Pith. sign in

REVIEW 3 major objections 4 minor 45 references

Bi-Level Collaborative Learning for Few-Shot Scribble-Supervised Medical Image Segmentation

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read With five scribble-annotated cases, a bi-level superpixel–segmentation loop reaches 0.828 Mean Dice on cardiac MRI and 0.534 on prostate MRI, surpassing prior scribble-supervised methods.

desk verdict Solid method paper for scribble-supervised segmentation, but headline gains rest on test-set-tuned hyperparameters and a single non-random few-shot split, so the size of the improvement is not yet trustworthy. read the letter →

arxiv 2607.25432 v1 pith:FNREE3ER submitted 2026-07-28 cs.CV

Xiang-Xiang Su , Yufan Ye , Yihang Zheng , Min Gan , Guang-Yong Chen This is my paper · ORCID
classification cs.CV
keywords medicalimagesegmentationscribblesupervisionfew-shotlearningsuperpixelspseudo-labelpropagationbileveloptimizationcardiacMRIprostate
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 argues that the dual challenge of few-shot and scribble supervision can be broken by coupling a learnable superpixel model with a segmentation model in a bi-level loop. The superpixel model supplies region-structural priors that turn sparse strokes into dense pseudo-labels; a spatial-prior-guided filter prevents those pseudo-labels from over-expanding into errors; and the segmentation model's learned semantics are fed back to reshape the superpixels so they align with anatomy rather than just low-level image edges. On ACDC and Prostate with only five scribble-annotated training cases, this raises Mean Dice by about five percentage points over prior methods and sharply lowers boundary error (HD95). A sympathetic reader would take this as evidence that structural priors, made learnable and task-aligned, can compensate for extreme supervision scarcity.

What carries the argument

The central mechanism is a bi-level loop: an upper-level differentiable Superpixel Sampling Network (SSN) generates soft pixel-to-superpixel assignments; the lower-level segmentation model (U-Net under Mean Teacher) uses those assignments to diffuse high-confidence pseudo-labels to whole superpixels; a spatial-prior-guided adaptive filter (Eqs. 8–10) caps each superpixel's expansion ratio as a Gaussian decay of its centroid's distance from the image center, and a center-dense seed reallocation (Eqs. 13–14) concentrates superpixels near the center. Semantic feedback from the segmentation model is added to the SSN's reconstruction objective, so superpixels evolve from generic boundary detector

What would settle it

Run the same five-case protocol on a dataset with off-center organs, such as liver or kidney segmentation from CT slices where organs appear at the periphery. If Mean Dice does not degrade relative to the centered cardiac/prostate results, the center prior is not load-bearing; if it collapses toward the no-filter ablation level (about 0.311 Dice), the assumption is confirmed as essential.

Watch

Extended reading notes

Core claim

Under a five-case scribble-supervised setting, the proposed BiSCL reports Mean Dice of 0.828 on ACDC versus 0.776 for the best prior method, and 0.534 on Prostate versus 0.478, with Mean HD95 dropping from 21.01 mm to 5.57 mm and from 32.19 mm to 16.09 mm respectively. Ablations attribute the gain to the spatial-prior-guided filtering: unconstrained superpixel diffusion collapses Mean Dice from 0.599 to 0.311, filtering restores it to 0.774, and adding the bi-level semantic feedback raises it to 0.828. The claim is that making superpixels learnable and driving them with segmentation feedback yields region priors aligned with anatomical semantics, enabling reliable dense pseudo-label generati

Load-bearing premise

The spatial-prior-guided filter and superpixel seeding assume target anatomical structures sit near the image center; if that assumption fails, the filter suppresses exactly the pseudo-label expansion the method relies on.

