Pith. sign in

REVIEW 4 major objections 7 minor 49 references

Decoupled Competitive Framework for Semi-supervised Medical Image Segmentation

T0 review · 4 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Two competing students alternately update a shared teacher and mentor the loser, and the paper reports this decoupled competitive framework sets new state-of-the-art Dice on LA, Pancreas-CT, and ISIC under low annotation.

desk verdict Plausible incremental co-training twist with a clean tutoring ablation, but the SOTA claim needs better baselines and variance analysis before I'd trust it. read the letter →

arxiv 2505.24667 v1 pith:2OQ2ZDTV submitted 2025-05-30 cs.CV

Jiahe Chen , Jiahe Ying , Shen Wang , Jianwei Zheng This is my paper · ORCID
classification cs.CV
keywords semi-supervisedlearningmedicalimagesegmentationmeanteacherdualstudentcompetitivepseudo-labelsupervisionconsistencyregularizationdecoupledframework
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that the two dominant architectures for semi-supervised medical image segmentation—Mean Teacher and Dual Student—share a fixable flaw: the teacher is too tightly coupled to one student, or the two students drift into mutually reinforcing errors. The proposed Decoupled Competitive Framework (DCF) instead lets two student networks compete on the labeled data, uses the winner's weights to update a shared teacher via exponential moving average, and has the teacher tutor the losing student with pseudo-labels. The authors report state-of-the-art Dice scores on three public benchmarks: left atrium MRI (89.94% with 10% labels), pancreas CT (78.94% with 10% labels), and ISIC dermoscopy (89.23% with 10% labels). A sympathetic reader would care because the framework needs no extra modules or parameters, so if correct it offers a simple replacement for EMA coupling in low-annotation medical settings.

What carries the argument

The load-bearing object is a competition function evaluated on the labeled subset each step. It compares the two students using metrics already computed during supervised training—Dice, cross-entropy, 95% Hausdorff distance—and declares a winner. The winner's parameters enter the teacher's exponential moving average, $\theta'_t=\alpha\theta'_{t-1}+(1-\alpha)\theta_t$, while the loser receives extra supervision through a mentoring loss $L_{ms}$ that matches its predictions on unlabeled data to the teacher's pseudo-labels; the two students also supervise each other through cross-pseudo supervision $L_{cps}$. The combined unsupervised loss $L_{unsup}=L_{cps}+L_{ms}$ is added to the supervised loss with a ramp-up weight $\lambda$.

What would settle it

Replace the competition with a random choice of which student updates the teacher, keeping the mentoring and cross-pseudo-supervision losses unchanged; if Dice on held-out test sets does not drop, the competition mechanism itself is doing none of the work attributed to it.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that the performance bottleneck in Mean Teacher comes from tying the EMA teacher to a single student, and the bottleneck in Dual Student comes from letting two students exchange knowledge without a referee. DCF introduces a dynamic referee: at every iteration the two students are scored on the labeled batch by Dice, cross-entropy, or 95% Hausdorff distance; the winner updates the teacher by EMA, and the teacher, acting as a mentor, generates pseudo-labels that supervise the loser. Because the winner changes over training, the teacher's parameters are not locked to any one student, and because the teacher mediates, the students can be aligned with a simple cross-pseudo-supervision loss without collapse. The paper reports that this mechanism improves Dice over previous methods on LA, Pancreas-CT, and ISIC under 5–20% label budgets, and that weight-distance curves show the teacher alternating between the two students.

Load-bearing premise

The winner of the competition is chosen by performance on the few labeled samples available (for example, 8 volumes on LA and 6 on Pancreas-CT), and the paper assumes this small-sample signal reliably identifies the better student; if that signal is noisy, a wrong winner would corrupt both the EMA teacher and the mentoring of the loser.

Editorial extensions

If this is right

  • With only 10% of LA labels, DCF reports a Dice of 89.94%, above the 89.62% of the best compared method (BCP); at 20% it reaches 91.44%.
  • On Pancreas-CT at 10% labels, DCF reports a Dice of 78.94% against 76.62% for RCPS, and the gap between its 10% and 20% results is the smallest among compared methods, indicating more efficient use of unlabeled data.
  • On the ISIC 2D dataset, DCF reports Dice of 88.88% (5% labels) and 89.23% (10% labels), ahead of UCMT's 88.22% and 88.46%.
  • The framework needs no extra network modules beyond one teacher and two students, so it adds no parameters and minimal extra computation, since the competition metrics are already computed for supervision.
  • The authors argue the same decoupled competitive design can serve as a plug-and-play module in other semi-supervised learning settings.

