pith. sign in

arxiv: 2606.01753 · v1 · pith:QWQLX2JOnew · submitted 2026-06-01 · 💻 cs.CV

Quality-Guided Semi-Supervised Learning for Medical Image Segmentation

Kumar Abhishek , Ghassan Hamarneh This is my paper

Pith reviewed 2026-06-28 15:42 UTC · model grok-4.3

classification 💻 cs.CV
keywords semi-supervised learningmedical image segmentationquality estimationpseudolabelingregularizationdeep learningmask quality
0
0 comments X

The pith

A quality predictor trained on corrupted and partial masks improves semi-supervised medical image segmentation when used for regularization and pseudolabel reweighting.

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

The paper introduces a quality predictor network that learns to score how good a segmentation mask is for a given image. This score replaces reliance on model confidence when deciding how to use unlabeled data in semi-supervised training. The predictor is built by showing it many imperfect masks created through controlled corruptions plus outputs from models that are only partly trained. These two mechanisms, a quality-aware loss term and reweighting of pseudolabels by their predicted quality, are added to standard SSL pipelines. The result is presented as a modular addition that raises performance on medical segmentation tasks without changing the underlying learner.

Core claim

The paper claims that a dedicated quality estimation network, trained on image-mask pairs whose masks come from synthetic corruptions and from checkpoints of incompletely trained segmenters, can be inserted into any existing semi-supervised segmentation pipeline through a quality-regularized loss and quality-weighted pseudolabel selection, producing higher accuracy than confidence-based or uncertainty-based SSL baselines on multiple medical datasets.

What carries the argument

The quality predictor network that maps an image and its candidate mask to a scalar quality score, trained on variable-quality masks generated by synthetic corruptions plus partial-model outputs.

If this is right

  • The quality-guided components act as a drop-in addition to existing SSL frameworks without altering their core training loop.
  • Performance gains appear consistently across five different medical image datasets.
  • The same quality predictor works with several different segmentation network architectures.
  • The approach raises accuracy beyond competing SSL methods that rely only on model confidence or uncertainty.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same quality-prediction idea could be tested in semi-supervised tasks outside medical imaging, such as natural scene segmentation or object detection.
  • If the quality scores prove reliable early in training, the method might allow SSL to succeed with even smaller labeled sets than currently demonstrated.
  • The quality predictor itself could be used to decide which unlabeled examples deserve human annotation, turning the framework toward active learning.

Load-bearing premise

Masks created by synthetic corruptions and by partially trained models produce error patterns that are representative enough for the quality predictor to judge real masks generated during SSL training.

What would settle it

An experiment in which the quality predictor's scores show low or negative correlation with actual segmentation metrics such as Dice on held-out labeled data, or in which adding the quality guidance produces no accuracy gain over the unmodified SSL baseline.

Figures

Figures reproduced from arXiv: 2606.01753 by Ghassan Hamarneh, Kumar Abhishek.

Figure 1
Figure 1. Figure 1: An overview of the proposed quality-guided semi-supervised segmentation methods, along with the scope of experiments present in this paper. 2 Method Our proposed approach has two phases: (Phase 1) training a quality predictor gϕ to estimate segmentation quality from image-mask pairs, and (Phase 2) using the frozen gϕ to guide semi-supervised segmentation training [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Scatter plot of the segmentation predictions’ Dice DSC(y, yˆ) and the corre￾sponding predicted quality estimates gϕ(x, yˆ) on the test set of CLI dataset, and four representative images (A-D) with the ground truth (green) and predicted (red) segmen￾tations. We observe a strong, stat. sig., positive linear correlation (ρ=0.69; p=1e-314). over 3 runs with different seeds. We compare QAR (Sec. 2.4 A) against … view at source ↗
read the original abstract

