Pith. sign in

REVIEW 4 major objections 6 minor 34 references

VoxelOpt: Voxel-Adaptive Message Passing for Discrete Optimization in Deformable Abdominal CT Registration

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read A discrete optimizer with entropy-guided adaptive smoothing registers abdominal CT scans as accurately as semi-supervised networks, without any training.

desk verdict Entropy-guided adaptive message passing is a real contribution, but the 'no label supervision' claim is undercut by reliance on a label-trained foundation model. read the letter →

arxiv 2506.19975 v1 pith:ALDEFYHY submitted 2025-06-24 eess.IV cs.AIcs.CVeess.SP

classification eess.IVcs.AIcs.CVeess.SP
keywords deformableimageregistrationdiscreteoptimizationmessagepassingdisplacemententropyabdominalCTmean-fieldinferencefoundationmodelfeaturesadaptivefiltering
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

VoxelOpt aims to close the gap between slow iterative registration and fast learning-based registration for deformable abdominal CT. The paper argues that the entropy of a local displacement cost volume measures how confident each voxel is in its own displacement, and that this signal should control how much each voxel listens to its neighbors: confident voxels keep their displacement, uncertain voxels borrow from neighbors. Using this entropy-adaptive message passing inside a multi-level discrete optimizer, with features from a pretrained segmentation model, VoxelOpt reports Dice scores on par with the best label-supervised learning method and better than the best iterative method, all in under one second and with no training on the target data. A sympathetic reader would take this as evidence that discrete optimization can be made both fast and accurate when the smoothing schedule is driven by per-voxel signal strength rather than a fixed isotropic regularizer.

What carries the argument

The load-bearing object is the per-voxel displacement entropy $E(x)$, computed by converting a 27-neighbor cost volume $C^1(x)$ into a probability distribution via softmax with temperature $\beta$ and taking the entropy of that distribution. Entropy is then normalized and mapped by $\sigma(x) = \alpha \log(E(x)/\max(E) + 1)$ with $\alpha = 1.5$ to a per-voxel Gaussian blur width used during message passing. This adaptive smoothing replaces the isotropic regularizer used in earlier discrete and learning-based registration, and it is what lets the optimizer propagate displacement signals from boundary and texture-rich voxels into uniform regions in only six iterations.

What would settle it

Run the same six-iteration VoxelOpt pipeline on a second abdominal CT dataset from a different scanner or contrast protocol, comparing the adaptive version against the non-adaptive version (ablation #3 vs #4): if the adaptive component does not consistently improve Dice, the entropy-to-smoothing map is not a general mechanism. A second check would be to compare the list of images the foundation model was pretrained on against the 30-scan evaluation set; any overlap would change the interpretation of the training-free claim.

Watch

Extended reading notes

Core claim

The central claim is that a training-free discrete optimization framework can match semi-supervised deep registration accuracy on abdominal CT by making the smoothing step of mean-field inference voxel-adaptive. Specifically, VoxelOpt computes a probabilistic cost volume over 27 neighbor displacements at each pyramid level, derives per-voxel displacement entropy, and maps that entropy through a fixed logarithmic rule to a per-voxel Gaussian blur width. Voxels with strong displacement signals (low entropy) are smoothed less, preserving their information; voxels with weak signals (high entropy) are smoothed more, receiving more neighbor influence. Together with a five-level image pyramid, a 27-neighbor search per level, and feature maps from a pretrained foundational segmentation model, this reaches 58.5% Dice on the testing split, compared with 53.6% for the best iterative baseline and 58.8% for the best semi-supervised network, at sub-second runtime.

Load-bearing premise

The whole accuracy gain from adaptivity rests on the untested assumption that the fixed logarithmic formula $\sigma(x)=1.5\log(E(x)/\max(E)+1)$ turns cost-volume entropy into the right per-voxel blur strength for every anatomy and scanning protocol.

Editorial extensions

If this is right

  • Sub-second deformable registration becomes achievable without any training or label supervision, which matters for low-data clinical settings where semi-supervised networks cannot be trained.
  • The entropy of a local cost volume provides a usable per-voxel confidence signal, so the same adaptive message-passing idea could be attached to other discrete optimization solvers.
  • Pretrained segmentation features can replace hand-crafted or contrastively learned features for registration, removing an entire training stage.
  • A fixed small search neighborhood (27 neighbors) with an image pyramid suffices; enlarging the kernel only adds runtime and degrades smoothness, so exponential search-space growth is avoidable.

