Pith. sign in

REVIEW 5 major objections 5 minor 32 references

Boundless Across Domains: A New Paradigm of Adaptive Feature and Cross-Attention for Domain Generalization in Medical Image Segmentation

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Adding AFB style mixing and DCAR cross-channel attention to a segmentation U-Net improves leave-one-domain-out Dice and surface distance on the Fundus and Prostate benchmarks.

desk verdict The average gains on two medical DG benchmarks look real, but DCAR is a black box: Eq. (4) is dimensionally wrong and the consistency loss is undefined, so the core second contribution cannot be evaluated. read the letter →

arxiv 2411.14883 v1 pith:TAV6ED4V submitted 2024-11-22 cs.CV cs.AIcs.LG

classification cs.CVcs.AIcs.LG
keywords domaingeneralizationmedicalimagesegmentationAdaptiveFeatureBlendingDualCross-AttentionRegularizationstyleaugmentationcross-channelattentionstatisticsleave-one-domain-outevaluation
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

The paper's goal is to make medical image segmentation models generalize to unseen scanners and hospitals without retraining. It proposes two add-ons to a U-shaped segmentation network: Adaptive Feature Blending (AFB), a style augmentation that replaces convex-combination statistics with random statistics sampled from a uniform distribution, and Dual Cross-Attention Regularization (DCAR), a cross-channel attention module that reconstructs original and style-perturbed deep features from each other and enforces consistency between them. The authors claim that this channel-level alignment produces domain-invariant representations and, in leave-one-domain-out tests on the Fundus and Prostate benchmarks, gives higher average Dice and lower average surface distance than six existing methods. If true, the contribution is a training-time regularizer that needs no adversarial training and extends the range of styles the model sees.

What carries the argument

The load-bearing machinery is two steps. Adaptive Feature Blending (AFB) perturbs the per-channel mean and standard deviation of encoder features using statistics sampled from $U(0,1)$ and mixes them with the original statistics through Bernoulli-selected masks, which is what allows augmented features to fall outside the source-domain distribution. Dual Cross-Attention Regularization (DCAR) applies cross-channel attention in both directions, with each of the original and generated feature sets serving in turn as queries against the other as keys and values, followed by channel self-attention. A mean-squared-error consistency loss then treats the two feature sets as having the same semantics. This explicit consistency constraint is the mechanism claimed to align channel responses and produce the reported generalization gains.

What would settle it

Measure the trained model's Dice on AFB-augmented source images as the Bernoulli mixing probability $\lambda$ varies; if Dice collapses as the random statistics fully replace the original channel statistics, the semantic-preservation assumption fails and the consistency loss would be aligning features to random content.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that random perturbation of feature-channel statistics, combined with a dual cross-channel attention reconstruction, makes the encoder's same-channel responses similar across domains. The authors hypothesize that an ideal generalized representation should exhibit similar pattern responses within the same channel for cross-domain images. AFB generates augmented features by drawing $\mu',\sigma' \sim U(0,1)$, selecting channel mixes with Bernoulli masks driven by $Beta(0.1,0.1)$, and replacing the original channel statistics with the mixed ones. DCAR then uses the original deep features as queries and the generated features as keys and values, repeats the construction in the opposite direction, applies channel self-attention, and trains with an MSE consistency term alongside Dice and cross-entropy losses. The reported results beat all six compared methods on both benchmarks, with Fundus average Dice/ASD of 89.44/9.18 and Prostate average Dice/ASD of 89.61/1.21.

Load-bearing premise

The method assumes that randomly perturbing a channel's mean and standard deviation, sometimes replacing them entirely with uniform random values, leaves the segmentation meaning of the image intact, so the original and AFB-augmented features should match.

Editorial extensions

If this is right

  • With AFB and DCAR combined, the paper reports leave-one-domain-out average Dice/ASD of 89.44/9.18 on Fundus and 89.61/1.21 on Prostate, beating all six comparison methods.
  • The Fundus ablation attributes +3.22 average Dice to AFB and another +2.32 to DCAR, so each component is claimed to contribute independently.
  • Because AFB samples statistics outside the source distribution, it is claimed to expand the training domain beyond the convex hull of source styles, which should help when the target domain lies outside that hull.
  • DCAR is a training-time regularizer, so it can be added to a U-shaped segmentation network without changing the architecture or adding adversarial training.