Reading between the lines

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

  • Editorial inference: the reported gains could come as much from having two students plus a teacher, rather than from the competition itself; a random-winner control would separate these factors.
  • Editorial inference: the competition signal is computed on the labeled subset at each iteration, so on datasets with 6–8 labeled volumes the winner may fluctuate; the paper does not analyze this sensitivity, and a wrong winner could propagate errors.
  • Editorial inference: the same decoupling idea transfers naturally to other teacher-student SSL tasks where a cheap per-step performance metric exists, such as classification accuracy on a labeled memory set.
  • Editorial inference: the antagonistic weight-distance trend between the teacher and the two students suggests the students take turns pulling the teacher; measuring this alternation rate could serve as a diagnostic for whether the competition is behaving as intended.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes a Decoupled Competitive Framework (DCF) for semi-supervised medical image segmentation. DCF maintains two student networks and one teacher network. At each iteration, a competition function evaluated on the labeled mini-batch selects the better-performing student; that student's weights update the teacher via exponential moving average, while the teacher provides pseudo-labels to mentor the weaker student. The framework is evaluated on three public datasets (LA MRI, Pancreas-CT, ISIC dermoscopy) under several low-label regimes, with reported Dice improvements over prior methods, and ablation studies on the tutoring mechanism and on the choice of competition metric.

Significance. If the reported results are reliable, the paper would make a useful contribution: it presents a conceptually simple mechanism to mitigate teacher-student coupling in Mean Teacher and Dual Student architectures, and it is evaluated on external benchmarks with publicly available datasets. The work also provides ablations for both the competition metric and the tutoring policy, which are informative despite the concerns below. The promised code availability (https://github.com/JiaheChen2002/DCF) is a strength, though the manuscript does not currently link to a working repository allowing verification. The central idea of dynamically selecting the student to update the teacher is plausible, but the paper does not yet establish the statistical reliability of the winner-selection signal under very small labeled sets, which is load-bearing for the claimed state-of-the-art results.