Reading between the lines

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

  • Inference beyond the paper: the same entropy-gated smoothing could be applied to other per-voxel discrete estimates, such as optical flow or stereo disparity, where confidence varies spatially.
  • Inference beyond the paper: the logarithmic mapping in Eq. (6) is one of many possible entropy-to-blur curves; scanning $\alpha$ or fitting the mapping on a small validation set would show how much of the 3.7% Dice gain depends on that particular choice.
  • Inference beyond the paper: because the evaluation set is a public 30-scan dataset, the pretrained segmentation model's training-data overlap with it is a concrete risk; checking for overlap would determine whether the reported gains transfer to unseen scanners.
  • Inference beyond the paper: since the optimizer is differentiable and runs in a single forward pass, it could be unrolled and fine-tuned end-to-end, which the paper mentions as a possibility but does not test.
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. The paper proposes VoxelOpt, a training-free discrete optimization framework for deformable abdominal CT registration. The method builds a 27-neighbor cost volume from feature maps, computes per-voxel displacement entropy from that cost volume, and uses the entropy to set a per-voxel Gaussian blur width that controls adaptive message passing. Large deformations are handled with a multi-level Laplacian pyramid, and the feature extractor is a pretrained foundational segmentation model. Experiments on 42 test pairs from a public abdominal CT dataset report 58.51% Dice with sub-second runtime, matching a semi-supervised RDP baseline and outperforming iterative and unsupervised learning baselines. Ablations attribute 12.8 Dice points to the foundation-model features and 3.7 Dice points to the adaptive message-passing component.

Significance. If the results hold, the paper demonstrates a useful point: a non-learning discrete optimizer with entropy-guided adaptive smoothing and features from a pretrained segmentation model can reach accuracy comparable to semi-supervised registration networks while running in under a second. The multi-level pyramid with fixed 27-neighbor cost volumes is a sensible way to avoid exponential search-space growth, and the entropy-based interpretation of feature quality for registration is an interesting diagnostic idea. The paper includes a concrete held-out test split and ablation studies for feature type, kernel size, adaptive message passing, and cost-volume pre-filtering. However, the central claim of label-supervision-free performance depends on a label-trained foundation model whose pretraining data overlap with the evaluation set is never analyzed, and the core adaptive mapping is introduced heuristically without sensitivity analysis. The empirical evidence is also based on a single small test set without significance testing.

