Pith. sign in

REVIEW 4 major objections 6 minor 42 references

Vision-Language Semantic Aggregation Leveraging Foundation Model for Generalizable Medical Image Segmentation

T0 review · 4 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper claims that clustering visual and textual features with an EM algorithm before fusing them at the deepest visual level substantially improves single-source domain generalization in medical image segmentation.

desk verdict Sensible combination of EM aggregation and text-guided decoding for medical SDG; small but consistent gains, but the unidentified text encoder undermines the stated mechanism until disclosed. read the letter →

arxiv 2509.08570 v1 pith:FXHJX5JW submitted 2025-09-10 cs.CV

Wenjun Yu , Yinchen Zhou , Jia-Xuan Jiang , Shubin Zeng , Yuee Li , Zhong Wang This is my paper · ORCID
classification cs.CV
keywords single-sourcedomaingeneralizationmedicalimagesegmentationvision-languagemodelsexpectation-maximizationaggregationsemantictext-guidedpixeldecodercardiacMRIfundusimaging
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 tries to establish that the poor performance of vision-language models on medical segmentation comes from feature dispersion and from a semantic gap between abstract text and fine-grained visual features, and that both can be addressed by semantic aggregation. It introduces an Expectation-Maximization (EM) Aggregation mechanism that clusters features from each modality into compact semantic centers, and a Text-Guided Pixel Decoder that fuses text queries with the deepest visual features. On cardiac MRI and fundus imaging benchmarks, the combined framework outperforms existing single-domain generalization methods, including the Rein baseline. A sympathetic reader would care because it offers a concrete architectural recipe—cluster first, then fuse deeply—for making segmentation models transfer to unseen scanners and imaging modalities.

What carries the argument

EM-Aggregation: an iterative expectation-maximization module that models features as a mixture of K latent semantic concepts, computes soft assignments in the E-step, updates prototypes in the M-step, and reconstructs refined features. It appears in two instances: Text-EMA clusters text embeddings into compact meta-semantic centers, and Vision-EMA clusters deep patch features to suppress domain-specific noise. This compaction is what enables the Text-Guided Pixel Decoder—a cross-modal attention module with text-derived queries and visual keys/values—to fuse modalities effectively at the deepest pyramid level f4.

What would settle it

Train the complete model twice, identical except for the initialization of the frozen text encoder ET: once with whatever pre-trained weights the authors used and once with random weights. If Dice on the unseen domains does not drop in the random-initialization run, the supposed domain-invariant textual semantics are not what drives the gain. A second control using semantically meaningless but visually matched text prompts would reveal whether the text stream contributes anything beyond the visual backbone.

Watch

Extended reading notes

Core claim

The paper's central claim is that deep, abstract visual features align with textual semantics far better than shallow features, and that EM-based clustering prepares both modalities for robust cross-modal fusion. Concretely, EM-Aggregation is applied to text embeddings and visual patch features: features are softly assigned to K latent prototypes, the prototypes are updated as weighted averages, and the features are reconstructed from those prototypes. This reduces feature variance and increases clustering tightness for both modalities. The Text-Guided Pixel Decoder then uses text embeddings as queries and the deepest visual features as keys and values in cross-modal attention, producing a t

Load-bearing premise

The paper assumes its frozen text encoder is a genuinely pre-trained semantic encoder, but it never says where that pre-training came from; if the encoder is random, the domain-invariant textual knowledge the method relies on does not exist.

Editorial extensions

If this is right

  • Fusing text guidance at the deepest visual feature level is more effective than at shallower levels, so future multimodal segmentation architectures should align language with the most abstract visual features rather than spreading guidance across all scales.
  • EM-Aggregation measurably compacts features: the paper reports about a 50% reduction in variance and roughly 30% improvement in clustering tightness for both textual and visual features.
  • Text guidance alone produces a large performance jump over the vision-only Rein baseline, and adding textual then visual EM-Aggregation yields further consistent gains, indicating the two mechanisms are complementary.
  • The framework generalizes across two anatomies and imaging modalities (cardiac MRI and fundus SLO/OCT), suggesting the approach is not tied to one organ or acquisition protocol.
  • The paper's stated future direction is adapting the framework to broader tasks such as tumor segmentation, with more adaptive fusion strategies.