major comments (4)
  1. [§4.4, Table 2] The Pancreas-CT results are obtained by directly adopting numbers reported in other papers, while the DCF results come from the authors' own re-implementation with their own preprocessing and training schedule. This protocol mismatch makes the cross-method comparison unreliable; differences in cropping, resampling, number of epochs, and optimizer settings can easily account for Dice differences of the size reported here (e.g., 78.94 vs. 76.62). The authors should re-run the baseline methods on the same data split and preprocessing, or provide a detailed protocol-compatibility argument; otherwise the Pancreas-CT state-of-the-art claim is not supported.
  2. [§5, Table 5; §4.2] The competition metric (Dice) is selected after inspecting the ablation study on the LA dataset with 20% labeled data, which is part of the reported test set. This is a form of test-set tuning. The sensitivity is not negligible: Table 5 shows that switching from Dice to Jaccard changes the final Dice by 2.12 points (91.44 to 89.32), while the headline 10% improvement over BCP on LA is only 0.32 Dice. The paper does not report results on a held-out validation set for metric selection, nor multiple random seeds, so the reader cannot tell whether the reported margin is above the noise floor. At minimum, the metric-selection process should be described as part of the training protocol and its variance documented.
  3. [§3.2, Algorithm 1; §4.2] The winner signal is computed on the current labeled mini-batch, which at 10% LA supervision contains only 2 labeled volumes (batch size 4 with 2 labeled), and on Pancreas-CT 6 labeled volumes in total. The choice of winner can therefore flip frequently due to mini-batch noise. A wrong winner simultaneously (a) updates the EMA teacher with the weaker student and (b) routes mentoring to the stronger student, directly inverting the intended dynamics. The manuscript provides no sensitivity analysis, no comparison with a random-winner control, and no oracle-winner upper bound. This is a load-bearing gap because the entire contribution rests on the reliability of this signal. I recommend adding a random-winner baseline and a multi-seed variance analysis, or at least a theoretical argument for why the mini-batch Dice estimate is stable enough.
  4. [§5, Table 4] The tutoring ablation shows a difference of 1.18 Dice between the proposed policy (row 5, 91.44) and the no-tutoring case (row 1, 91.10), and other rows differ by even less. Without multiple random seeds or statistical significance testing, these differences could be within run-to-run variation. The paper's conclusion that tutoring the underperforming student is 'optimal' should be supported by error bars or repeated runs, especially since the same table is used to justify a core design choice.
minor comments (7)
  1. [Abstract and §1] The abstract contains the phrase 'the severe cognitive bias between two students' but the term 'cognitive bias' is used metaphorically and never defined; consider replacing with a more precise term such as 'performance gap' or 'prediction disagreement'.
  2. [§4.4, Figure 4] Figure 4 is cited as showing a 'significantly smaller performance gap' but it appears to be a bar chart without error bars or numeric labels; the claim of significance is not supported by the figure alone.
  3. [Algorithm 1] In Algorithm 1, the notation for pseudo-labels (Y-hat) is not explicitly defined in the algorithm box; it first appears in Equation (2), but a reader scanning the algorithm may be confused about which network's output is used as the pseudo-label.
  4. [§4.1] The ISIC dataset section says the training set has 1815 images and validation 779, but Table 3 lists 90/181 labeled images for 5%/10% settings; the relationship to the 1815 training images should be stated explicitly (e.g., whether validation images are excluded).
  5. [§3.2, Eq. (4)] Equation (4) defines L_unsup as L_cps + L_ms, but the manuscript does not state whether the mentoring loss is applied only to the current loser (which is described in the text) or to both students; please clarify the exact assignment in the equation.
  6. [§6] The conclusion acknowledges that the dynamic interplay between students 'gradually wanes' over training, but this limitation is not analyzed or quantified, and it is not connected to the earlier figures showing prediction distance; consider reporting the temporal dynamics quantitatively.
  7. [Overall] Several tables use inconsistent capitalization for V-Net (e.g., Vnet vs. V-Net) and 'Volumes' appears as 'V olumes' in Tables 1 and 3; these typos should be fixed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DCF's central claim is an empirical result on held-out benchmarks, not a derivation from its own inputs.

full rationale

The paper proposes a training scheme (DCF) and evaluates it on three public datasets. The claimed state-of-the-art performance is a measured outcome against held-out test splits, with no step in which a test metric is constructed from the method's own definitions. Equations (1)-(5) define training losses (supervised segmentation, mentoring, cross-pseudo-supervision, and total loss); they are not predictions of the Table 1-3 results. The competition mechanism uses a metric on the labeled mini-batch to choose which student updates the EMA teacher, but the reported Dice/Jaccard/95HD/ASD numbers come from the evaluation protocol, not from that competition function. The ablation in Table 5 chooses Dice as the competition metric by comparing variants on the LA benchmark; this is model selection on an evaluation set and a legitimate methodological weakness, but it does not make the final result equivalent to the method's inputs by construction. The acknowledged limitation that the student interplay wanes during later training is an empirical observation, not a circular claim. There are no load-bearing self-citations, no imported uniqueness theorems, and no renamed known result used as a first-principles derivation. Therefore no circular step can be exhibited.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The framework introduces no new physical or model entities. It relies on standard SSL assumptions and hyperparameters, several of which are not reported or are tuned on the test set.

free parameters (3)
  • EMA decay α = not reported
    Controls the teacher update rate in Eq. (1) and is central to the EMA mechanism; its value is not stated in the paper.
  • Unsupervised loss weight λ = not reported
    Balances supervised and unsupervised losses in Eq. (5); value not specified.
  • Competition metric selection = Dice
    The choice of which metric (Dice, CE, Jac, etc.) determines the winner is made after evaluating ablation performance on LA (Table 5), effectively a fitted choice on the test set.
assumptions (4)
  • domain assumption Smoothness, cluster, and manifold assumptions of semi-supervised learning hold for medical images.
    Section 2.1 lists these assumptions as the foundation of the method.
  • domain assumption The competition metric computed on the small labeled batch is a reliable indicator of student performance on the data distribution.
    Section 3.2 uses this to select the winner for EMA teacher update; with only 6 or 8 labeled volumes, the signal may be noisy.
  • domain assumption Pseudo-labels from the EMA teacher are sufficiently accurate to guide the underperforming student.
    Used in Eq. (2) for the mentoring loss; no confidence thresholding or uncertainty weighting is applied.
  • domain assumption The alternating EMA update prevents collapse and maintains diverse students.
    Section 3.3 asserts this based on weight distance plots, but no proof or rigorous analysis is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Decoupled Competitive Framework for Semi-supervised Medical Image Segmentation." pith.science (2026). https://pith.science/paper/2OQ2ZDTV