major comments (4)
  1. [Sections 3.1, 3.2, and 4] The claim that VoxelOpt achieves state-of-the-art accuracy 'without relying on label supervision' is not established as stated. The dominant accuracy gain comes from features extracted by the foundational segmentation model [19], which was trained with dense organ-segmentation labels on large CT corpora, and Table 2 shows that this feature source contributes 12.8 Dice points (#1 vs #3). The evaluation dataset [28] is a public abdominal CT dataset with 13-organ segmentations, but the paper never checks whether [19]'s pretraining data includes scans from [28] or from the same patient cohort. If such overlap exists, the reported 58.51 Dice partly reflects memorized segmentation knowledge leaking into the cost volume. The authors should either demonstrate non-overlap of the pretraining and evaluation data and quantify the sensitivity to that overlap, or reframe the contribution as registration without training a dedicated network and provide results with features that are not label-trained.
  2. [Equation (6) and Section 3.4] The voxel-adaptive message-passing contribution rests entirely on the mapping sigma(x) = alpha * log(E(x)/max(E) + 1), but the paper provides no derivation, optimality argument, or sensitivity analysis for this mapping. The ablation in Table 2 (#3 vs #4) reports only 3.7 Dice points for the adaptive term, and it is unclear whether this gain is robust to the choices of alpha = 1.5, beta, and the theta schedule used in Eq. (5). Please report results for a range of alpha and beta values, and ideally compare the log mapping against a simpler alternative such as a linear or rank-based mapping, so readers can assess whether the specific functional form is load-bearing for the claimed improvement.
  3. [Tables 1 and 2, Section 3.3] All accuracy comparisons are based on 42 test pairs from a single dataset, and no statistical significance tests, confidence intervals, or per-pair variance measures are reported. Several headline comparisons are close (VoxelOpt at 58.51 Dice vs. RDP semi-supervised at 58.77 Dice), and the claimed parity/outperformance should be supported by paired tests (e.g., Wilcoxon signed-rank) and confidence intervals for Dice and HD95. Without this, the differences between methods in Table 1 may be within noise, and the conclusion that VoxelOpt 'matches' or 'outperforms' specific baselines is not yet quantitatively supported.
  4. [Sections 2.4, 2.5, and 3.2] The implementation of the proposed spatially varying Gaussian filtering is underspecified. Equation (6) defines a per-voxel sigma field, but the paper does not state how a space-variant Gaussian is applied to the 6D cost volume or to the displacement field; the statement that adaptive 3D Gaussian filtering is implemented as three separable 1D filters is insufficient because space-variant filtering is not separable in general. Since this filtering is the core mechanism of the method, please provide the exact discretization, including how per-voxel sigma is converted to filter kernels and how border conditions are handled, or make the code available in the review version.
minor comments (6)
  1. [Throughout] There are several spacing artifacts such as 'V oxelOpt' and 'T able 1' that should be corrected during production.
  2. [Section 3.4] The text says the cost volume grows from (2*1+1)^3 to (2*2+1)^3 and (3*1+1)^3; the last expression should be (2*3+1)^3 for kernel size k = 3.
  3. [Sections 1 and 3.4] The paper uses both '27-neighbor' and '26-neighborhood' for the same k = 1 cost volume; please standardize the terminology.
  4. [Equations (2)-(4)] The symbol L1 is used both for the label of the displacement neighborhood and for the L1 dissimilarity function mentioned in Section 3.2, which is confusing; please disambiguate the notation.
  5. [Section 2.5 and Figure 2] The pre-filtering of the cost volume (the step that produces C-tilde from Ck) is mentioned in the framework description and ablation but is not defined by an equation; a precise definition would improve reproducibility.
  6. [Section 3.2] Please clarify whether 'pre-softmax features' are taken from a single layer or from multiple layers of the foundation model, and which layer is used for each pyramid level.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: VoxelOpt's derivation is self-contained and evaluated on held-out pairs; the label-supervision caveat is a confound, not a circular step.

full rationale

The derivation chain is self-contained with respect to the paper's claims. VoxelOpt has no trainable parameters; the entropy field E(x) is computed from the cost volume P^1 via Eq. (4), and sigma(x) is a fixed logarithmic mapping (Eq. (6), alpha=1.5). The adaptive message passing filters C^k with this sigma, producing a new cost volume; the output is not equal to the input, and no quantity in the method is defined in terms of the final Dice or HD95 evaluation. Accuracy is measured on 42 test pairs from dataset [28] that are disjoint from the 380 training pairs used by the compared learning baselines, and VoxelOpt itself is not trained on any pairs, so there is no fitted parameter renamed as a prediction. Self-citations ([8], [30], [31], [32], [33]) are used as related-work motivation for segmentation-derived features and spatially adaptive filtering, but the paper's own Fig. 1 and Table 2 ablate these choices; no load-bearing claim rests solely on a self-citation. The closest self-referential design element is that entropy is derived from the same cost volume that it later filters; this is a confidence-guided smoothing heuristic, not a definitional equivalence. Two non-circular caveats are worth flagging: the conclusion's phrase 'without relying on label supervision or complex training' is overstated because the feature extractor [19] is a label-supervised foundation model and the paper does not analyze pretraining/evaluation overlap with [28]; and the alpha=1.5 entropy-to-smoothing mapping in Eq. (6) is a heuristic with no sensitivity analysis. These are correctness and generalization risks, not instances of the derivation reducing to its own inputs.

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

The method introduces no new physical entities. Its free parameters are all optimization hyperparameters (alpha, beta, theta schedule, pyramid settings), and its axioms are the small-displacement assumption, the Gaussian approximation of the regularizer, and the two design hypotheses about entropy and pretrained features. The central claim depends most heavily on the entropy-to-sigma mapping being broadly valid.