Editorial extensions

If this is right

  • With only five scribble-annotated training cases, BiSCL outperforms existing scribble-supervised methods on ACDC and Prostate in both region overlap (Dice) and boundary error (HD95).
  • The spatial-prior-guided filter is load-bearing: removing it turns superpixel diffusion from a benefit into a liability, dropping Mean Dice from 0.599 to 0.311 on ACDC.
  • Bilevel semantic feedback further improves the filtered diffusion, from 0.774 to 0.828 Mean Dice, indicating that superpixels aligned to segmentation semantics produce more reliable pseudo-labels.
  • The segmentation backbone is a plain U-Net, so the reported gains come from the supervision machinery rather than a specialized decoder, suggesting the approach transfers across architectures.

Reading between the lines

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

  • A direct test of the center prior: run BiSCL on a dataset with off-center target organs (e.g., liver or kidneys in CT slices where structures appear at the periphery). If performance degrades toward the no-filter level, the center assumption is confirmed as the bottleneck; if not, the filter is more flexible than the paper suggests.
  • Because the bilevel training deliberately drops the implicit Jacobian term in the upper-level gradient, a natural extension is a small-scale comparison against a full hypergradient estimator to quantify the bias introduced by this approximation.
  • The framework's reliance on low-level reconstruction plus semantic feedback could be carried over to other sparse-supervision regimes, such as point-click or box annotations, without changing the bilevel structure.
  • The large HD95 reduction (about 16 mm on both datasets) implies boundary quality improves substantially; a clinic-facing extension would check whether these contour changes alter downstream volume measurements in cardiac or prostate assessment.
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

3 major / 4 minor

Summary. The paper proposes BiSCL, a bi-level collaborative learning framework for few-shot scribble-supervised medical image segmentation. An upper-level learnable superpixel network (based on SSN) provides region-level structural priors; the lower-level segmentation network (U-Net in a Mean Teacher setting) uses superpixel-guided pseudo-label diffusion with a spatial-prior-guided adaptive filter to suppress unreliable expansion. Segmentation-derived semantic feedback is then used to refine the superpixel representation. The method is validated on ACDC and Prostate with only five scribble-annotated training cases, reporting Mean Dice improvements of 5.2 pp over QMaxViT-Unet+ on ACDC and 5.6 pp over ScribbleVC on Prostate, with corresponding HD95 reductions.

Significance. If the reported gains are reproducible, the framework is a useful contribution to weakly supervised medical image segmentation: it addresses a realistic setting where both annotation sparsity and annotated-sample scarcity coexist, and the superpixel/segmentation feedback loop is an interesting mechanism beyond static superpixel priors. The paper includes a clear formalization of the bi-level objective, a concrete training algorithm, and an informative ablation study showing that unconstrained superpixel diffusion hurts and the proposed filter restores performance. The central idea is therefore worth pursuing. However, the current experimental evidence is not sufficient to support the strong claim of 'significantly outperforms,' because hyperparameters are selected on the test set and the few-shot setting is evaluated on a single non-random split.