Training accurate medical image segmentation models requires large amounts of densely annotated data, which is costly and time-consuming to obtain. Semi-supervised learning (SSL) alleviates this by learning from both abundant unlabeled data and limited labeled data. However, most modern SSL methods rely on pseudolabels for unlabeled data, and typically assess their reliability through model confidence or uncertainty, measures that are self-referential and lack explicit grounding in segmentation quality. Instead, we propose a quality-guided SSL framework that trains a dedicated network to estimate segmentation quality from image-mask pairs. The predictor is trained on variable-quality masks generated through synthetic corruptions augmented with imperfect outputs from partially trained segmentation models, capturing realistic error patterns encountered during training. We integrate the quality predictor into SSL through two complementary mechanisms: a quality-aware regularization loss and a quality-based pseudolabel sample reweighting scheme. We show that our method serves as a drop-in enhancement to existing SSL frameworks. Extensive experiments across five datasets and multiple architectures demonstrate consistent improvements over competing SSL methods, advancing the state-of-the-art in semi-supervised medical image segmentation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a quality-guided SSL framework for medical image segmentation. A dedicated quality predictor network is trained on variable-quality masks created via synthetic corruptions plus outputs from partially trained segmentation models. This predictor is integrated into existing SSL methods through a quality-aware regularization loss and quality-based pseudolabel reweighting. The method is presented as a drop-in enhancement, with experiments across five datasets and multiple architectures showing consistent gains over competing SSL approaches.

Significance. If the quality estimates prove reliable on actual SSL pseudolabels, the approach offers an explicit, non-self-referential alternative to confidence- or uncertainty-based weighting, which could improve robustness in annotation-scarce medical imaging. The multi-dataset, multi-architecture evaluation and drop-in design are positive features that would support broader adoption if the core assumption holds.

major comments (2)
  1. [Methods (quality predictor training subsection)] The central claim that the quality predictor produces scores that meaningfully reflect true segmentation quality during SSL training rests on the training data construction (synthetic corruptions + partial-model outputs). No direct validation is provided that the resulting error patterns, spatial distributions, or severity match those of pseudolabels generated in the full SSL loop; without such a calibration or distribution-comparison experiment, gains cannot be confidently attributed to the proposed guidance mechanisms rather than incidental regularization effects.
  2. [Experiments and results] In the experimental results, consistent improvements are reported over baselines, but the paper does not include an ablation isolating the contribution of the quality predictor's accuracy (e.g., by replacing it with random or oracle scores) or statistical significance tests across the five datasets. This leaves open whether the reported SOTA advances are load-bearing on the quality estimates or on other implementation details.
minor comments (2)
  1. [Method] Notation for the quality score q and its integration into the loss (e.g., the exact form of the quality-aware regularization term) should be defined with an equation number for clarity.
  2. [Experiments] Figure captions for the qualitative results should explicitly state the dataset, architecture, and labeled-data ratio used in each panel to aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We address each major comment below, agreeing where the manuscript can be strengthened through revisions.

read point-by-point responses

  1. Referee: [Methods (quality predictor training subsection)] The central claim that the quality predictor produces scores that meaningfully reflect true segmentation quality during SSL training rests on the training data construction (synthetic corruptions + partial-model outputs). No direct validation is provided that the resulting error patterns, spatial distributions, or severity match those of pseudolabels generated in the full SSL loop; without such a calibration or distribution-comparison experiment, gains cannot be confidently attributed to the proposed guidance mechanisms rather than incidental regularization effects.

    Authors: We acknowledge that an explicit comparison of error patterns (types, spatial distributions, and severity) between the quality predictor's training data and actual pseudolabels from the SSL training loop would provide stronger support for the claim. Our construction using synthetic corruptions and partial-model outputs is intended to capture common segmentation error modes, but we agree a direct calibration experiment would reduce ambiguity regarding attribution of gains. In the revised manuscript, we will add such an analysis, including quantitative and qualitative comparisons of error distributions. revision: yes

  2. Referee: [Experiments and results] In the experimental results, consistent improvements are reported over baselines, but the paper does not include an ablation isolating the contribution of the quality predictor's accuracy (e.g., by replacing it with random or oracle scores) or statistical significance tests across the five datasets. This leaves open whether the reported SOTA advances are load-bearing on the quality estimates or on other implementation details.