free parameters (4)
  • alpha (blurring strength cap) = 1.5
    Hand-set in Eq. (6); controls how strongly entropy modulates the Gaussian blur. No sensitivity analysis is provided.
  • beta (softmax temperature) = unspecified
    Used in Eq. (4) to convert cost volume into probabilities; value is never given in the paper, leaving the entropy computation under-determined.
  • theta schedule = {150, 50, 15, 5, 1.5, 0.5}
    Decaying interaction weights for the 6 iterations of coordinate descent, taken from prior work without adaptation to this dataset.
  • pyramid levels and integration steps = 5 levels, 7 scaling-and-squaring steps
    Architectural choices set by hand; no ablation on the number of levels is reported.
assumptions (4)
  • domain assumption At each pyramid level, the largest residual displacement is within one voxel, so a 27-neighbor cost volume is sufficient.
    Stated in section 2.2 and used to design the cost volume at every level. If residual displacements exceed one voxel, the search space is insufficient.
  • domain assumption The u-subproblem solution can be approximated by Gaussian filtering of the v field.
    Used in section 2.4 to replace the fixed-point iteration for the total-variation regularizer. This is a standard approximation but introduces a modeling error not quantified.
  • ad hoc to paper Displacement entropy computed from the local cost volume measures displacement signal strength.
    Core hypothesis of the paper, introduced in section 2.2. No proof or independent validation beyond the ablation in Table 2.
  • domain assumption The pretrained foundational segmentation model [19] provides semantically useful features for registration without further training.
    Adopted in section 2.3 based on qualitative entropy plots (Fig. 1). The paper does not analyze whether the model's pretraining data includes the evaluation dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VoxelOpt: Voxel-Adaptive Message Passing for Discrete Optimization in Deformable Abdominal CT Registration." pith.science (2026). https://pith.science/paper/ALDEFYHY

@misc{pith2026250619975,
  author       = {Pith},
  title        = {Pith review of: VoxelOpt: Voxel-Adaptive Message Passing for Discrete Optimization in Deformable Abdominal CT Registration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ALDEFYHY}},
  note         = {Machine review of arXiv:2506.19975}
}
read the original abstract

Recent developments in neural networks have improved deformable image registration (DIR) by amortizing iterative optimization, enabling fast and accurate DIR results. However, learning-based methods often face challenges with limited training data, large deformations, and tend to underperform compared to iterative approaches when label supervision is unavailable. While iterative methods can achieve higher accuracy in such scenarios, they are considerably slower than learning-based methods. To address these limitations, we propose VoxelOpt, a discrete optimization-based DIR framework that combines the strengths of learning-based and iterative methods to achieve a better balance between registration accuracy and runtime. VoxelOpt uses displacement entropy from local cost volumes to measure displacement signal strength at each voxel, which differs from earlier approaches in three key aspects. First, it introduces voxel-wise adaptive message passing, where voxels with lower entropy receives less influence from their neighbors. Second, it employs a multi-level image pyramid with 27-neighbor cost volumes at each level, avoiding exponential complexity growth. Third, it replaces hand-crafted features or contrastive learning with a pretrained foundational segmentation model for feature extraction. In abdominal CT registration, these changes allow VoxelOpt to outperform leading iterative in both efficiency and accuracy, while matching state-of-the-art learning-based methods trained with label supervision. The source code will be available at https://github.com/tinymilky/VoxelOpt

Figures

Figures reproduced from arXiv: 2506.19975 by the authors.

Figure 1
Figure 1. Visual examples of how features affect entropy distribution. The fixed image is shown in the first column; the moving image is manually shifted by one voxel. Cost volumes are computed using a kernel size of 1, and per-voxel entropies are derived via Eq. (4), then normalized via Eq. (6). voxels. Specifically, voxels with stronger signals should prevent their information from being diluted, while those with weak signa… view at source ↗
Figure 2
Figure 2. Schematic of the VoxelOpt framework. As shown in [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 22 canonical work pages

  1. [19]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Liu, J., Zhang, Y., Chen, J.N., Xiao, J., Lu, Y., A Landman, B., Yuan, Y., Yuille, A., Tang, Y., Zhou, Z.: Clip-driven universal model for organ segmentation and tumor detection. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 21152–21164 (2023)

  2. [28]

    IEEE Transactions on Biomedical Engineering 63(8), 1563–1572 (2016)

    Xu, Z., Lee, C.P., Heinrich, M.P., Modat, M., Rueckert, D., Ourselin, S., Abram- son, R.G., Landman, B.A.: Evaluation of six registration methods for the human abdomen on clinically acquired ct. IEEE Transactions on Biomedical Engineering 63(8), 1563–1572 (2016)

  3. [1]

    IEEE transactions on medical imaging38(2), 394–405 (2018)