major comments (3)
  1. [§4.5, Table 4] The final reported results (Tables 1 and 2) use hyperparameters sigma=0.3 and gamma_center=6, which are selected in Table 4 because they 'achieve the best performance' on the ACDC test set. Using the test set for model selection makes the headline comparison optimistic: the 5.2 pp gain over the second-best baseline on ACDC may be inflated by test-set-specific tuning. A proper validation-based selection (e.g., choosing sigma and gamma_center on the existing validation split and only then reporting test performance) is required before the 'significantly outperforms' claim is supported.
  2. [§4.1, Tables 1–2] The few-shot evaluation uses only the first five training cases as labeled data, with no repeated random splits or multiple training seeds. The reported standard deviations are across test slices, not across training runs or split choices. With only five labeled cases, a single split can produce large variance in the comparison, so the 5.2–5.6 pp differences may not be stable. The authors should report results over multiple random few-shot splits and/or multiple seeds, and ideally include a paired significance test, to substantiate the claimed improvement over the second-best method.
  3. [§3.4, Eqs. (8)–(9) and (13)–(14)] Both the spatial-prior-guided filter and the non-uniform superpixel initialization rely on an unvalidated assumption that target anatomical structures lie near the image center. Equation (9) limits allowable expansion proportionally to a Gaussian of the centroid's distance from the center, and Eq. (14) redistributes superpixel seeds toward the center. The ablation (Table 3) shows the filter is indispensable: without it Mean Dice drops from 0.599 to 0.311. If the target organ is off-center in many slices, the method could suppress valid pseudo-label expansion and lose the only dense supervision signal. The paper should provide evidence about the actual centroid distribution of the target structures in ACDC and Prostate, and report sensitivity to this prior (e.g., by comparing with a version that uses a learned or data-dependent spatial prior, or by evaluating on slices where the target is
minor comments (4)
  1. [Eq. (7)] The formula for the expansion ratio r_k appears without a visible fraction bar in the text: r_k = |Ω_k| Σ ... . It should be written as a fraction with the denominator being the sum of high-confidence seeds. Please clarify.
  2. [Eq. (5)] The dominant seed class is computed as argmax over c in {1,...,C-1}, which excludes the background class (assumed to be 0). If scribble annotations never label background, this is fine, but it should be stated explicitly; otherwise the exclusion of background in the diffusion rule is not justified.
  3. [Algorithm 1] The reuse of theta_0 and w_0 as both initial parameters and loop-indexed variables is confusing. For example, line 8 sets theta_0 <- theta_hat while the outer loop still refers to theta_0; renaming would improve readability.
  4. [General] The paper states in a footnote that it has been accepted for publication at ACM MM 2026. This is not a technical issue, but it is unusual in a submitted manuscript and may raise editorial questions about prior disclosure; please verify venue policies.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity found: the bi-level loop is a standard co-training/self-training mechanism, and the reported test numbers are not derived from their own target values.

full rationale

I walked the claimed derivation chain. The upper-level superpixel network produces superpixel partitions; the teacher's confidence-filtered pseudo-labels (Eq. 4) are propagated within superpixels (Eqs. 5-6), filtered by a spatial-prior expansion-ratio bound (Eqs. 7-10), and used as dense supervision in the lower-level objective Lseg (Eq. 12). The segmentation model's teacher pseudo-labels are then used in the upper-level semantic reconstruction loss Lsem (Eq. 18), closing the loop. This is a mutually coupled self-training/co-training scheme, not a mathematical derivation of the claimed result from its own target: the ACDC and Prostate test ground truths enter only in Tables 1-2 for evaluation, not in any training loss (Eqs. 2-19). The spatial-prior-guided filtering is an explicit anatomical assumption, not a fitted parameter renamed as a prediction. There is no load-bearing self-citation: the only practice-justifying citations [10,11] are external works on bilevel optimization, and the learnable superpixel backbone SSN is adopted from external work [8]. The main caveat is evaluative rather than construction-circular: Section 4.5 selects sigma=0.3 and gamma_center=6 because they 'achieve the best performance' on the ACDC test set, and the few-shot split is a single non-random split, so the headline margin over baselines may be optimistic and should be confirmed on held-out validation splits. However, no equation or claim in the paper reduces by construction to its own inputs, so the circularity score is 0.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim is an empirical method improvement, so the ledger is dominated by hand-chosen hyperparameters and practical assumptions about pseudo-label reliability and anatomical centrality. No new physical or mathematical entities are introduced. The most fragile item is the center-of-image prior because the ablation shows the filter is the component that makes superpixel propagation viable.