    Authors: We agree that an ablation study replacing the quality predictor with random scores (or an oracle) would more clearly isolate its contribution, and that statistical significance testing across datasets would strengthen the empirical claims. These elements were not included in the original submission. In the revised manuscript, we will add the requested ablations and report statistical significance (e.g., paired t-tests or Wilcoxon tests with p-values) for the main results. revision: yes

Circularity Check

0 steps flagged

No circularity; quality predictor trained independently

full rationale

The paper trains a dedicated quality network on masks generated via synthetic corruptions plus outputs from partially trained models, then integrates the resulting scores into existing SSL frameworks via a regularization loss and reweighting. This training occurs prior to and independently of the target SSL loop. Central claims rest on empirical results across five datasets and multiple architectures rather than any definitional loop, fitted parameter renamed as prediction, or load-bearing self-citation. No equations or steps reduce the claimed improvements to the inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim depends on the effectiveness of the quality predictor, which rests on the domain assumption about error patterns and possibly several hyperparameters not detailed in the abstract.

free parameters (1)
  • Quality loss weight
    Likely a hyperparameter balancing the quality regularization loss, though not specified in abstract.
axioms (1)
  • domain assumption Synthetic corruptions and partial model outputs produce error patterns representative of those during actual training.
    This is invoked to justify the training of the quality predictor.
invented entities (1)
  • Dedicated quality predictor network no independent evidence
    purpose: To estimate segmentation quality independently of model confidence.
    New component introduced in the framework.

pith-pipeline@v0.9.1-grok · 5715 in / 1341 out tokens · 32918 ms · 2026-06-28T15:42:43.850344+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 26 canonical work pages

  1. [1]

    Arti- ficial intelligence review54(1), 137–178 (Jun 2021).https://doi.org/10.1007/ s10462-020-09854-1

    Asgari Taghanaki, S., Abhishek, K., Cohen, J.P., Cohen-Adad, J., Hamarneh, G.: Deep semantic segmentation of natural and medical images: a review. Arti- ficial intelligence review54(1), 137–178 (Jun 2021).https://doi.org/10.1007/ s10462-020-09854-1

    2021

  2. [2]

    In: Celebi, M.E., Schaefer, G

    Ballerini, L., Fisher, R.B., Aldridge, B., Rees, J.: A Color and Texture Based Hier- archical K-NN Approach to the Classification of Non-melanoma Skin Lesions. In: Celebi, M.E., Schaefer, G. (eds.) Color Medical Image Analysis. vol. 6, pp. 63–86. Springer Netherlands (2013).https://doi.org/10.1007/978-94-007-5389-1_4

    work page doi:10.1007/978-94-007-5389-1_4 2013

  3. [3]

    saliency maps from physicians

    Bernal, J., Sánchez, F.J., Fernández-Esparrach, G., Gil, D., Rodríguez, C., Vilar- iño, F.: Wm-dova maps for accurate polyp highlighting in colonoscopy: Validation vs. saliency maps from physicians. Computerized Medical Imaging and Graphics 43, 99–111 (Jul 2015).https://doi.org/10.1016/j.compmedimag.2015.02.007

    work page doi:10.1016/j.compmedimag.2015.02.007 2015

  4. [4]

    Abhishek et al

    Bernal, J., Sánchez, J., Vilarino, F.: Towards automatic polyp detection with a polypappearancemodel.PatternRecognition45(9),3166–3182(Sep2012).https: //doi.org/10.1016/j.patcog.2012.03.002 10 K. Abhishek et al

    work page doi:10.1016/j.patcog.2012.03.002 2012

  5. [5]

    In: Statistical Atlases and Computational Models of the Heart

    Byrne, N., Clough, J.R., Montana, G., King, A.P.: A persistent homology-based topological loss function for multi-class CNN segmentation of cardiac MRI. In: Statistical Atlases and Computational Models of the Heart. M&Ms and EMIDEC Challenges. p. 3–13. Springer International Publishing (2021).https://doi.org/ 10.1007/978-3-030-68107-4_1

    work page doi:10.1007/978-3-030-68107-4_1 2021

  6. [6]

    In: European Conference on Computer Vision

    Cao, H., Wang, Y., Chen, J., Jiang, D., Zhang, X., Tian, Q., Wang, M.: Swin- Unet: Unet-like pure transformer for medical image segmentation. In: European Conference on Computer Vision. pp. 205–218. Springer (2022).https://doi.org/ 10.1007/978-3-031-25066-8_9

    work page doi:10.1007/978-3-031-25066-8_9 2022

  7. [7]

    Nobile, Paolo Cazzaniga, Daniela Besozzi, Riccardo Colombo, Gi- ancarlo Mauri, and Gabriella Pasi