Reading between the lines

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

  • The paper leaves open whether semantic preservation holds when the Bernoulli mask replaces whole channels, so the method's success may depend on how much of the task-relevant signal lives in low-frequency channel statistics rather than in the discarded parts of each channel.
  • The reported gain over the strongest baseline is larger on Prostate (+0.69 average Dice) than on Fundus (+0.24), suggesting the benefit may scale with the amount of style shift, a relationship the paper does not test.
  • A natural extension is to apply the same cross-channel reconstruction to multiple source domains simultaneously, or to intermediate encoder layers, which would make the consistency constraint denser without changing the loss.
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

5 major / 5 minor

Summary. This paper proposes two components for domain generalization in medical image segmentation: Adaptive Feature Blending (AFB), a style-augmentation technique that mixes original per-channel feature statistics with statistics sampled from U(0,1) under a Bernoulli mask, and Dual Cross-Attention Regularization (DCAR), a channel-attention module intended to reconstruct original and generated deep features under a semantic consistency constraint. The combined method is evaluated in leave-one-domain-out fashion on the Fundus and Prostate benchmarks, comparing Dice and ASD against six prior methods, with an ablation study on Fundus. The central performance claim is that the combined method achieves superior average Dice and ASD over prior state-of-the-art methods on both benchmarks.

Significance. If the reported results are reproducible, the paper would offer a simple and computationally light augmentation-plus-regularizer recipe that improves Dice and ASD over strong baselines on two standard medical-image domain-generalization benchmarks. The AFB idea of sampling augmented statistics outside the convex hull of source-domain styles is a reasonable departure from MixStyle-style augmentation, and a channel-attention consistency regularizer is a plausible direction. However, the significance is currently limited by missing definitions, an invalid equation, and the absence of variance estimates, so the empirical gains cannot yet be causally attributed to the proposed mechanisms. The paper does not ship code, machine-checked proofs, or other reproducibility artifacts; the experimental protocol is standard but the method is underspecified.