free parameters (8)
  • gamma_center = 6
    Maximum expansion ratio allowed near the image center in the spatial-prior filter; tuned on ACDC test set in Table 4.
  • sigma = 0.3
    Decay rate of expansion allowance with distance from center; tuned on ACDC test set in Table 4.
  • gamma = 1.8
    Exponent in inverse radial transform for center-dense superpixel initialization; chosen by hand.
  • beta = 2.0
    Boundary emphasis weight in the low-level feature reconstruction loss.
  • tau = 0.9
    Confidence threshold for teacher pseudo-labels; chosen by hand.
  • lambda_low, lambda_sem, lambda_c = 0.5, 1, 1e-4
    Upper-level loss weights for low-level reconstruction, semantic reconstruction, and compactness; chosen by hand.
  • N_l and N_u = 50 and 5
    Lower- and upper-level update steps in the alternating optimization; chosen by hand.
  • K = 100
    Number of superpixels; chosen by hand.
assumptions (6)
  • domain assumption The Superpixel Sampling Network (SSN) provides valid differentiable pixel-to-superpixel associations.
    The upper-level model relies on SSN [8] as an existing tool; its behavior is assumed correct for medical images.
  • domain assumption Mean Teacher EMA produces stable and useful pseudo-labels for self-training.
    The lower-level segmentation uses teacher pseudo-labels filtered by confidence; this semi-supervised assumption is standard but unproven in the few-shot scribble regime.
  • domain assumption Superpixel-wise dominant-class diffusion yields reliable dense pseudo-labels when filtered.
    The core pseudo-label propagation (Eqs. 5-6) assumes that a dominant seed class in a superpixel is representative of the whole region.
  • ad hoc to paper Target anatomical structures are located near the image center.
    The spatial-prior filtering (Eq. 9) and the center-dense superpixel initialization (Eqs. 13-14) encode a centrality prior that is not justified from anatomy and may fail on off-center organs.
  • domain assumption Dropping the implicit Jacobian term in the upper-level update is acceptable.
    Eqs. 22-23 discard the gradient through the lower-level optimum, citing prior work [10, 11]. The paper acknowledges bias but assumes the partial gradient works in practice.
  • domain assumption Unlabeled training slices carry learnable signal in this few-shot setting.
    The method treats 65 (ACDC) and 50 (Prostate) unlabeled training slices as useful for improving segmentation; this is a standard semi-supervised assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bi-Level Collaborative Learning for Few-Shot Scribble-Supervised Medical Image Segmentation." pith.science (2026). https://pith.science/paper/FNREE3ER

@misc{pith2026260725432,
  author       = {Pith},
  title        = {Pith review of: Bi-Level Collaborative Learning for Few-Shot Scribble-Supervised Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FNREE3ER}},
  note         = {Machine review of arXiv:2607.25432}
}
read the original abstract

Scribble annotations offer an efficient alternative to costly pixel-wise labeling for medical image segmentation, yet in real clinical scenarios, scribble-annotated samples are often still limited, imposing the dual challenges of sparse supervision and annotated sample scarcity. These compounded constraints severely deprive models of the structural evidence needed for complete region recovery and precise boundary delineation. To break this bottleneck, we propose a bi-level collaborative learning framework for few-shot scribble-supervised medical image segmentation. Specifically, an upper-level learnable superpixel model is introduced to provide region-structural priors for lower-level segmentation, while superpixel-based region-wise pseudo-label propagation and a spatial-prior-guided filtering strategy are performed to generate reliable dense pseudo-labels for segmentation learning. Meanwhile, the anatomical semantics learned by the lower-level segmentation model under the guidance of the current superpixels are fed back to the upper level, further driving it to learn region-structural representations better aligned with the segmentation task. Through bidirectional interaction and collaborative learning between the upper and lower levels, the proposed framework significantly outperforms existing state-of-the-art scribble-supervised methods on the ACDC and Prostate datasets under the few-shot scribble-supervised setting.

Figures

Figures reproduced from arXiv: 2607.25432 by the authors.

Figure 1
Figure 1. (a) Dual challenges in few-shot scribble-supervised segmentation: sparse supervision and limited annotated samples. (b) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of the proposed BiSCL framework. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of our method with state-of-the-art methods on the ACDC and Prostate test sets under the [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visual comparison of superpixel partitioning be [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of different ablation vari [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 2 linked inside Pith

  1. [1]

    Radhakrishna Achanta, Appu Shaji, Kevin Smith, Aurelien Lucchi, Pascal Fua, and Sabine Süsstrunk. 2012. SLIC superpixels compared to state-of-the-art superpixel methods.IEEE transactions on pattern analysis and machine intelligence34, 11 (2012), 2274–2282

