Programs in Biomedicine, 98(3):278–284, 2010. HP-MICCAI 2008. 11 arXiv Template A P REPRINT

    work page 2010

  42. [43]

    Tony C. W. Mok and Albert C. S. Chung. Large deformation diffeomorphic image registration with laplacian pyramid networks, 2020

    work page 2020

  43. [44]

    Pluim, J.B.A

    J.P .W. Pluim, J.B.A. Maintz, and M.A. Viergever. Mutual-information-based registration of medical images: a survey. IEEE Transactions on Medical Imaging, 22(8):986–1004, 2003

    work page 2003

  44. [45]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y . K. Li, Y . Wu, and Daya Guo. Deepseek- math: Pushing the limits of mathematical reasoning in open language models, 2024

    work page 2024

  45. [46]

    Davatzikos

    Dinggang Shen and C. Davatzikos. Hammer: hierar- chical attribute matching mechanism for elastic reg- istration. IEEE Transactions on Medical Imaging , 21(11):1421–1439, 2002

    work page 2002

  46. [47]

    Medical image registration based on˘auncoupled learning and accumulative enhance- ment

    Y ucheng Shu, Hao Wang, Bin Xiao, Xiuli Bi, and Weisheng Li. Medical image registration based on˘auncoupled learning and accumulative enhance- ment. In Marleen de Bruijne, Philippe C. Cattin, Stéphane Cotin, Nicolas Padoy, Stefanie Speidel, Y efeng Zheng, and Caroline Essert, editors, Medi- cal Image Computing and Computer Assisted Inter- vention – MICCAI...

    work page 2021

  47. [48]

    Hessam Sokooti, Bob de V os, Floris Berendsen, Mohsen Ghafoorian, Sahar Y ousefi, Boudewijn P . F. Lelieveldt, Ivana Isgum, and Marius Staring. 3d con- volutional neural networks image registration based on efficient supervised learning from artificial defor- mations, 2019

    work page 2019

  48. [49]

    Hessam Sokooti, Bob de V os, Floris Berendsen, Boudewijn P . F. Lelieveldt, Ivana Išgum, and Mar- ius Staring. Nonrigid image registration using multi- scale 3d convolutional neural networks. In Maxime Descoteaux, Lena Maier-Hein, Alfred Franz, Pierre Jannin, D. Louis Collins, and Simon Duchesne, edi- tors, Medical Image Computing and Computer As- sisted ...

    work page 2017

  49. [51]

    Deformable medical image registration: A survey

    Aristeidis Sotiras, Christos Davatzikos, and Nikos Paragios. Deformable medical image registration: A survey. IEEE Transactions on Medical Imaging , 32(7):1153–1190, 2013

    work page 2013

  50. [52]

    Robust mul- timodal image registration using deep recurrent re- inforcement learning

    Shanhui Sun, Jing Hu, Mingqing Y ao, Jinrong Hu, Xiaodong Y ang, Qi Song, and Xi Wu. Robust mul- timodal image registration using deep recurrent re- inforcement learning. In C. V . Jawahar, Hongdong Li, Greg Mori, and Konrad Schindler, editors, Com- puter Vision – ACCV 2018 , pages 511–526, Cham,

    work page 2018

  51. [54]

    A framework for deformable image registration validation in radiotherapy clinical applications

    Raj V aradhan, Grigorios Karangelis, Karthik Krish- nan, and Susanta Hui. A framework for deformable image registration validation in radiotherapy clinical applications. Journal of Applied Clinical Medical Physics, 14(1):192–213, 01 2013

    work page 2013

  52. [55]

    Attention Is All You Need

    Ashish V aswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. CoRR, abs/1706.03762, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  53. [56]

    Deepflash: An efficient network for learning-based medical image registration, 2020

    Jian Wang and Miaomiao Zhang. Deepflash: An efficient network for learning-based medical image registration, 2020

    work page 2020

  54. [57]