@misc{pith2026250524667,
  author       = {Pith},
  title        = {Pith review of: Decoupled Competitive Framework for Semi-supervised Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2OQ2ZDTV}},
  note         = {Machine review of arXiv:2505.24667}
}
read the original abstract

Confronting the critical challenge of insufficiently annotated samples in medical domain, semi-supervised medical image segmentation (SSMIS) emerges as a promising solution. Specifically, most methodologies following the Mean Teacher (MT) or Dual Students (DS) architecture have achieved commendable results. However, to date, these approaches face a performance bottleneck due to two inherent limitations, \textit{e.g.}, the over-coupling problem within MT structure owing to the employment of exponential moving average (EMA) mechanism, as well as the severe cognitive bias between two students of DS structure, both of which potentially lead to reduced efficacy, or even model collapse eventually. To mitigate these issues, a Decoupled Competitive Framework (DCF) is elaborated in this work, which utilizes a straightforward competition mechanism for the update of EMA, effectively decoupling students and teachers in a dynamical manner. In addition, the seamless exchange of invaluable and precise insights is facilitated among students, guaranteeing a better learning paradigm. The DCF introduced undergoes rigorous validation on three publicly accessible datasets, which encompass both 2D and 3D datasets. The results demonstrate the superiority of our method over previous cutting-edge competitors. Code will be available at https://github.com/JiaheChen2002/DCF.

Figures

Figures reproduced from arXiv: 2505.24667 by the authors.

Figure 1
Figure 1. Different SSL structures. (a) Mean Teacher. (b) Dual Student. (c) Cross Pseudo Supervision. (d) Perturbed and Strict Mean Teachers. (e) Uncertainty-guided Collaborative Mean Teacher. (f) Decoupled Competitive Framework (ours). Mean Teacher model [30] has dominated in this field for a long time, inspiring numerous followers seeking notable advancements [42, 19, 27, 39]. Yet, Ke et al. [13] and Zhao et al. [46] have d… view at source ↗
Figure 3
Figure 3. (a) denotes the weight distance between teacher and student. (b) represents the prediction distance among three networks. For simplicity, the simple Euclidean distance is used for weight and prediction distances. The overall framework of the proposed DCF is shown in Fig￾ure 2, which mainly comprises three networks: a teacher network f(·; θt) and two student networks f(·; θs1) and f(·; θs2), all initial￾ized randomly… view at source ↗
Figure 2
Figure 2. Block A introduces our proposed DCF. Upon receiving input data, DCF undergoes two random data augmentations. Then, three separate networks follow: a student network and two teacher networks. Operating on a co-teaching scheme, DCF fosters cross-pseudo supervision between two student results. In Block B, a competitive mechanism employing metrics such as Dice, Cross-entropy, and 95HD is elaborated during training to co… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Dice scores for 10% and 20% labeled data across various models on the Pancreas-CT dataset. Our method exhibits a significantly smaller performance gap between these two cases. were preprocessed, involving adjustments of Hounsfield Units (HU) to ranges of [−125, 275] or…
Figure 5
Figure 5. Figure 5: Comparison of visualization results on LA Dataset. The first row shows the results in 2D form, while the second row provides the visualizations in 3D form, where certain details have been enlarged for better clarity. (a) UAMT (b) SASSNet (c) DTC (d) MC-Net+ (e) RCPS (f…
Figure 6
Figure 6. Figure 6: Comparison of visual results on Pancreas-CT Dataset. The first row shows the results in 2D form, while the second row provides the visualizations in 3D form, where certain details have been enlarged for better clarity [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages

  1. [1]

    Y . Bai, D. Chen, Q. Li, W. Shen, and Y . Wang. Bidirectional copy-paste for semi-supervised medical image segmentation. InProc.CVPR, pages 11514–11524, 2023

    work page 2023

  2. [2]