    2012

  2. [2]

    Olivier Bernard, Alain Lalande, Clement Zotti, Frederick Cervenansky, Xin Yang, Pheng-Ann Heng, Irem Cetin, Karim Lekadir, Oscar Camara, Miguel Angel Gon- zalez Ballester, et al. 2018. Deep learning techniques for automatic MRI cardiac multi-structures segmentation and diagnosis: is the problem solved?IEEE trans- actions on medical imaging37, 11 (2018), 2514–2525

    2018

  3. [3]

    Xiaokang Chen, Yuhui Yuan, Gang Zeng, and Jingdong Wang. 2021. Semi- supervised semantic segmentation with cross pseudo supervision. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2613–2622

    2021

  4. [4]

    Kenneth Clark, Bruce Vendt, Kirk Smith, John Freymann, Justin Kirby, Paul Koppel, Stephen Moore, Stanley Phillips, David Maffitt, Michael Pringle, et al

  5. [5]

    Terrance DeVries and Graham W Taylor. 2017. Improved regularization of convolutional neural networks with cutout.arXiv preprint arXiv:1708.04552 (2017)

    arXiv 2017

  6. [6]

    Xin Fan, Xiaolin Wang, Jiaxin Gao, Jia Wang, Zhongxuan Luo, and Risheng Liu. 2024. Bi-level learning of task-specific decoders for joint registration and one-shot medical image segmentation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 11726–11735

    2024

  7. [7]

    Leo Grady. 2006. Random walks for image segmentation.IEEE transactions on pattern analysis and machine intelligence28, 11 (2006), 1768–1783

    2006

  8. [8]

    Varun Jampani, Deqing Sun, Ming-Yu Liu, Ming-Hsuan Yang, and Jan Kautz

Show all 45 references
  1. [9]

    Zhanghexuan Ji, Yan Shen, Chunwei Ma, and Mingchen Gao. 2019. Scribble- based hierarchical weakly supervised learning for brain tumor segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 175–183

    2019

  2. [10]

    Liuyuan Jiang, Quan Xiao, Lisha Chen, and Tianyi Chen. 2025. Beyond Value Functions: Single-Loop Bilevel Optimization under Flatness Conditions. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems

    2025

  3. [11]

    Liuyuan Jiang, Quan Xiao, Lisha Chen, and Tianyi Chen. 2025. Efficient Penalty- Based Bilevel Methods: Improved Analysis, Novel Updates, and Flatness Condi- tion.arXiv preprint arXiv:2511.16796(2025)

    2025

  4. [12]

    Shuailin Li, Zhitong Gao, and Xuming He. 2021. Superpixel-guided iterative learning from noisy labels for medical image segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 525–535

    2021

  5. [13]

    Shijie Li, Yunbin Tu, Qingyuan Xiang, and Zheng Li. 2024. MAGIC: Rethinking dynamic convolution design for medical image segmentation. InProceedings of the 32nd ACM International Conference on Multimedia. 9106–9115

    2024

  6. [14]

    Shiman Li, Jiayue Zhao, Shaolei Liu, Xiaokun Dai, Chenxi Zhang, and Zhijian Song. 2024. SP3: Superpixel-propagated pseudo-label learning for weakly semi- supervised medical image segmentation.arXiv preprint arXiv:2411.11636(2024)

    2024 arXiv

  7. [15]

    Zhixun Li, Jiancheng Fang, Ruiyun Qiu, and Huiling Gong. 2025. Scribble- supervised medical image segmentation based on dynamically generated pseudo labels via multi-scale superpixels.Biomedical Signal Processing and Control105 (2025), 107668

    2025

  8. [16]

    Zihan Li, Yuan Zheng, Xiangde Luo, Dandan Shan, and Qingqi Hong. 2023. ScribbleVC: Scribble-supervised medical image segmentation with vision-class embedding. InProceedings of the 31st ACM International Conference on Multimedia. 3384–3393