    On-the-Fly Guidance Training for Medical Image Registration

    Y uelin Xin, Yicheng Chen, Shengxiang Ji, Kun Han, and Xiaohui Xie. On-the-Fly Guidance Training for Medical Image Registration. arXiv e-prints , page arXiv:2308.15216, August 2023

    work page arXiv 2023

  55. [58]

    Eval- uation of six registration methods for the human ab- domen on clinically acquired ct

    Z Xu, CP Lee, MP Heinrich, M Modat, D Rueckert, S Ourselin, RG Abramson, and BA Landman. Eval- uation of six registration methods for the human ab- domen on clinically acquired ct. IEEE Trans Biomed Eng, 63(8):1563–1572, Aug 2016

    work page 2016

  56. [59]

    Fast predictive image registration

    Xiao Y ang, Roland Kwitt, and Marc Nietham- mer. Fast predictive image registration. In Gus- tavo Carneiro, Diana Mateus, Loïc Peter, Andrew Bradley, João Manuel R. S. Tavares, V asileios Be- lagiannis, João Paulo Papa, Jacinto C. Nascimento, Marco Loog, Zhi Lu, Jaime S. Cardoso, and Julien Cornebise, editors, Deep Learning and Data Label- ing for Medical...

    work page

  57. [60]

    Springer International Publishing

    work page

  58. [61]

    Recurrent inference ma- chine for medical image registration

    Yi Zhang, Yidong Zhao, Hui Xue, Peter Kellman, Stefan Klein, and Qian Tao. Recurrent inference ma- chine for medical image registration. Medical Image Analysis, 106:103748, 2025

    work page 2025

  59. [62]

    Recursive cascaded networks for unsuper- vised medical image registration

    Shengyu Zhao, Y ue Dong, Eric I-Chao Chang, and Y an Xu. Recursive cascaded networks for unsuper- vised medical image registration. In Proceedings of the IEEE International Conference on Computer Vi- sion (ICCV), 2019

    work page 2019

  60. [63]

    Unsupervised 3d end-to-end medical image registration with volume tweening network

    Shengyu Zhao, Tingfung Lau, Ji Luo, Eric I Chang, and Y an Xu. Unsupervised 3d end-to-end medical image registration with volume tweening network. IEEE Journal of Biomedical and Health Informatics, 2019

    work page 2019

  61. [64]

    Unsupervised deformable medi- cal image registration via pyramidal residual defor- mation fields estimation

    Y ujia Zhou, Shumao Pang, Jun Cheng, Y uhang Sun, Yi Wu, Lei Zhao, Y aqin Liu, Zhentai Lu, Wei Y ang, and Qianjin Feng. Unsupervised deformable medi- cal image registration via pyramidal residual defor- mation fields estimation. CoRR, abs/2004.07624, 2020

    work page arXiv 2004

  62. [65]

    Z. Zhou, B. Hong, X. Qian, et al. macjnet: weakly- supervised multimodal image deformable registra- tion using joint learning framework and multi- sampling cascaded mind. BioMed Eng OnLine , 22:91, 2023

    work page 2023

  63. [66]

    Functional mag- netic resonance imaging progressive deformable reg- istration based on a cascaded convolutional neural network

    Qiaoyun Zhu, Guoye Lin, Y uhang Sun, Yi Wu, Y ujia Zhou, and Qianjin Feng. Functional mag- netic resonance imaging progressive deformable reg- istration based on a cascaded convolutional neural network. Quantitative Imaging in Medicine and Surgery, 11(8), 2021. 12 arXiv Template A P REPRINT

    work page 2021

  64. [67]

    Swin-voxelmorph: A sym- metric unsupervised learning model for ˘adeformable medical image registration using swin transformer

    Y ongpei Zhu and Shi Lu. Swin-voxelmorph: A sym- metric unsupervised learning model for ˘adeformable medical image registration using swin transformer. In Linwei Wang, Qi Dou, P . Thomas Fletcher, Ste- fanie Speidel, and Shuo Li, editors, Medical Im- age Computing and Computer Assisted Intervention – MICCAI 2022, pages 78–87, Cham, 2022. Springer Nature S...

    work page 2022