    Basak and Z

    H. Basak and Z. Yin. Pseudo-label guided contrastive learning for semi- supervised medical image segmentation. InProc.CVPR, pages 19786– 19797, 2023

    work page 2023

  3. [3]

    Cascante-Bonilla, F

    P. Cascante-Bonilla, F. Tan, Y . Qi, and V . Ordonez. Curriculum labeling: Revisiting pseudo-labeling for semi-supervised learning. InProc.AAAI, volume 35, pages 6912–6920, 2021

    work page 2021

  4. [4]

    D. Chen, Y . Bai, W. Shen, Q. Li, L. Yu, and Y . Wang. Magicnet: Semi- supervised multi-organ segmentation via magic-cube partition and re- covery. InProc.CVPR, pages 23869–23878, 2023

    work page 2023

  5. [5]

    X. Chen, Y . Yuan, G. Zeng, and J. Wang. Semi-supervised semantic seg- mentation with cross pseudo supervision. InProc.CVPR, pages 2613– 2622, 2021

    work page 2021

  6. [6]

    N. C. Codella, D. Gutman, M. E. Celebi, B. Helba, M. A. Marchetti, S. W. Dusza, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler, et al. Skin lesion analysis toward melanoma detection: A challenge at the 2017 international symposium on biomedical imaging (isbi), hosted by the international skin imaging collaboration (isic). InProc.ISBI, pages 168– 172, 2018

    work page 2017

  7. [7]

    D.-P. Fan, T. Zhou, G.-P. Ji, Y . Zhou, G. Chen, H. Fu, J. Shen, and L. Shao. Inf-net: Automatic covid-19 lung infection segmentation from ct images.IEEE Transactions on Medical Imaging, 39(8):2626–2637, 2020

    work page 2020

  8. [8]

    S. Gao, Z. Zhang, J. Ma, Z. Li, and S. Zhang. Correlation-aware mutual learning for semi-supervised medical image segmentation. In Proc.MICCAI, pages 98–108. Springer, 2023

    work page 2023

Show all 49 references
  1. [9]

    Grünberg, O

    K. Grünberg, O. A. J. del Toro, A. Jakab, G. Langs, T. S. Fernandez, M. Winterstein, M.-A. Weber, and M. Krenn. Annotating medical image data. InCloud-Based Benchmarking of Medical Image Analysis, 2017

    2017

  2. [10]

    Huang, Z

    F. Huang, Z. Yao, and W. Zhou. Dtbs: Dual-teacher bi-directional self- training for domain adaptation in nighttime semantic segmentation. In Proc.ECAI, 2024

    2024

  3. [11]

    Huang, C

    W. Huang, C. Chen, Z. Xiong, Y . Zhang, X. Chen, X. Sun, and F. Wu. Semi-supervised neuron segmentation via reinforced consistency learn- ing.IEEE Transactions on Medical Imaging, 41(11):3016–3028, 2022

    2022

  4. [12]

    Huang, Z

    X. Huang, Z. Deng, D. Li, X. Yuan, and Y . Fu. Missformer: An effective transformer for 2d medical image segmentation.IEEE Transactions on Medical Imaging, 42(5):1484–1494, 2023

    2023

  5. [13]

    Z. Ke, D. Wang, Q. Yan, J. Ren, and R. W. Lau. Dual student: Breaking the limits of the teacher in semi-supervised learning. InProc.ICCV, pages 6728–6736, 2019

    2019

  6. [14]

    Kwon and S

    D. Kwon and S. Kwak. Semi-supervised semantic segmentation with error localization network. InProc.CVPR, pages 9957–9967, 2022

    2022

  7. [15]

    Laine and T

    S. Laine and T. Aila. Temporal ensembling for semi-supervised learn- ing. InProc.ICLR, 2017

    2017

  8. [16]

    D.-H. Lee. Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks. 2013. URL https://api. semanticscholar.org/CorpusID:18507866

    2013

  9. [17]

    S. Li, C. Zhang, and X. He. Shape-aware semi-supervised 3d semantic segmentation for medical images. InProc.MICCAI, pages 552–561, 2020

    2020

  10. [18]