major comments (5)
  1. [III-B, Eq. (4)] As written, Eq. (4) is dimensionally invalid. With f-bar, f-bar_AFB in R^{HW x C} and w_q, w_k, w_v in R^{C x 2C}, the projections q, k, v are in R^{HW x 2C}; then q^T k is in R^{2C x 2C}, and v^T w_out cannot be formed with w_out in R^{2C x C} because v^T has HW columns while w_out has 2C rows. No bracketing of (q^T k) v^T w_out yields a tensor in R^{HW x C}. Consequently the reconstructed feature f-hat and the DCAR mechanism are not defined by the manuscript, and the Table III gain of 2.32 from DCAR cannot be attributed to the stated cross-channel attention mechanism.
  2. [III-C, Eq. (5)] The consistency loss L_consist in Eq. (5) is never defined. The text states only that a mean squared error is used as a semantic consistency loss, without specifying which tensors are compared (e.g., original versus generated features, reconstructed versus original features, or decoder outputs) or at which layers and attention heads. Because the DCAR contribution in Table III is computed with this unspecified loss, any MSE penalty could in principle produce the effect; the experiment does not isolate the proposed dual cross-attention mechanism as written.
  3. [III-A, Eqs. (1)-(2)] AFB is incompletely specified. The text samples b ~ Beta(alpha, alpha) and says b is used as the probability to generate a Bernoulli distribution, but then defines lambda ~ Bern(P) without ever giving P or stating that P = b. In addition, sampling mu' and sigma' from U(0,1) is not justified: post-ReLU activation means are nonnegative and can exceed 1, so U(0,1) does not obviously encompass most feature statistics, and Eq. (2) lacks an epsilon in the normalization denominator, which is numerically problematic for near-constant channels. These ambiguities prevent reproduction of the AFB augmentation and of the 3.22-point ablation gain in Table III.
  4. [III-B, semantic-preservation assumption] The DCAR consistency constraint rests on the assumption that AFB changes only style and preserves semantics. This is not guaranteed: because lambda is a per-channel Bernoulli mask, a channel can have its mean and variance fully replaced by random values, and with many channels a substantial fraction of the feature map can be replaced simultaneously. The paper provides no test of semantic preservation, such as measuring whether AFB-transformed features still support the original segmentation or whether the MSE loss penalizes task-relevant domain-specific features. This is a load-bearing assumption for interpreting the reported gains and should be validated or relaxed.
  5. [Tables I-III] All reported results are averages over three runs with no standard deviations or per-run values. For a claim of superior performance over prior methods, differences of roughly 0.2-0.7 Dice (e.g., WT-PSE 89.20 vs Ours 89.44 on Fundus, and WT-PSE 88.92 vs Ours 89.61 on Prostate) need to be assessed against run-to-run variability; without variance information, the headline improvements are not statistically grounded and the ablation increments in Table III are not interpretable as reliable effects.
minor comments (5)
  1. [Abstract and Introduction] The phrase 'display constraint' appears repeatedly; it should be 'explicit constraint' or 'explicit regularization constraint'.
  2. [IV-B, Table I] The text says the method 'consistently achieves higher Dice coefficients,' but in Table I the Domain 3 optic-cup Dice of 86.46 is lower than WT-PSE's 87.90; the claim should be qualified to averages or to the majority of domains.
  3. [Fig. 1] Figure 1 is difficult to interpret because it has no axis labels and does not explain how lambda and 1-lambda relate to the plotted statistics; adding axes and a caption description would help.
  4. [III-B] The notation 'psi denote the instance normalization' is grammatically unclear, and psi is not otherwise defined; please define the normalization operator explicitly.
  5. [IV-A] The paper should state the input resolution, the exact U-Net/ResNet-34 configuration, and how the hyperparameters alpha and the consistency-loss weight are chosen; the current dependence on 'previous studies [3]' is insufficient for reproduction.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: AFB/DCAR are training objectives validated on external benchmarks; undefined L_consist and Eq. (4) dimensional issues are rigor flaws, not circular reasoning.

full rationale

The paper's central claim is empirical: AFB and DCAR improve Dice/ASD on held-out domains compared with external baselines. AFB is a stochastic augmentation defined by Eqs. (1)-(2), sampling statistics from U(0,1) and mixing them with original channel statistics; DCAR is a consistency regularization that compares original and augmented deep features. Neither component is a fitted parameter renamed as a prediction, and no equation reduces to its own input by construction. The 'ideal generalized representation' hypothesis motivates the design of the loss, but the reported results are measured on unseen target domains and are not derived from that hypothesis. One reference ([10]) shares a co-author name with the first author, but it is cited only as general related work on domain-invariant representation learning and is not load-bearing. The manuscript does contain serious rigor gaps: L_consist in Eq. (5) is never explicitly defined, and Eq. (4) appears dimensionally inconsistent under the declared shapes (with f-bar, f-bar_AFB in R^{HW x C} and w_q, w_k, w_v in R^{C x 2C}, the projections are R^{HW x 2C}, so q^T k and the subsequent products do not conform). These are correctness and reproducibility concerns that prevent attributing the ablation gains to the stated mechanism, but they are not circularity: an undefined or uncomputable loss is an omission, not a case where the output is equal to the input by definition or by self-citation. Accordingly, no circular step is established and the score is 0.

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

The central claim rests on several unverified modeling assumptions, chiefly semantic preservation under random-statistic augmentation and the representational power of channel statistics. AFB introduces hand-chosen distribution parameters (alpha and U(0,1)) and an unspecified Bernoulli mixing probability. No new physical entities are introduced.