    Chaitanya, K., Erdil, E., Karani, N., Konukoglu, E.: Local contrastive loss with pseudo-label based self-training for semi-supervised medical image segmentation. Medical Image Analysis87, 102792 (Jul 2023).https://doi.org/10.1016/j. media.2023.102792

    work page doi:10.1016/j 2023

  8. [8]

    Journal of Imaging8(5), 121 (Apr 2022).https://doi.org/10.3390/jimaging8050121

    Chen, S., Urban, G., Baldi, P.: Weakly supervised polyp segmentation in colonoscopy images using deep neural networks. Journal of Imaging8(5), 121 (Apr 2022).https://doi.org/10.3390/jimaging8050121

    work page doi:10.3390/jimaging8050121 2022

  9. [9]

    In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Chen, X., Yuan, Y., Zeng, G., Wang, J.: Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 2613–2622. IEEE, Nashville, TN, USA (Jun 2021).https://doi.org/10.1109/CVPR46437.2021.00264

    work page doi:10.1109/cvpr46437.2021.00264 2021

  10. [10]

    arXiv preprint arXiv:1807.00502 pp

    DeVries, T., Taylor, G.W.: Leveraging uncertainty estimates for predicting seg- mentation quality. arXiv preprint arXiv:1807.00502 pp. 1–9 (Jul 2018)

    Pith/arXiv arXiv 2018

  11. [11]

    In: Medical Image Com- puting and Computer Assisted Intervention – MICCAI 2023

    Du, S., Bayasi, N., Hamarneh, G., Garbi, R.: MDViT: Multi-domain vision trans- former for small medical image segmentation datasets. In: Medical Image Com- puting and Computer Assisted Intervention – MICCAI 2023. p. 448–458. Springer Nature Switzerland (2023).https://doi.org/10.1007/978-3-031-43901-8_43

    work page doi:10.1007/978-3-031-43901-8_43 2023

  12. [12]

    Cancers16(6), 1120 (Mar 2024).https://doi.org/10

    Dzieniszewska, A., Garbat, P., Piramidowicz, R.: Improving skin lesion segmen- tation with self-training. Cancers16(6), 1120 (Mar 2024).https://doi.org/10. 3390/cancers16061120

    2024

  13. [13]

    In: IEEE/CVF International Conference on Computer Vision (ICCV), pp

    Girshick, R.: Fast R-CNN. In: 2015 IEEE International Conference on Computer Vision (ICCV). p. 1440–1448. IEEE (Dec 2015).https://doi.org/10.1109/iccv. 2015.169

    work page doi:10.1109/iccv 2015

  14. [14]

    IEEE Transactions on Biomedical Engineering60(7), 1873–1883 (Jul 2013)

    Glaister, J., Amelard, R., Wong, A., Clausi, D.A.: MSIM: Multistage illumination modeling of dermatological photographs for illumination-corrected skin lesion anal- ysis. IEEE Transactions on Biomedical Engineering60(7), 1873–1883 (Jul 2013). https://doi.org/10.1109/tbme.2013.2244596

    work page doi:10.1109/tbme.2013.2244596 2013

  15. [15]

    In: International Conference on Machine Learning

    Guo, C., Pleiss, G., Sun, Y., Weinberger, K.Q.: On calibration of modern neu- ral networks. In: International Conference on Machine Learning. pp. 1321–1330. PMLR (2017)

    2017

  16. [16]

    Nature Methods18(2), 203–211 (Feb 2021).https://doi.org/10.1038/ s41592-020-01008-z

    Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H.: nnU-Net: a self-configuring method for deep learning-based biomedical image segmenta- tion. Nature Methods18(2), 203–211 (Feb 2021).https://doi.org/10.1038/ s41592-020-01008-z

    2021

  17. [17]

    In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).p.9796–9805.IEEE(Jun2020).https://doi.org/10.1109/ cvpr42600.2020.00982

    Kirillov, A., Wu, Y., He, K., Girshick, R.: PointRend: Image segmentation as rendering. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).p.9796–9805.IEEE(Jun2020).https://doi.org/10.1109/ cvpr42600.2020.00982

    arXiv 2020

  18. [18]

    Kohlberger, T., Singh, V., Alvino, C., Bahlmann, C., Grady, L.: Evaluating Seg- mentation Error without Ground Truth. In: Hutchison, D., Kanade, T., Kit- tler, J., Kleinberg, J.M., Mattern, F., Mitchell, J.C., Naor, M., Nierstrasz, O., Quality-Guided Semi-Supervised Learning for Medical Image Segmentation 11 Pandu Rangan, C., Steffen, B., Sudan, M., Terzo...