Reading between the lines

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

  • An unstated dependency: the frozen text encoder ET is described only as a 3-layer Transformer with 256-dimensional embeddings, and the paper does not identify its pre-training source. If ET is randomly initialized, the claimed domain-invariant textual knowledge is vacuous, and the source of the reported gains would need a different explanation.
  • A testable extension is to apply EM-Aggregation before fusion in other multimodal tasks, such as vision-language question answering or report generation; if compactness is the true mechanism, it should improve cross-modal alignment there too.
  • The finding that shallow features are poor fusion partners for text suggests a clearer division of labor: shallow layers could be reserved for spatial and structural decoding, while text guidance acts only on deep semantics. That could simplify multi-scale fusion designs.
  • The reported variance and clustering-tightness metrics could be broken down per class and per target domain; if compaction predicts per-domain Dice, that would directly support the mechanistic story rather than just the aggregate numbers.
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 a single-source domain generalization (SDG) framework for medical image segmentation that combines a frozen DINOv2 visual backbone with a parallel text stream. Two main mechanisms are introduced: (i) an Expectation-Maximization (EM) Aggregation module that iteratively clusters visual and textual features around latent prototypes (Eqs. 2–4), and (ii) a Text-Guided Pixel Decoder that uses class-level text queries in cross-attention over deep visual features (Eqs. 6–8), with fusion performed only at the deepest pyramid level f4. Training uses a BCE+Dice segmentation loss plus a text regularization loss (Eqs. 9–11). Experiments are reported on cardiac MRI (ACDC-bSSFP as source, four unseen target domains) and fundus (FairDomain SLO↔OCT), where the method is compared with CNN- and ViT-based SDG baselines, along with ablations and t-SNE/PCA dispersion analyses.

Significance. If the results are reproducible, the paper offers a reasonable architectural recipe for injecting language guidance into medical image segmentation under single-source DG. The strengths are that evaluation uses strictly held-out target domains, the EM and cross-attention equations are stated plainly, and the full model shows consistent gains over the vision-only Rein baseline in Table 2. The two anatomical sites give some breadth. However, the central explanatory claim—that a frozen vision-language text encoder provides domain-invariant semantic knowledge—rests on an unverified assumption about the text encoder's pre-training. The paper does not identify the text encoder's initialization or pre-training source, so the mechanism may be no more than a learned text-conditioned query decoder. The numerical claims also lack error bars for the cardiac experiments. These issues are localizable and fixable, but they are load-bearing for the paper's stated contribution.