    2023

  9. [17]

    Zihan Li, Yuan Zheng, Dandan Shan, Shuzhou Yang, Qingde Li, Beizhan Wang, Yuanting Zhang, Qingqi Hong, and Dinggang Shen. 2024. Scribformer: Trans- former makes cnn work better for scribble-based medical image segmentation. IEEE Transactions on Medical Imaging43, 6 (2024), 2254–2265

    2024

  10. [18]

    Shiying Lin, Rong Hu, Zuoyong Li, Qinghua Lin, Jiawei Wu, and Changqing Zhang. 2025. Gradient-Aware Revitalization of Non-Effective Samples in Medical Image Segmentation. InProceedings of the 33rd ACM International Conference on Multimedia. 4728–4737

    2025

  11. [19]

    Risheng Liu, Jiaxin Gao, Jin Zhang, Deyu Meng, and Zhouchen Lin. 2021. Inves- tigating bi-level optimization for learning and vision from a unified perspective: A survey and beyond.IEEE Transactions on Pattern Analysis and Machine Intelli- gence44, 12 (2021), 10045–10067

    2021

  12. [20]

    Yong-Jin Liu, Cheng-Chi Yu, Min-Jing Yu, and Ying He. 2016. Manifold SLIC: A fast method to compute content-sensitive superpixels. InProceedings of the IEEE conference on computer vision and pattern recognition. 651–659

    2016

  13. [21]

    Ilya Loshchilov and Frank Hutter. 2019. Decoupled Weight Decay Regularization. InInternational Conference on Learning Representations

    2019

  14. [22]

    Xiangde Luo, Minhao Hu, Wenjun Liao, Shuwei Zhai, Tao Song, Guotai Wang, and Shaoting Zhang. 2022. Scribble-supervised medical image segmentation via dual-branch network and dynamically mixed pseudo labels supervision. In International Conference on Medical Image Computing and...

    2022

  15. [23]

    Xiangde Luo, Wenjun Liao, Jianghong Xiao, Jieneng Chen, Tao Song, Xiaofan Zhang, Kang Li, Dimitris N Metaxas, Guotai Wang, and Shaoting Zhang. 2022. WORD: A large scale dataset, benchmark and clinical applicable study for ab- dominal organ segmentation from CT image.Medical Im...

    2022

  16. [24]

    Thien B Nguyen-Tat, Hoang-An Vo, and Phuoc-Sang Dang. 2025. QMaxViT- Unet+: A query-based MaxViT-Unet with edge enhancement for scribble- supervised segmentation of medical images.Computers in Biology and Medicine 187 (2025), 109762

    2025

  17. [25]

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical image computing and computer-assisted intervention. Springer, 234–241

    2015

  18. [26]

    Nima Tajbakhsh, Laura Jeyaseelan, Qian Li, Jeffrey N Chiang, Zhihao Wu, and Xiaowei Ding. 2020. Embracing imperfect datasets: A review of deep learning solutions for medical image segmentation.Medical image analysis63 (2020), 101693

    2020

  19. [27]

    Meng Tang, Abdelaziz Djelouah, Federico Perazzi, Yuri Boykov, and Christopher Schroers. 2018. Normalized cut loss for weakly-supervised cnn segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition. 1818–1827

    2018

  20. [28]

    Antti Tarvainen and Harri Valpola. 2017. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.Advances in neural information processing systems30 (2017)

    2017

  21. [29]

    Bethany H Thompson, Gaetano Di Caterina, and Jeremy P Voisey. 2022. Pseudo- label refinement using superpixels for semi-supervised brain tumour segmen- tation. In2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI). IEEE, 1–5

    2022

  22. [30]

    Gabriele Valvano, Andrea Leo, and Sotirios A Tsaftaris. 2021. Learning to segment from scribbles using multi-scale adversarial attention gates.IEEE Transactions on Medical Imaging40, 8 (2021), 1990–2001

    2021

  23. [31]