free parameters (4)
  • Beta concentration parameter alpha = 0.1
    Controls the Beta(alpha,alpha) distribution that generates the mixing coefficients in AFB. Copied from MixStyle [4] as an empirical setting, not tuned on the target task. The paper depends on this value for the augmentation distribution.
  • Bernoulli mixing probability P = unspecified
    The paper defines b~Beta(alpha,alpha) as the probability for a Bernoulli draw, then writes lambda~Bern(P) without defining P. This choice determines whether feature channels are fully replaced or continuously blended, and it is central to AFB's behavior.
  • Support of augmentation statistics U(0,1) = range [0,1]
    The augmented feature statistics mu' and sigma' are sampled from U(0,1). The paper asserts, without measurement, that this range encompasses most feature statistics. The choice determines how far AFB moves outside the training distribution and is central to the claim of generating out-of-distribution samples.
  • Consistency loss weight = 1.0 (implicit)
    Eq. (5) sums L_seg and L_consist without a weighting coefficient, so the relative weight is implicitly 1. This is a hand-chosen balance that affects the DCAR strength; no ablation over the weight is reported.
assumptions (4)
  • domain assumption An ideal generalized representation should exhibit similar pattern responses within the same channel across cross-domain images.
    States the motivation for DCAR in Section III-B. It is not derived or tested directly; the entire cross-attention regularization is built on this premise.
  • domain assumption AFB-generated images are merely style-transformed versions of the originals, with semantics unchanged.
    Section III-B states this. If false, the consistency loss between original and generated features could harm segmentation. This is the weakest assumption.
  • domain assumption Channel mean and standard deviation of CNN features capture visual style and domain information.
    Invoked in Section III-A based on prior work [20]-[22]. The AFB method operates on these statistics; if these statistics do not capture domain style, the augmentation is not meaningful.
  • ad hoc to paper U(0,1) encompasses most feature statistics of medical image encoders.
    Section III-A states this without empirical evidence. The paper's Fig. 1 is a schematic illustration, not measured feature distributions. This assumption is needed for AFB to produce in-distribution and out-of-distribution samples.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Boundless Across Domains: A New Paradigm of Adaptive Feature and Cross-Attention for Domain Generalization in Medical Image Segmentation." pith.science (2026). https://pith.science/paper/TAV6ED4V

@misc{pith2026241114883,
  author       = {Pith},
  title        = {Pith review of: Boundless Across Domains: A New Paradigm of Adaptive Feature and Cross-Attention for Domain Generalization in Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TAV6ED4V}},
  note         = {Machine review of arXiv:2411.14883}
}
read the original abstract

Domain-invariant representation learning is a powerful method for domain generalization. Previous approaches face challenges such as high computational demands, training instability, and limited effectiveness with high-dimensional data, potentially leading to the loss of valuable features. To address these issues, we hypothesize that an ideal generalized representation should exhibit similar pattern responses within the same channel across cross-domain images. Based on this hypothesis, we use deep features from the source domain as queries, and deep features from the generated domain as keys and values. Through a cross-channel attention mechanism, the original deep features are reconstructed into robust regularization representations, forming an explicit constraint that guides the model to learn domain-invariant representations. Additionally, style augmentation is another common method. However, existing methods typically generate new styles through convex combinations of source domains, which limits the diversity of training samples by confining the generated styles to the original distribution. To overcome this limitation, we propose an Adaptive Feature Blending (AFB) method that generates out-of-distribution samples while exploring the in-distribution space, significantly expanding the domain range. Extensive experimental results demonstrate that our proposed methods achieve superior performance on two standard domain generalization benchmarks for medical image segmentation.

Figures

Figures reproduced from arXiv: 2411.14883 by the authors.

Figure 1
Figure 1. Visualization of synthetic feature statistics samples using MixStyle, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Visual comparison for Fundus and Prostate segmentation task. The [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 12 canonical work pages

  1. [1]

    Shape-aware meta-learning for gener- alizing prostate mri segmentation to unseen domains,

    Q. Liu, Q. Dou, and P.-A. Heng, “Shape-aware meta-learning for gener- alizing prostate mri segmentation to unseen domains,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part II 23. Springer, 2020, pp. 475–485

  2. [2]

    Dofe: Domain-oriented feature embedding for generalizable fundus image seg- mentation on unseen datasets,

    S. Wang, L. Yu, K. Li, X. Yang, C.-W. Fu, and P.-A. Heng, “Dofe: Domain-oriented feature embedding for generalizable fundus image seg- mentation on unseen datasets,” IEEE Transactions on Medical Imaging , vol. 39, no. 12, pp. 4237–4248, 2020

  3. [3]