major comments (4)
  1. [§3.4.1, Eq. (5) and §4.2] The text encoder E_T is the stated source of “domain-invariant textual knowledge,” yet Section 4.2 describes it only as a 3-layer Transformer with a 256-dim word embedding, and no pre-training source, checkpoint, or initialization is given. Section 3.4.1 says the encoder is frozen to preserve pre-trained knowledge, but freezing a randomly initialized network does not preserve any semantics—it merely fixes a random projection. If E_T is random, the text queries in Eq. (5) are not domain-invariant semantic vectors, and the claimed foundation-model motivation is vacuous; the gains could be attributed entirely to the learnable prompts and the cross-attention decoder acting as a class-conditioned query mechanism. Please specify the initialization (e.g., CLIP, BiomedCLIP, or other VLM text encoder) and, ideally, report an ablation that replaces it with a randomly initialized encoder to show th
  2. [Tables 1 and 3, §4.3] The paper claims the method “consistently outperforms” SOTA methods, but no error bars, standard deviations, or significance tests are reported for the cardiac results in Table 1. Several reported gains are small (e.g., average LV Dice 0.7800 vs. tqdm's 0.7652; average RV 0.5351 vs. 0.4294 is larger, but the margin varies by domain). The only significance statement is a Wilcoxon test on the fundus set versus tqdm, and it is not reported for all comparisons. Please report results over multiple seeds with standard deviations and, where appropriate, paired significance tests; otherwise the claim of consistent superiority is not statistically supported.
  3. [Table 2, §4.4] The text states that adding Textual EM-Aggregation “yields another consistent performance boost,” but the table does not support “consistent.” Comparing row (2) and row (3), average Dice_Myo decreases from 0.6589 to 0.6582, average Dice_RV increases by only 0.0009, and several per-domain entries drop (e.g., Domain 1 LV and Myo, Domain 3 RV). Only average Dice_LV shows a clear improvement (0.7521 to 0.7629). This weakens the claimed synergy between the EM modules and text guidance. Please either temper the claim, provide per-seed paired comparisons, or analyze which components drive the improvements more carefully.
  4. [Figs. 8 and 9, §4.5] The dispersion analysis reports large reductions in “Variance” and “Clustering Tightness” (e.g., 50.76% and 30.00% for text features), but these metrics are never defined, and the figures do not state which features were used, from which layers, on which domains, or how the values were computed. As presented, these numbers cannot be verified, and their connection to segmentation generalization is asserted rather than demonstrated. Define the metrics and the feature sets, and ideally report the same statistics on source and target domains separately.
minor comments (6)
  1. [Eq. (10), §3.5.2] The regularization loss is said to use embeddings from the “trainable text encoder,” but §3.4.1 states that E_T is frozen. Clarify whether the trainable part is only the prompts p, the projection layers, or some other component.
  2. [§4.1] The description says “T2-weighted MRI images” for the combined dataset, but the listed domains include ACDC-bSSFP and MSCMR-LGE, which are not typically T2-weighted. Please clarify the imaging protocols.
  3. [Table 3] The backbone column is ambiguous: some CNN-based methods appear to be evaluated with a U-Net backbone and Transformer-based methods with ViT, but the table layout makes it unclear which backbone is used for each row. Please make the pairing explicit and consistent.
  4. [References] Reference [42] appears malformed (“P. Reviewed, S. Blitzer, J. Crammer...”). The citation should be cleaned up.
  5. [Typos] Minor typos include “embbeding” (Fig. 4 caption), “mutimodal” (§3.2), “foudation” (§3.4.1), and “Cross-Model” (Fig. 2 caption).
  6. [Figure 1] The caption says “on our samples” without specifying which dataset or domain. Since this figure motivates the choice of f4, specify the data source and evaluation split.

Circularity Check

0 steps flagged · score 2.0 of 10

No substantive circularity: held-out target-domain benchmarks support the central claims; remaining issues are missing implementation details and non-load-bearing self-citations.

full rationale

The paper's central generalization claims are evaluated on strictly unseen target domains (ACDC source ACDC-bSSFP with MSCMR/EMIDEC targets; FairDomain cross-modality SLO<->OCT), with hyperparameters selected via 3-fold cross-validation on the source domain only. The headline Dice improvements are therefore not fitted predictions or re-statements of training data. The EM-Aggregation module is a self-contained application of standard E/M updates (Eqs. 2-4) and does not encode target-domain information. The text-query generation (Eq. 5) is a frozen encoder with learnable prompts; the paper's assertion that this supplies 'domain-invariant textual knowledge' is an unverified premise because Section 4.2 only specifies a 3-layer Transformer with dimension 256 and gives no pre-training source or initialization. This is a support gap and a correctness/robustness concern, not a circular reduction: even a randomly initialized ET would still define the same equations, and the reported gains could partly come from the learnable prompts and cross-attention decoder. Self-citations [8,9,19] appear in related-work and motivation contexts but are not load-bearing for the method's architecture or results. The only mildly self-referential element is that the compactness improvements shown in Figs. 8-9 are largely a mathematical consequence of the EM reconstruction defined by Eqs. (3)-(4), so those visualizations are consistency checks rather than independent evidence; nevertheless, they are not used to inflate the held-out benchmark numbers. Overall, no derivation reduces to its own inputs by construction, so circularity is minimal.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The core contribution (EM aggregation and text-guided fusion) rests on standard clustering and cross-attention math, but several inputs are unspecified: the frozen text encoder's pre-training, the fixed anchors R in Eq. 10, and the hand-set hyperparameters (K, alpha, loss weights). The invented-entities list is empty because no new physical entities are introduced.