    Aggarwal, H.K., Mani, M.P., Jacob, M.: Modl: Model-based deep learning architec- ture for inverse problems. IEEE transactions on medical imaging38(2), 394–405 (2018)

  4. [2]

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

    Arsigny, V., Commowick, O., Pennec, X., Ayache, N.: A log-euclidean framework for statistics on diffeomorphisms. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 924–931. Springer (2006)

  5. [3]

    Neuroimage 38(1), 95–113 (2007)

    Ashburner, J.: A fast diffeomorphic image registration algorithm. Neuroimage 38(1), 95–113 (2007)

  6. [4]

    Medical image analysis12(1), 26–41 (2008)

    Avants, B.B., Epstein, C.L., Grossman, M., Gee, J.C.: Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Medical image analysis12(1), 26–41 (2008)

  7. [5]

    IEEE transactions on medical imaging38(8), 1788–1800 (2019)

    Balakrishnan, G., Zhao, A., Sabuncu, M.R., Guttag, J., Dalca, A.V.: Voxelmorph: a learning framework for deformable medical image registration. IEEE transactions on medical imaging38(8), 1788–1800 (2019)

  8. [6]

    International journal of computer vision61, 139–157 (2005)

    Beg, M.F., Miller, M.I., Trouv´ e, A., Younes, L.: Computing large deformation metric mappings via geodesic flows of diffeomorphisms. International journal of computer vision61, 139–157 (2005)

Show all 34 references
  1. [7]

    Journal of Mathematical imaging and vision20, 89–97 (2004)

    Chambolle, A.: An algorithm for total variation minimization and applications. Journal of Mathematical imaging and vision20, 89–97 (2004)

  2. [8]

    IEEE Transactions on Neural Networks and Learning Systems pp

    Chen, X., Liu, M., Wang, R., Hu, R., Liu, D., Li, G., Zhang, H.: Spatially covariant image registration with text prompts. IEEE Transactions on Neural Networks and Learning Systems pp. 1–11 (2024)

  3. [9]

    Machine Vision and Applications31, 1–18 (2020)

    Haskins, G., Kruger, U., Yan, P.: Deep learning in medical image registration: a survey. Machine Vision and Applications31, 1–18 (2020)

  4. [10]

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

    Heinrich, M.P.: Closing the gap between deep and conventional image registra- tion using probabilistic dense displacement networks. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 50–58. Springer (2019)

  5. [11]

    Medical image analysis16(7), 1423–1435 (2012)

    Heinrich, M.P., Jenkinson, M., Bhushan, M., Matin, T., Gleeson, F.V., Brady, M., Schnabel, J.A.: Mind: Modality independent neighbourhood descriptor for multi- modal deformable registration. Medical image analysis16(7), 1423–1435 (2012)

  6. [12]

    IEEE transactions on medical imaging32(7), 1239–1248 (2013)

    Heinrich, M.P., Jenkinson, M., Brady, M., Schnabel, J.A.: Mrf-based deformable registration and ventilation estimation of lung ct. IEEE transactions on medical imaging32(7), 1239–1248 (2013)

  7. [13]

    In: Biomedical Image Registration: 6th In- ternational Workshop, WBIR 2014, London, UK, July 7-8, 2014

    Heinrich, M.P., Papie˙ z, B.W., Schnabel, J.A., Handels, H.: Non-parametric discrete registration with convex optimisation. In: Biomedical Image Registration: 6th In- ternational Workshop, WBIR 2014, London, UK, July 7-8, 2014. Proceedings 6. pp. 51–61. Springer (2014)

  8. [14]

    Artificial intelligence17(1-3), 185–203 (1981)

    Horn, B.K., Schunck, B.G.: Determining optical flow. Artificial intelligence17(1-3), 185–203 (1981)

  9. [15]

    In: International Conference on Learning Representations (2024)

    Hu, J., Gan, W., Sun, Z., An, H., Kamilov, U.: A plug-and-play image registration network. In: International Conference on Learning Representations (2024)

  10. [16]

    Jena, R., Sethi, D., Chaudhari, P., Gee, J.: Deep learning in medical image reg- istration: Magic or mirage? In: The Thirty-eighth Annual Conference on Neural Information Processing Systems (2024)

  11. [17]

    In: Proceedings of the AAAI Conference on Artificial Intelligence (2023) Title Suppressed Due to Excessive Length 11