    Michael Van den Bergh, Xavier Boix, Gemma Roig, Benjamin De Capitani, and Luc Van Gool. 2012. Seeds: Superpixels extracted via energy-driven sampling. In European conference on computer vision. Springer, 13–26

    2012

  24. [32]

    Jiacheng Wang, Xiaomeng Li, Yiming Han, Jing Qin, Liansheng Wang, and Zhou Qichao. 2022. Separated contrastive learning for organ-at-risk and gross- tumor-volume segmentation with limited annotation. InProceedings of the AAAI conference on artificial intelligence, Vol. 36. 2459–2467

    2022

  25. [33]

    Tao Wang, Xinlin Zhang, Zhenxuan Zhang, Yuanbo Zhou, Yuanbin Chen, Longx- uan Zhao, Chaohui Xu, Shun Chen, Guang Yang, and Tong Tong. 2024. Scrib- blevs: Scribble-supervised medical image segmentation via dynamic competitive pseudo label selection.arXiv preprint arXiv:2411.10237(2024)

    2024

  26. [34]

    Song Wu, Xiaoyu Wei, Xinyue Chen, Yazhou Ren, Jing He, and Xiaorong Pu. 2024. Cross-view mutual learning for semi-supervised medical image segmentation. In Proceedings of the 32nd ACM International Conference on Multimedia. 9253–9261

    2024

  27. [35]

    Sen Xu, Shikui Wei, Tao Ruan, and Yao Zhao. 2023. ESNet: An efficient framework for superpixel segmentation.IEEE Transactions on Circuits and Systems for Video Technology34, 7 (2023), 5389–5399

    2023

  28. [36]

    Fengting Yang, Qian Sun, Hailin Jin, and Zihan Zhou. 2020. Superpixel segmenta- tion with fully convolutional networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 13964–13973. Su et al

    2020

  29. [37]

    Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Chun, Junsuk Choe, and Youngjoon Yoo. 2019. Cutmix: Regularization strategy to train strong classifiers with localizable features. InProceedings of the IEEE/CVF international conference on computer vision. 6023–6032

    2019

  30. [38]

    Ke Zhang and Vishal M Patel. 2024. Modelmix: A new model-mixup strategy to minimize vicinal risk across tasks for few-scribble based cardiac segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 456–466

    2024

  31. [39]

    Ke Zhang and Xiahai Zhuang. 2022. Cyclemix: A holistic strategy for medical image segmentation from scribble supervision. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 11656–11665

    2022

  32. [40]

    Xiao Zhang, Shaoxuan Wu, Peilin Zhang, Zhuo Jin, Xiaosong Xiong, Qirong Bu, Jingkun Chen, and Jun Feng. 2025. HELPNet: Hierarchical perturbations consistency and entropy-guided ensemble for scribble supervised medical image segmentation.Medical Image Analysis105 (2025), 103719

    2025

  33. [41]

    Yihua Zhang, Prashant Khanduri, Ioannis Tsaknakis, Yuguang Yao, Mingyi Hong, and Sijia Liu. 2024. An Introduction to Bilevel Optimization: Foundations and applications in signal processing and machine learning.IEEE Signal Processing Magazine41, 1 (2024), 38–59

    2024

  34. [42]

    Amy Zhao, Guha Balakrishnan, Fredo Durand, John V Guttag, and Adrian V Dalca. 2019. Data augmentation using learned transformations for one-shot med- ical image segmentation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 8543–8553

    2019

  35. [43]

    Meng Zhou, Zhe Xu, Kang Zhou, and Raymond Kai-yu Tong. 2023. Weakly supervised medical image segmentation via superpixel-guided scribble walking and class-wise contrastive regularization. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention....

    2023

  36. [2013]

    The Cancer Imaging Archive (TCIA): maintaining and operating a public information repository.Journal of digital imaging26, 6 (2013), 1045–1057

    2013

  37. [2018]

    InProceedings of the European conference on computer vision (ECCV)

    Superpixel sampling networks. InProceedings of the European conference on computer vision (ECCV). 352–368

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.