free parameters (6)
  • Number of text clusters K_text = 8
    Set in Section 4.2; controls granularity of text semantic aggregation.
  • Number of visual clusters K_vis = 6
    Set in Section 4.2.
  • EM iteration counts = text: 6, visual: 1
    Set in Section 4.2; uncertain whether one visual iteration is enough for convergence.
  • Strength factor alpha = 0.1
    Set in Section 4.2; used in the VisionEMAggregation module but not defined in the equations.
  • Loss weights lambda_bce, lambda_dice, lambda_reg = 1.0, 0.3, 0.05
    Empirically set in Section 3.5.3.
  • Temperature tau = not specified
    Appears in Eq. 2 and Eq. 10; value is not reported.
assumptions (4)
  • domain assumption Frozen text encoder ET provides domain-invariant semantics
    Section 3.4.1 claims ET is pre-trained and frozen, but no pre-training source or initialization is specified. If ET is random, semantic guidance is vacuous.
  • domain assumption Anchor embeddings R are pre-defined and fixed
    Eq. 10 uses R as fixed anchors; the source and construction of R are unspecified.
  • ad hoc to paper Fusion at deepest feature level f4 is optimal
    Based on the paper's own Fig. 1 experiments; if this choice is wrong, the deep alignment design fails.
  • standard math EM algorithm produces useful clusters with the chosen iteration counts
    EM convergence is standard, but one iteration for visual features is a heuristic adopted without justification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Vision-Language Semantic Aggregation Leveraging Foundation Model for Generalizable Medical Image Segmentation." pith.science (2026). https://pith.science/paper/FXHJX5JW

@misc{pith2026250908570,
  author       = {Pith},
  title        = {Pith review of: Vision-Language Semantic Aggregation Leveraging Foundation Model for Generalizable Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FXHJX5JW}},
  note         = {Machine review of arXiv:2509.08570}
}
read the original abstract

Multimodal models have achieved remarkable success in natural image segmentation, yet they often underperform when applied to the medical domain. Through extensive study, we attribute this performance gap to the challenges of multimodal fusion, primarily the significant semantic gap between abstract textual prompts and fine-grained medical visual features, as well as the resulting feature dispersion. To address these issues, we revisit the problem from the perspective of semantic aggregation. Specifically, we propose an Expectation-Maximization (EM) Aggregation mechanism and a Text-Guided Pixel Decoder. The former mitigates feature dispersion by dynamically clustering features into compact semantic centers to enhance cross-modal correspondence. The latter is designed to bridge the semantic gap by leveraging domain-invariant textual knowledge to effectively guide deep visual representations. The synergy between these two mechanisms significantly improves the model's generalization ability. Extensive experiments on public cardiac and fundus datasets demonstrate that our method consistently outperforms existing SOTA approaches across multiple domain generalization benchmarks.

Figures

Figures reproduced from arXiv: 2509.08570 by the authors.

Figure 1
Figure 1. The bar chart compares the Dice Score on our samples for each cardiac structure [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of our proposed text-guided framework with EM. The [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. The overall architecture of visual encoder. The framework consists of a progress [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The overall architecture of textual encoder. The framework consists of a tok [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: The overall architecture of our text-guided pixel decoder. The framework con [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of segmentation results on challenging unseen samples from [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Visual comparison of segmentation results on representative unseen samples [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Analysis of the effect of EM-Aggregation on text features. This figure compares [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Analysis of the effect of EM-Aggregation on visual features. This figure illus [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 2 linked inside Pith

  1. [22]

    Mamba as a bridge: Where vision foun- dation models meet vision language models for domain-generalized se- mantic segmentation,

    X. Zhang and T. Robby T., “Mamba as a bridge: Where vision foun- dation models meet vision language models for domain-generalized se- mantic segmentation,” June 2025

    2025

  2. [1]

    A survey of unsupervised deep domain adaptation,

    G. Wilson and D. J. Cook, “A survey of unsupervised deep domain adaptation,”ACM Transactions on Intelligent Systems and Technology (TIST), vol. 11, no. 5, pp. 1–46, 2020