    Jia, X., Bartlett, J., Chen, W., Song, S., Zhang, T., Cheng, X., Lu, W., Qiu, Z., Duan, J.: Fourier-net: Fast image registration with band-limited deformation. In: Proceedings of the AAAI Conference on Artificial Intelligence (2023) Title Suppressed Due to Excessive Length 11

  12. [18]

    Advances in neural information processing systems24(2011)

    Kr¨ ahenb¨ uhl, P., Koltun, V.: Efficient inference in fully connected crfs with gaussian edge potentials. Advances in neural information processing systems24(2011)

  13. [20]

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

    Meng, M., Feng, D., Bi, L., Kim, J.: Correlation-aware coarse-to-fine mlps for de- formable medical image registration. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 9645–9654 (2024)

  14. [21]

    IEEE Transactions on Medical Imaging (2024)

    Siebert, H., Großbr¨ ohmer, C., Hansen, L., Heinrich, M.P.: Convexadam: Self- configuring dual-optimisation-based 3d multitask medical image registration. IEEE Transactions on Medical Imaging (2024)

  15. [22]

    IEEE transactions on medical imaging32(7), 1153–1190 (2013)

    Sotiras, A., Davatzikos, C., Paragios, N.: Deformable medical image registration: A survey. IEEE transactions on medical imaging32(7), 1153–1190 (2013)

  16. [23]

    In: 2009 IEEE 12th International Conference on Computer Vision

    Steinbr¨ ucker, F., Pock, T., Cremers, D.: Large displacement optical flow computa- tion without warping. In: 2009 IEEE 12th International Conference on Computer Vision. pp. 1609–1614. IEEE (2009)

  17. [24]

    IEEE Transactions on Medical Imaging (2023)

    Van Harten, L.D., Stoker, J., Iˇ sgum, I.: Robust deformable image registration using cycle-consistent implicit representations. IEEE Transactions on Medical Imaging (2023)

  18. [25]

    Viergever, M.A., Maintz, J.A., Klein, S., Murphy, K., Staring, M., Pluim, J.P.: A survey of medical image registration–under review (2016)

  19. [26]

    IEEE Transactions on Medical Imaging (2024)

    Wang, H., Ni, D., Wang, Y.: Recursive deformable pyramid network for unsuper- vised medical image registration. IEEE Transactions on Medical Imaging (2024)

  20. [27]

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

    Wu, Y., Jiahao, T.Z., Wang, J., Yushkevich, P.A., Hsieh, M.A., Gee, J.C.: Nodeo: A neural ordinary differential equation based optimization framework for deformable image registration. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 2...

  21. [29]

    In: Pattern Recognition: 29th DAGM Symposium, Heidelberg, Germany, September 12-14, 2007

    Zach, C., Pock, T., Bischof, H.: A duality based approach for realtime tv-l 1 opti- cal flow. In: Pattern Recognition: 29th DAGM Symposium, Heidelberg, Germany, September 12-14, 2007. Proceedings 29. pp. 214–223. Springer (2007)

  22. [30]

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

    Zhang, H., Chen, X., Hu, R., Liu, D., Li, G., Wang, R.: Memwarp: Discontinuity- preserving cardiac registration with memorized anatomical filters. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 671–681. Springer (2024)

  23. [31]

    arXiv preprint arXiv:2506.10813 (2025)

    Zhang, H., Chen, X., Hu, R., Wang, R., Zhang, J., Liu, M., Wang, Y., Li, G., Cheng, X., Duan, J.: Unsupervised deformable image registration with structural nonparametric smoothing. arXiv preprint arXiv:2506.10813 (2025)

  24. [32]

    arXiv preprint arXiv:2401.09833 (2024)

    Zhang, H., Chen, X., Wang, R., Hu, R., Liu, D., Li, G.: Slicer networks. arXiv preprint arXiv:2401.09833 (2024)

  25. [33]

    In: Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence

    Zhang, H., Wang, R., Zhang, J., Liu, D., Li, C., Li, J.: Spatially covariant lesion segmentation. In: Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence. pp. 1713–1721 (2023)

  26. [34]

    NeuroImage268, 119886 (2023)

    Zhang, J., Spincemaille, P., Zhang, H., Nguyen, T.D., Li, C., Li, J., Kovanlikaya, I., Sabuncu, M.R., Wang, Y.: Laro: Learned acquisition and reconstruction optimiza- tion to accelerate quantitative susceptibility mapping. NeuroImage268, 119886 (2023)

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.