    work page doi:10.1007/978-3-642-33415-3_65 2012

  19. [19]

    In: ICML 2013 Workshop on Challenges in Representation Learning

    Lee, D.H.: Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks. In: ICML 2013 Workshop on Challenges in Representation Learning. vol. 3 (2), pp. 1–6. Atlanta, Georgia, USA (Jun 2013)

    2013

  20. [20]

    PeerJ Computer Science10, e1935 (Mar 2024).https: //doi.org/10.7717/peerj-cs.1935

    Li, Y., Tian, T., Hu, J., Yuan, C.: SUTrans-NET: a hybrid transformer approach to skin lesion segmentation. PeerJ Computer Science10, e1935 (Mar 2024).https: //doi.org/10.7717/peerj-cs.1935

    work page doi:10.7717/peerj-cs.1935 2024

  21. [21]

    Future Generation Computer Sys- tems177, 108228 (Apr 2026).https://doi.org/10.1016/j.future.2025.108228

    Liang, P., Chen, J., Wu, Y., Wu, R., Chen, Z., Pu, B., Chang, Q., Ran, G.: A multilevel alignment and cross-fusion knowledge distillation framework for vision transformer-based medical image segmentation. Future Generation Computer Sys- tems177, 108228 (Apr 2026).https://doi.org/10.1016/j.future.2025.108228

    work page doi:10.1016/j.future.2025.108228 2026

  22. [22]

    Medical Image Analysis42, 60–88 (Dec 2017).https: //doi.org/10.1016/j.media.2017.07.005

    Litjens, G., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., Van Der Laak, J.A., Van Ginneken, B., Sánchez, C.I.: A survey on deep learning in medical image analysis. Medical Image Analysis42, 60–88 (Dec 2017).https: //doi.org/10.1016/j.media.2017.07.005

    work page doi:10.1016/j.media.2017.07.005 2017

  23. [23]

    In: International Conferenceon LearningRepresentations(2019),https://openreview.net/forum? id=Bkg6RiCqY7

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. In: International Conferenceon LearningRepresentations(2019),https://openreview.net/forum? id=Bkg6RiCqY7

    2019

  24. [24]

    IEEE Transactions on Medical Imaging39(12), 3868–3878 (Dec 2020).https://doi.org/10.1109/tmi.2020.3006437

    Mehrtash, A., Wells, W.M., Tempany, C.M., Abolmaesumi, P., Kapur, T.: Con- fidence calibration and predictive uncertainty estimation for deep medical image segmentation. IEEE Transactions on Medical Imaging39(12), 3868–3878 (Dec 2020).https://doi.org/10.1109/tmi.2020.3006437

    work page doi:10.1109/tmi.2020.3006437 2020

  25. [25]

    In: IEEE Engineering in Medicine and Biology Society

    Mendonça, T., Ferreira, P.M., Marques, J.S., Marcal, A.R.S., Rozeira, J.: PH2 - A dermoscopic image database for research and benchmarking. In: IEEE Engineering in Medicine and Biology Society. pp. 5437–5440 (Jul 2013).https://doi.org/10. 1109/embc.2013.6610779

    arXiv 2013

  26. [26]

    In: Medical Imaging with Deep Learning (2018),https://openreview.net/forum?id=Skft7cijM

    Oktay, O., Schlemper, J., Folgoc, L.L., Lee, M., Heinrich, M., Misawa, K., Mori, K., McDonagh, S., Hammerla, N.Y., Kainz, B., Glocker, B., Rueckert, D.: Attention u-net: Learning where to look for the pancreas. In: Medical Imaging with Deep Learning (2018),https://openreview.net/forum?id=Skft7cijM

    2018

  27. [27]

    In: Computer Vision – ECCV 2024 Workshops

    Perera, S., Erzurumlu, Y., Gulati, D., Yilmaz, A.: MobileUNETR: A lightweight end-to-end hybrid vision transformer for efficient medical image segmentation. In: Computer Vision – ECCV 2024 Workshops. p. 281–299. Springer Nature Switzer- land (2025).https://doi.org/10.1007/978-3-031-91721-9_18

    work page doi:10.1007/978-3-031-91721-9_18 2024

  28. [28]