    2020

  3. [2]

    A comprehensive survey on source-free domain adaptation,

    J. Li, Z. Yu, Z. Du, L. Zhu, and H. T. Shen, “A comprehensive survey on source-free domain adaptation,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 8, pp. 5743–5762, 2024. 24

    2024

  4. [3]

    Domaingeneralizationthroughmeta- learning: a survey,

    A.G.Khoee, Y.Yu, andR.Feldt, “Domaingeneralizationthroughmeta- learning: a survey,”Artificial Intelligence Review, vol. 57, no. 10, p. 285, 2024

    2024

  5. [4]

    Open domain gen- eralization with domain-augmented meta-learning,

    Y. Shu, Z. Cao, C. Wang, J. Wang, and M. Long, “Open domain gen- eralization with domain-augmented meta-learning,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 9624–9633, 2021

    2021

  6. [5]

    Reducing domain gap in frequency and spatial domain for cross-modality domain adaptation on medical image segmentation,

    S. Liu, S. Yin, L. Qu, and M. Wang, “Reducing domain gap in frequency and spatial domain for cross-modality domain adaptation on medical image segmentation,” Nov 2022

    2022

  7. [6]

    Domain generaliza- tion: A survey,

    K. Zhou, Z. Liu, Y. Qiao, T. Xiang, and C. C. Loy, “Domain generaliza- tion: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, p. 1–20, Jan 2022

    2022

  8. [7]

    Adversarial consistency for single domain generalization in medical image segmentation,

    Y. Xu, S. Xie, M. Reynolds, M. Ragoza, M. Gong, and K. Batmanghe- lich, “Adversarial consistency for single domain generalization in medical image segmentation,”

Show all 42 references
  1. [8]

    Structure-aware single-source gen- eralization with pixel-level disentanglement for joint optic disc and cup segmentation,

    J.-X. Jiang, Y. Li, and Z. Wang, “Structure-aware single-source gen- eralization with pixel-level disentanglement for joint optic disc and cup segmentation,”Biomedical Signal Processing and Control, vol. 99, p. 106801, 2025

    2025

  2. [9]

    Multi- receptive field feature disentanglement with distance-aware gaussian brightness augmentation for single-source domain generalization in med- ical image segmentation,

    Z. Wang, J.-x. Jiang, S. Zeng, L. Zhou, Y. Li, and Z. Wang, “Multi- receptive field feature disentanglement with distance-aware gaussian brightness augmentation for single-source domain generalization in med- ical image segmentation,”Neurocomputing, vol. 638, p. 130120, 2025

    2025

  3. [10]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  4. [11]

    Learningtransferablevisualmodelsfromnaturallanguagesupervision,

    A. Radford, J. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sas- try, A. Amanda, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learningtransferablevisualmodelsfromnaturallanguagesupervision,” Cornell University - arXiv,Cornell University - arXiv, Feb 2021. 25

    2021

  5. [12]

    Textual query-driven mask transformer for domain generalized segmentation,

    B. Pak, B. Woo, S. Kim, D.-h. Kim, and H. Kim, “Textual query-driven mask transformer for domain generalized segmentation,” inEuropean Conference on Computer Vision, pp. 37–54, Springer, 2024

    2024

  6. [13]

    Uni- fied contrastive learning in image-text-label space,

    J. Yang, C. Li, P. Zhang, B. Xiao, C. Liu, L. Yuan, and J. Gao, “Uni- fied contrastive learning in image-text-label space,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 19163–19173, 2022

    2022

  7. [14]

    Pyramidclip: Hierarchical feature alignment for vision-language model pretraining,

    Y. Gao, J. Liu, Z. Xu, J. Zhang, K. Li, R. Ji, and C. Shen, “Pyramidclip: Hierarchical feature alignment for vision-language model pretraining,” Advances in neural information processing systems, vol. 35, pp. 35959– 35970, 2022

    2022

  8. [15]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. Vo, M. Szafraniec, V. Khali- dov, P.Fernandez, D.Haziza, F.Massa, A.El-Nouby, M.Assran, N.Bal- las, W. Galuba, R. Howes, P.-Y. Huang, S.-W. Li, I. Misra, M. Rab- bat, V. Sharma, G. Synnaeve, H. Xu, H. Jegou, J. Mairal, P. Labatut, A. Joul...