    Generalizable medical image segmentation via random amplitude mixup and domain-specific image restoration,

    Z. Zhou, L. Qi, and Y . Shi, “Generalizable medical image segmentation via random amplitude mixup and domain-specific image restoration,” in European Conference on Computer Vision . Springer, 2022, pp. 420– 436

  4. [4]

    Domain generalization with mixstyle,

    K. Zhou, Y . Yang, Y . Qiao, and T. Xiang, “Domain generalization with mixstyle,” arXiv preprint arXiv:2104.02008 , 2021

  5. [5]

    Fda: Fourier domain adaptation for semantic segmentation,

    Y . Yang and S. Soatto, “Fda: Fourier domain adaptation for semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 4085–4095

  6. [6]

    Adversarial style augmen- tation for domain generalized urban-scene segmentation,

    Z. Zhong, Y . Zhao, G. H. Lee, and N. Sebe, “Adversarial style augmen- tation for domain generalized urban-scene segmentation,” Advances in neural information processing systems , vol. 35, pp. 338–350, 2022

  7. [7]

    Adversarial style aug- mentation for domain generalization,

    Y . Zhang, B. Deng, R. Li, K. Jia, and L. Zhang, “Adversarial style aug- mentation for domain generalization,” arXiv preprint arXiv:2301.12643, 2023

  8. [8]

    Global and local tex- ture randomization for synthetic-to-real semantic segmentation,

    D. Peng, Y . Lei, L. Liu, P. Zhang, and J. Liu, “Global and local tex- ture randomization for synthetic-to-real semantic segmentation,” IEEE Transactions on Image Processing , vol. 30, pp. 6594–6608, 2021

Show all 32 references
  1. [9]

    Domain invariant representation learning with domain density transformations,

    A. T. Nguyen, T. Tran, Y . Gal, and A. G. Baydin, “Domain invariant representation learning with domain density transformations,” Advances in Neural Information Processing Systems , vol. 34, pp. 5264–5275, 2021

  2. [10]

    Knowledge distillation-based domain-invariant representation learning for domain generalization,

    Z. Niu, J. Yuan, X. Ma, Y . Xu, J. Liu, Y .-W. Chen, R. Tong, and L. Lin, “Knowledge distillation-based domain-invariant representation learning for domain generalization,” IEEE Transactions on Multimedia , 2023

  3. [11]

    Learning to balance specificity and invariance for in and out of domain generalization,

    P. Chattopadhyay, Y . Balaji, and J. Hoffman, “Learning to balance specificity and invariance for in and out of domain generalization,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16 . Springer, 2020, pp. 301–318

  4. [12]

    Treasure in distribution: a domain randomization based multi-source domain generalization for 2d medical image segmentation,

    Z. Chen, Y . Pan, Y . Ye, H. Cui, and Y . Xia, “Treasure in distribution: a domain randomization based multi-source domain generalization for 2d medical image segmentation,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 20...

  5. [13]

    Robust and generalizable visual representation learning via random convolutions,

    Z. Xu, D. Liu, J. Yang, C. Raffel, and M. Niethammer, “Robust and generalizable visual representation learning via random convolutions,” arXiv preprint arXiv:2007.13003 , 2020

  6. [14]

    Domain generalization with correlated style uncertainty,

    Z. Zhang, B. Wang, D. Jha, U. Demir, and U. Bagci, “Domain generalization with correlated style uncertainty,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2024, pp. 2000–2009

  7. [15]

    Localized adversarial domain generalization,

    W. Zhu, L. Lu, J. Xiao, M. Han, J. Luo, and A. P. Harrison, “Localized adversarial domain generalization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 7108–7118

  8. [16]

    Domain adversarial neural net- works for domain generalization: When it works and how to improve,

    A. Sicilia, X. Zhao, and S. J. Hwang, “Domain adversarial neural net- works for domain generalization: When it works and how to improve,” Machine Learning, vol. 112, no. 7, pp. 2685–2721, 2023

  9. [17]

    Feddg: Federated do- main generalization on medical image segmentation via episodic learn- ing in continuous frequency space,