    In: Greenspan, H., Madabhushi, A., Mousavi, P., Salcudean, S., Duncan, J., Syeda-Mahmood, T., Taylor, R

    Qiu, P., Chakrabarty, S., Nguyen, P., Ghosh, S.S., Sotiras, A.: QCResUNet: Joint Subject-Level and Voxel-Level Prediction of Segmentation Quality. In: Greenspan, H., Madabhushi, A., Mousavi, P., Salcudean, S., Duncan, J., Syeda-Mahmood, T., Taylor, R. (eds.) Medical Image Computing and Computer Assisted Intervention – MICCAI2023.vol.14223,pp.173–182.Sprin...

    work page doi:10.1007/978-3-031-43901-8_17 2023

  29. [29]

    In: International Conference on Machine Learning

    Ren,M.,Zeng,W.,Yang,B.,Urtasun,R.:Learningtoreweightexamplesforrobust deep learning. In: International Conference on Machine Learning. pp. 4334–4343. PMLR (2018)

    2018

  30. [30]

    Abhishek et al

    Robinson, R., Oktay, O., Bai, W., Valindria, V.V., Sanghvi, M.M., Aung, N., Paiva, J.M., Zemrak, F., Fung, K., Lukaschuk, E., et al.: Real-time prediction 12 K. Abhishek et al. of segmentation quality. In: International Conference on Medical Image Comput- ing and Computer-Assisted Intervention. pp. 578–585. Springer (2018).https: //doi.org/10.1007/978-3-0...

    work page doi:10.1007/978-3-030-00937-3_66 2018

  31. [31]

    In: International Conference on Medical image com- puting and computer-assisted intervention

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomed- ical image segmentation. In: International Conference on Medical image com- puting and computer-assisted intervention. pp. 234–241. Springer (2015).https: //doi.org/10.1007/978-3-319-24574-4_28

    work page doi:10.1007/978-3-319-24574-4_28 2015

  32. [32]

    Scientific Data8(1), 34 (Jan 2021).https://doi.org/10.1038/ s41597-021-00815-z

    Rotemberg, V., Kurtansky, N., Betz-Stablein, B., Caffery, L., Chousakos, E., Codella, N., Combalia, M., Dusza, S., Guitera, P., Gutman, D., Halpern, A., Helba, B., Kittler, H., Kose, K., Langer, S., Lioprys, K., Malvehy, J., Musthaq, S., Nanda, J., Reiter, O., Shih, G., Stratigos, A., Tschandl, P., Weber, J., Soyer, H.P.: A patient-centric dataset of imag...

    2021

  33. [33]

    arXiv preprint arXiv:2409.14874 pp

    Senbi,A.,Huang,T.,Lyu,F.,Li,Q.,Tao,Y.,Shao,W.,Chen,Q.,Wang,C.,Wang, S., Zhou, T., Zhang, Y.: Towards ground-truth-free evaluation of any segmentation in medical images. arXiv preprint arXiv:2409.14874 pp. 1–17 (2024)

    arXiv 2024

  34. [34]

    Advances in Neural Informa- tion Processing Systems32(2019)

    Shu, J., Xie, Q., Yi, L., Zhao, Q., Zhou, S., Xu, Z., Meng, D.: Meta-weight-net: Learning an explicit mapping for sample weighting. Advances in Neural Informa- tion Processing Systems32(2019)

    2019

  35. [35]

    Advances in Neural Information Processing Systems 33, 596–608 (2020)

    Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C.A., Cubuk, E.D., Kurakin, A., Li, C.L.: FixMatch: Simplifying semi-supervised learning with consistency and confidence. Advances in Neural Information Processing Systems 33, 596–608 (2020)

    2020

  36. [36]

    Information Fusion131, 104182 (Jul 2026).https: //doi.org/10.1016/j.inffus.2026.104182

    Song, P., Wang, Z., Zhang, J., Fu, S., Zhang, Y., Wu, W., Bao, F.: SU-RMT: Toward bridging semantic representation and structural detail modeling for med- ical image segmentation. Information Fusion131, 104182 (Jul 2026).https: //doi.org/10.1016/j.inffus.2026.104182

    work page doi:10.1016/j.inffus.2026.104182 2026

  37. [37]