  9. [16]

    Stronger fewer & superior: Harnessing vision foundation models for domain generalized semantic segmentation,

    Z. Wei, L. Chen, Y. Jin, X. Ma, T. Liu, P. Ling, B. Wang, H. Chen, and J. Zheng, “Stronger fewer & superior: Harnessing vision foundation models for domain generalized semantic segmentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition...

    2024

  10. [17]

    Causality-inspired single-source domain generalization for medical im- age segmentation,

    C. Ouyang, C. Chen, S. Li, Z. Li, C. Qin, W. Bai, and D. Rueckert, “Causality-inspired single-source domain generalization for medical im- age segmentation,”

  11. [18]

    Devil is in channels: Contrastive single domain generalization for medical image segmentation,

    S. Hu, Z. Liao, and Y. Xia, “Devil is in channels: Contrastive single domain generalization for medical image segmentation,” Jun 2023

    2023

  12. [19]

    Single domain generalization for multimodal cross-cancer prognosis via dirac rebalancer and distribution entanglement,

    J.-X. Jiang, J. Liu, H. Wu, Y. Wu, Z. Wang, Q. Bi, and Y. Zheng, “Single domain generalization for multimodal cross-cancer prognosis via dirac rebalancer and distribution entanglement,” inProceedings of the 33rd ACM International Conference on Multimedia, 2025

    2025

  13. [20]

    Denseclip: Language-guided dense prediction with context-aware 26 prompting,

    Y. Rao, W. Zhao, G. Chen, Y. Tang, Z. Zhu, G. Huang, J. Zhou, and J. Lu, “Denseclip: Language-guided dense prediction with context-aware 26 prompting,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

    2022

  14. [21]

    Maximum likelihood from incomplete data via the em algorithm,

    A. P. Dempster, N. M. Laird, and D. B. Rubin, “Maximum likelihood from incomplete data via the em algorithm,”Journal of the Royal Sta- tistical Society. Series B (Methodological), vol. 39, no. 1, pp. 1–38, 1977

    1977

  15. [23]

    Acceleration of the em algorithm by using quasi-newton methods,

    M. Jamshidian and R. I. Jennrich, “Acceleration of the em algorithm by using quasi-newton methods,”Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 59, no. 3, pp. 569–587, 1997

    1997

  16. [24]

    Prompt tuning for parameter- efficient medical image segmentation,

    M. Fischer, A. Bartler, and B. Yang, “Prompt tuning for parameter- efficient medical image segmentation,”Medical Image Analysis, vol. 91, p. 103024, 2024

    2024

  17. [25]

    Cat: Coordi- nating anatomical-textual prompts for multi-organ and tumor segmen- tation,

    Z. Huang, Y. Jiang, R. Zhang, S. Zhang, and X. Zhang, “Cat: Coordi- nating anatomical-textual prompts for multi-organ and tumor segmen- tation,”Advances in Neural Information Processing Systems, vol. 37, pp. 3588–3610, 2024

    2024

  18. [26]

    Multi-modality cross attention network for image and sentence matching,

    X. Wei, T. Zhang, Y. Li, Y. Zhang, and F. Wu, “Multi-modality cross attention network for image and sentence matching,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10941–10950, 2020

    2020

  19. [27]

    Hierarchical self-attention network for industrial data series modeling with different sampling rates between the input and output sequences,

    X. Yuan, Z. Jia, Z. Xu, N. Xu, L. Ye, K. Wang, Y. Wang, C. Yang, W. Gui, and F. Shen, “Hierarchical self-attention network for industrial data series modeling with different sampling rates between the input and output sequences,”IEEE Transactions on Neural Networks and Learn- ...