    C. Liu, C. Gao, F. Liu, P. Li, D. Meng, and X. Gao. Hierarchical su- pervision and shuffle data augmentation for 3d semi-supervised object detection. InProc.CVPR, pages 23819–23828, 2023

    2023

  11. [19]

    Y . Liu, Y . Tian, Y . Chen, F. Liu, V . Belagiannis, and G. Carneiro. Per- turbed and strict mean teachers for semi-supervised semantic segmen- tation. InProc.CVPR, pages 4258–4267, 2022

    2022

  12. [20]

    X. Luo, J. Chen, T. Song, and G. Wang. Semi-supervised medical image segmentation through dual-task consistency. InProc.AAAI, volume 35, pages 8801–8809, 2021

    2021

  13. [21]

    F. Lyu, M. Ye, J. F. Carlsen, K. Erleben, S. Darkner, and P. C. Yuen. Pseudo-label guided image synthesis for semi-supervised covid-19 pneumonia infection segmentation.IEEE Transactions on Medical Imaging, 42(3):797–809, 2023. doi: 10.1109/TMI.2022.3217501

    2023

  14. [22]

    J. Miao, C. Chen, F. Liu, H. Wei, and P.-A. Heng. Caussl: Causality- inspired semi-supervised learning for medical image segmentation. Proc.ICCV, pages 21369–21380, 2023

    2023

  15. [23]

    Mittal, M

    S. Mittal, M. Tatarchenko, and T. Brox. Semi-supervised semantic seg- mentation with high-and low-level consistency.IEEE transactions on pattern analysis and machine intelligence, 43(4):1369–1379, 2019

    2019

  16. [24]

    Ouali, C

    Y . Ouali, C. Hudelot, and M. Tami. Semi-supervised semantic segmen- tation with cross-consistency training. InProc.CVPR, pages 12674– 12684, 2020

    2020

  17. [25]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox. U-net: Convolutional networks for biomedical image segmentation. InProc.MICCAI, pages 234–241. Springer, 2015

    2015

  18. [26]

    H. R. Roth, L. Lu, A. A. Farag, H.-C. Shin, J. Liu, E. B. Turkbey, and R. M. Summers. Deeporgan: Multi-level deep convolutional networks for automated pancreas segmentation. InProc.MICCAI, page 556–564,

  19. [27]

    Z. Shen, P. Cao, H. Yang, X. Liu, J. Yang, and O. R. Zaiane. Co-training with high-confidence pseudo labels for semi-supervised medical image segmentation. InProc.IJCAI, pages 4199–4207, 2023

    2023

  20. [28]

    K. Sohn, D. Berthelot, N. Carlini, Z. Zhang, H. Zhang, C. A. Raffel, E. D. Cubuk, A. Kurakin, and C.-L. Li. Fixmatch: Simplifying semi- supervised learning with consistency and confidence. InProc.NeurIPS, volume 33, pages 596–608, 2020

    2020

  21. [29]

    Taleb, W

    A. Taleb, W. Loetzsch, N. Danz, J. Severin, T. Gaertner, B. Bergner, and C. Lippert. 3d self-supervised methods for medical imaging. In Proc.NeurIPS, volume 33, pages 18158–18172, 2020

    2020

  22. [30]

    Tarvainen and H

    A. Tarvainen and H. Valpola. Mean teachers are better role mod- els: Weight-averaged consistency targets improve semi-supervised deep learning results. InProc.NeurIPS, page 1195–1204, 2017

    2017

  23. [31]

    G. Wang, M. A. Zuluaga, W. Li, R. Pratt, P. A. Patel, M. Aertsen, T. Doel, A. L. David, J. Deprest, S. Ourselin, et al. Deepigeos: a deep interactive geodesic framework for medical image segmentation.IEEE transactions on pattern analysis and machine intelligence, 41(7):1559– 1...

    2018

  24. [32]

    G. Wang, S. Zhai, G. Lasio, B. Zhang, B. Yi, S. Chen, T. J. Macvit- tie, D. Metaxas, J. Zhou, and S. Zhang. Semi-supervised segmentation of radiation-induced pulmonary fibrosis from lung ct scans with multi- scale guided dense attention.IEEE Transactions on Medical Imaging, 41...