    Q. Liu, C. Chen, J. Qin, Q. Dou, and P.-A. Heng, “Feddg: Federated do- main generalization on medical image segmentation via episodic learn- ing in continuous frequency space,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 1013– 1023

  10. [18]

    Domain generalization with adversarial feature learning,

    H. Li, S. J. Pan, S. Wang, and A. C. Kot, “Domain generalization with adversarial feature learning,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 5400–5409

  11. [19]

    Robustnet: Improving domain generalization in urban-scene segmentation via in- stance selective whitening,

    S. Choi, S. Jung, H. Yun, J. T. Kim, S. Kim, and J. Choo, “Robustnet: Improving domain generalization in urban-scene segmentation via in- stance selective whitening,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 11 580–11 590

  12. [20]

    Instance normalization: The missing ingredient for fast stylization,

    D. Ulyanov, A. Vedaldi, and V . Lempitsky, “Instance normalization: The missing ingredient for fast stylization,”arXiv preprint arXiv:1607.08022, 2016

  13. [21]

    Arbitrary style transfer in real-time with adaptive instance normalization,

    X. Huang and S. Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” in Proceedings of the IEEE interna- tional conference on computer vision , 2017, pp. 1501–1510

  14. [22]

    A learned representation for artistic style,

    V . Dumoulin, J. Shlens, and M. Kudlur, “A learned representation for artistic style,” arXiv preprint arXiv:1610.07629 , 2016

  15. [23]

    Uctransnet: rethinking the skip connections in u-net from a channel-wise perspective with trans- former,

    H. Wang, P. Cao, J. Wang, and O. R. Zaiane, “Uctransnet: rethinking the skip connections in u-net from a channel-wise perspective with trans- former,” in Proceedings of the AAAI conference on artificial intelligence, vol. 36, no. 3, 2022, pp. 2441–2449

  16. [24]

    Davit: Dual attention vision transformers,

    M. Ding, B. Xiao, N. Codella, P. Luo, J. Wang, and L. Yuan, “Davit: Dual attention vision transformers,” in European conference on com- puter vision. Springer, 2022, pp. 74–92

  17. [25]

    V-net: Fully convolutional neural networks for volumetric medical image segmentation,

    F. Milletari, N. Navab, and S.-A. Ahmadi, “V-net: Fully convolutional neural networks for volumetric medical image segmentation,” in 2016 fourth international conference on 3D vision (3DV) . Ieee, 2016, pp. 565–571

  18. [26]

    K. P. Murphy, Machine learning: a probabilistic perspective . MIT press, 2012

  19. [27]

    Domain specific convolution and high frequency reconstruction based unsupervised domain adaptation for medical image segmentation,

    S. Hu, Z. Liao, and Y . Xia, “Domain specific convolution and high frequency reconstruction based unsupervised domain adaptation for medical image segmentation,” in International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2022, pp. 650–659

  20. [28]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

  21. [29]

    Hcdg: A hierarchical consistency framework for domain generalization on medical image segmentation,

    Y . Yang, S. Wang, L. Zhu, and L. Yu, “Hcdg: A hierarchical consistency framework for domain generalization on medical image segmentation,” arXiv preprint arXiv:2109.05742 , 2021

  22. [30]

    Tvconv: Efficient translation variant convolution for layout-aware visual process- ing,

    J. Chen, T. He, W. Zhuo, L. Ma, S. Ha, and S.-H. G. Chan, “Tvconv: Efficient translation variant convolution for layout-aware visual process- ing,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 12 548–12 558

  23. [31]

    Cddsa: Contrastive domain disentanglement and style augmentation for generalizable medical image segmentation,

    R. Gu, G. Wang, J. Lu, J. Zhang, W. Lei, Y . Chen, W. Liao, S. Zhang, K. Li, D. N. Metaxas et al., “Cddsa: Contrastive domain disentanglement and style augmentation for generalizable medical image segmentation,” Medical Image Analysis , vol. 89, p. 102904, 2023

  24. [32]

    Learning robust shape regularization for generalizable medical image segmentation,

    K. Chen, T. Qin, V . H.-F. Lee, H. Yan, and H. Li, “Learning robust shape regularization for generalizable medical image segmentation,” IEEE Transactions on Medical Imaging , 2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.