    2024

  20. [28]

    Transformers in vision: A survey,

    S. Khan, M. Naseer, M. Hayat, S. W. Zamir, F. S. Khan, and M. Shah, “Transformers in vision: A survey,”ACM Computing Surveys, p. 1–41, Jan 2022. 27

    2022

  21. [29]

    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,” in2016 Fourth International Conference on 3D Vision (3DV), Oct 2016

    2016

  22. [30]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,”International Journal of Computer Vision, vol. 130, no. 9, pp. 2337–2348, 2022

    2022

  23. [31]

    A similarity paradigm through textual regularization without forgetting,

    F. Cui, J. Fong, R. Zeng, X. Tian, and J. Yu, “A similarity paradigm through textual regularization without forgetting,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, pp. 16100–16108, 2025

    2025

  24. [32]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  25. [33]

    Deep learning techniques for automatic mri cardiac multi-structures segmen- tation and diagnosis: Is the problem solved?,

    O. Bernard, A. Lalande, C. Zotti, F. Cervenansky, X. Yang, P.-A. Heng, I. Cetin, K. Lekadir, O. Camara, M. A. Gonzalez Ballester, G. San- roma, S. Napel, S. Petersen, G. Tziritas, E. Grinias, M. Khened, V. A. Kollerathu, G. Krishnamurthi, M.-M. Rohé, X. Pennec, M. Sermesant, F...

    2018

  26. [34]

    Fairdomain: Achievingfairnessincross- domain medical image segmentation and classification,

    Y. Tian, C. Wen, M. Shi, M. M. Afzal, H. Huang, M. O. Khan, Y.Luo, Y.Fang, andM.Wang, “Fairdomain: Achievingfairnessincross- domain medical image segmentation and classification,”arXiv preprint arXiv:2407.08813, 2024

    2024 arXiv

  27. [35]

    U-net: Convolutional net- works for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional net- works for biomedical image segmentation,” inMedical Image Comput- ing and Computer-Assisted Intervention – MICCAI 2015(N. Navab, J. Hornegger, W. M. Wells, and A. F. Frangi, eds.), (Cham), pp. 234– 241, Spring...

    2015

  28. [36]

    H2former: An efficient hierarchical hybrid transformer for medical image segmentation,

    A. He, K. Wang, T. Li, C. Du, S. Xia, and H. Fu, “H2former: An efficient hierarchical hybrid transformer for medical image segmentation,”IEEE Transactions on Medical Imaging, vol. 42, no. 9, pp. 2763–2775, 2023. 28

    2023

  29. [37]

    Improved regularization of convolutional neu- ral networks with cutout,

    T. Devries and G. Taylor, “Improved regularization of convolutional neu- ral networks with cutout,”

  30. [38]

    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,” Cornell University - arXiv,Cornell University - arXiv, Jul 2020

    2020

  31. [39]

    Domain generalization with mixstyle,

    K. Zhou, Y. Yang, Y. Qiao, and T. Xiang, “Domain generalization with mixstyle,”Learning,Learning, Apr 2021

    2021

  32. [40]

    Rethinking data augmentation for single-source domain generalization in medical image segmentation,

    Z. Su, K. Yao, X. Yang, Q. Wang, J. Sun, and K. Huang, “Rethinking data augmentation for single-source domain generalization in medical image segmentation,” Nov 2022

    2022

  33. [41]

    Prompting seg- ment anything model with domain-adaptive prototype for generalizable medical image segmentation,

    Z. Wei, W. Dong, P. Zhou, Y. Gu, Z. Zhao, and Y. Xu, “Prompting seg- ment anything model with domain-adaptive prototype for generalizable medical image segmentation,” inMedical Image Computing and Com- puter Assisted Intervention – MICCAI 2024(M. G. Linguraru, Q. Dou, A. Ferag...

    2024

  34. [42]

    A theory of learning from different domains,

    P. Reviewed, S. Blitzer, J. Crammer, K. Kulesza, A. Pereira, F. Vaughan, J. Wortman, S. Ben-David, J. Blitzer, K. Crammer, A. Kulesza, F. Pereira, N. Cesa-Bianchi, D. Hardoon, G. Leen, D. Cheri- ton, and J. Vaughan, “A theory of learning from different domains,” 29

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.