    Computerized Medical Imaging and Graphics75, 24–33 (Jul 2019).https://doi.org/10.1016/j.compmedimag.2019.04.005

    Taghanaki, S.A., Zheng, Y., Zhou, S.K., Georgescu, B., Sharma, P., Xu, D., Co- maniciu, D., Hamarneh, G.: Combo loss: Handling input and output imbalance in multi-organ segmentation. Computerized Medical Imaging and Graphics75, 24–33 (Jul 2019).https://doi.org/10.1016/j.compmedimag.2019.04.005

    work page doi:10.1016/j.compmedimag.2019.04.005 2019

  38. [38]

    Medical image analysis63, 101693 (Jul 2020).https://doi.org/10.1016/ j.media.2020.101693

    Tajbakhsh, N., Jeyaseelan, L., Li, Q., Chiang, J.N., Wu, Z., Ding, X.: Embracing imperfect datasets: A review of deep learning solutions for medical image segmen- tation. Medical image analysis63, 101693 (Jul 2020).https://doi.org/10.1016/ j.media.2020.101693

    arXiv 2020

  39. [39]

    In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R

    Tarvainen, A., Valpola, H.: Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Advances in Neural Information Processing Systems. vol. 30. Curran Associates, Inc. (2017)

    2017

  40. [40]

    IEEE Transactions on Medi- cal Imaging36(8), 1597–1606 (Aug 2017).https://doi.org/10.1109/TMI.2017

    Valindria, V.V., Lavdas, I., Bai, W., Kamnitsas, K., Aboagye, E.O., Rockall, A.G., Rueckert, D., Glocker, B.: Reverse Classification Accuracy: Predicting Segmenta- tion Performance in the Absence of Ground Truth. IEEE Transactions on Medi- cal Imaging36(8), 1597–1606 (Aug 2017).https://doi.org/10.1109/TMI.2017. 2665165

    work page doi:10.1109/tmi.2017 2017

  41. [41]

    Neural Networks145, 90–106 (Jan 2022).https://doi.org/10.1016/j.neunet.2021

    Verma, V., Kawaguchi, K., Lamb, A., Kannala, J., Solin, A., Bengio, Y., Lopez- Paz, D.: Interpolation consistency training for semi-supervised learning. Neural Networks145, 90–106 (Jan 2022).https://doi.org/10.1016/j.neunet.2021. 10.008 Quality-Guided Semi-Supervised Learning for Medical Image Segmentation 13

    work page doi:10.1016/j.neunet.2021 2022

  42. [42]

    In: Shen, D., Liu, T., Peters, T.M., Staib, L.H., Essert, C., Zhou, S., Yap, P.T., Khan, A

    Yu, L., Wang, S., Li, X., Fu, C.W., Heng, P.A.: Uncertainty-Aware Self-ensembling Model for Semi-supervised 3D Left Atrium Segmentation. In: Shen, D., Liu, T., Peters, T.M., Staib, L.H., Essert, C., Zhou, S., Yap, P.T., Khan, A. (eds.) Medical ImageComputingandComputerAssistedIntervention–MICCAI2019.vol.11765, pp. 605–613. Springer International Publishin...

    work page doi:10.1007/978-3-030-32245-8_67 2019

  43. [43]

    In: Machine Learning in Medical Imaging

    Zheng, Z., Wang, X., Zhang, X., Zhong, Y., Yao, X., Zhang, Y., Wang, Y.: Semi-supervised segmentation with self-training based on quality estimation and refinement. In: Machine Learning in Medical Imaging. p. 30–39. Springer In- ternational Publishing (2020).https://doi.org/10.1007/978-3-030-59861-7_4, http://dx.doi.org/10.1007/978-3-030-59861-7_4

    work page doi:10.1007/978-3-030-59861-7_4 2020

  44. [44]

    IEEE Transac- tions on Medical Imaging39(6), 1856–1867 (2019).https://doi.org/10.1109/ tmi.2019.2959609

    Zhou, Z., Siddiquee, M.M.R., Tajbakhsh, N., Liang, J.: Unet++: Redesigning skip connections to exploit multiscale features in image segmentation. IEEE Transac- tions on Medical Imaging39(6), 1856–1867 (2019).https://doi.org/10.1109/ tmi.2019.2959609

    arXiv 2019