    2021

  25. [33]

    T. Wang, J. Lu, Z. Lai, J. Wen, and H. Kong. Uncertainty-guided pixel contrastive learning for semi-supervised medical image segmentation. In L. D. Raedt, editor,Proc.IJCAI, pages 1444–1450, 7 2022

    2022

  26. [34]

    Y . Wang, Y . Zhang, J. Tian, C. Zhong, Z. Shi, Y . Zhang, and Z. He. Double-uncertainty weighted method for semi-supervised learning. In Proc.MICCAI, pages 542–551, 2020

    2020

  27. [35]

    Y . Wang, H. Chen, Q. Heng, W. Hou, Y . Fan, Z. Wu, J. Wang, M. Sav- vides, T. Shinozaki, B. Raj, B. Schiele, and X. Xie. Freematch: Self- adaptive thresholding for semi-supervised learning. InProc.ICLR, 2023

    2023

  28. [36]

    H. Wu, Z. Wang, Y . Song, L. Yang, and J. Qin. Cross-patch dense contrastive learning for semi-supervised segmentation of cellular nuclei in histopathologic images. InProc.CVPR, pages 11656–11665, 2022

    2022

  29. [37]

    Y . Wu, Z. Ge, D. Zhang, M. Xu, L. Zhang, Y . Xia, and J. Cai. Mutual consistency learning for semi-supervised medical image segmentation. Medical Image Analysis, 81:102530, 2022

    2022

  30. [38]

    Y . Xia, D. Yang, Z. Yu, F. Liu, J. Cai, L. Yu, Z. Zhu, D. Xu, A. Yuille, and H. Roth. Uncertainty-aware multi-view co-training for semi- supervised medical image segmentation and domain adaptation.Medi- cal image analysis, 65:101766, 2020

    2020

  31. [39]

    Xiang, P

    J. Xiang, P. Qiu, and Y . Yang. Fussnet: Fusing two sources of uncertainty for semi-supervised medical image segmentation. In Proc.MICCAI, pages 481–491, 2022

    2022

  32. [40]

    Xiong, Q

    Z. Xiong, Q. Xia, Z. Hu, N. Huang, S. Vesal, N. Ravikumar, A. Maier, C. Li, Q. Tong, W. Si, et al. A global benchmark of algorithms for seg- menting late gadolinium-enhanced cardiac magnetic resonance imag- ing.Medical Image Analysis, page 101832, 2020

    2020

  33. [41]

    Yang and S

    Z. Yang and S. Farsiu. Directional connectivity-based segmentation of medical images. InProc.CVPR, pages 11525–11535, 2023

    2023

  34. [42]

    L. Yu, S. Wang, X. Li, C.-W. Fu, and P.-A. Heng. Uncertainty-aware self-ensembling model for semi-supervised 3d left atrium segmentation. InProc.MICCAI, pages 605–613, 2019

    2019

  35. [43]

    Zhang, Y

    B. Zhang, Y . Wang, W. Hou, H. Wu, J. Wang, M. Okumura, and T. Shi- nozaki. Flexmatch: Boosting semi-supervised learning with curriculum pseudo labeling. InProc.NeurIPS, 2021

    2021

  36. [44]

    X. Zhao, C. Fang, D.-J. Fan, X. Lin, F. Gao, and G. Li. Cross-level con- trastive learning and consistency constraint for semi-supervised medical image segmentation. InProc.ISBI, pages 1–5. IEEE, 2022

    2022

  37. [45]

    X. Zhao, Z. Qi, S. Wang, Q. Wang, X. Wu, Y . Mao, and L. Zhang. Rcps: Rectified contrastive pseudo supervision for semi-supervised medical image segmentation.IEEE Journal of Biomedical and Health Infor- matics, 2023

    2023

  38. [46]

    Y . Zhao, K. Lu, J. Xue, S. Wang, and J. Lu. Semi-supervised medi- cal image segmentation with voxel stability and reliability constraints. IEEE Journal of Biomedical and Health Informatics, 27(8):3912–3923,

  39. [47]

    Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, and J. Liang. Unet++: Redesigning skip connections to exploit multiscale features in image segmentation.IEEE Transactions on Medical Imaging, 2019

    2019

  40. [2015]

    URL https://api.semanticscholar.org/CorpusID:5776545

  41. [2023]

    doi: 10.1109/JBHI.2023.3273609

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.