Pith. sign in

REVIEW 4 major objections 6 minor 43 references

SWDL: Stratum-Wise Difference Learning with Deep Laplacian Pyramid for Semi-Supervised 3D Intracranial Hemorrhage Segmentation

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A dual-decoder network segments brain hemorrhages with just 2% labeled CT scans, reaching 89.32% Dice.

desk verdict New architecture and ablations are real, but test-set hyperparameter tuning and a CI copy-paste error undermine the SOTA claim. read the letter →

arxiv 2506.10325 v1 pith:Y6NPTH3G submitted 2025-06-12 eess.IV cs.CVcs.LG

Cheng Wang , Siqi Chen , Donghua Mi , Yang Chen , Yudong Zhang , Yinsheng Li This is my paper · ORCID
classification eess.IVcs.CVcs.LG
keywords intracranialhemorrhagesegmentationsemi-supervisedlearningLaplacianpyramiddeepconvolutionalupsamplingdifference3DCTdual-decoderlow-labelregime
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 proposes SWDL-Net, a semi-supervised framework for 3D intracranial hemorrhage (ICH) segmentation that works with very few labeled CT volumes. The method trains two decoders—one using deep transposed convolutions and one using Laplacian pyramid upsampling—and feeds the stratum-wise difference between their features back into the encoder. The authors claim this difference learning lets the model use abundant unlabeled data effectively, outperforming existing semi-supervised methods (UAMT, URPC, LeFeD) on a 271-case clinical dataset with only 2% labels and on the public BHSD benchmark with 5% labels. If correct, the approach offers a practical path to accurate ICH segmentation in settings where expert annotations are scarce.

What carries the argument

The key mechanism is 'stratum-wise difference learning' (SWDL) combined with a dual-decoder architecture: a deep convolutional (DC) decoder and a Deep Laplacian Pyramid Upsampling (DelPU) decoder. At each iteration, the difference $\Delta^{(s,p)}$ between the two decoders' feature maps at stratum $s$ is computed and added into the encoder's next-iteration feature computation via $y^s_{E,p} = f^s_{\theta_E}(y^{s-1}_{E,p} + \xi \Delta^{(s-1,p-1)})$. The DelPU decoder builds a Laplacian pyramid with Gaussian downsampling, inter-level differential extraction, and weighted reconstruction governed by an edge-sharpening hyperparameter $\mu$. A stratum-adaptive rule chooses pyramid depth based on input dimension. The unsupervised loss is MSE between the two decoders' outputs, while the supervised loss combines Dice, cross-entropy, and deep supervision on the DC path.

What would settle it

A concrete test would be to replace the difference term in Equation 2 with random Gaussian noise of the same magnitude (or with a fixed non-informative constant) and measure whether Dice on the PBGTH 2% setting drops significantly; if it does not, the specific difference signal is not responsible for the reported gain.

Watch

Extended reading notes

Core claim

SWDL-Net establishes that a hybrid dual-decoder architecture with stratum-wise difference learning can achieve near-fully-supervised performance in low-label regimes. With only 2% labeled data on the private PBGTH dataset, it attains Dice of 89.32% versus 92.48% for fully supervised VNet, i.e., 96.6% of the fully supervised score, and improves HD95 by 41.3–66.9% over the compared SSL baselines. On the public BHSD dataset, with 5% labeled data, it reaches 51.81% Dice, out-performing LeFeD by 1.11 points. The central claim is that feeding the difference between the DC decoder and DelPU decoder features back into the encoder (Equation 2) provides a supervisory signal that leverages the Laplacian pyramid's edge-sharpening and the convolution path's detail precision.

Load-bearing premise

The load-bearing premise is that the difference between the two decoders' features is a meaningful, learnable signal for unlabeled data, rather than noise that simply adds variance to the encoder's features.

Editorial extensions

If this is right

  • If the results hold, semi-supervised ICH segmentation can be performed accurately with only a handful of labeled scans (2–5%), greatly reducing annotation cost in emergency neuroimaging.
  • The difference-learning mechanism between Laplacian-pyramid and convolutional upsampling decoders suggests that other anisotropic or edge-critical medical segmentation tasks (e.g., vessel or tumor boundary delineation) could benefit from the same dual-decoder discrepancy signal.
  • The reported Dice of 89.32% on PBGTH and 51.81% on BHSD with so few labels indicates that the method could serve as a strong baseline for future SSL work on volumetric CT segmentation.
  • Because inference requires only the encoder and primary decoder, the method could be deployed in clinical workflows without the computational overhead of the pyramid module.

Reading between the lines

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

  • A natural testable extension is to apply the same stratum-wise difference feedback to other multi-decoder consistency frameworks (e.g., different upsampling kernels or transformer-based decoders) to see whether the benefit is specific to Laplacian-vs-convolutional diversity or generic to any two sufficiently different decoders.
  • The paper does not isolate how much of the gain comes from the difference-feedback loop versus simply using two decoders with different losses; a controlled experiment with feature concatenation instead of subtraction would clarify the mechanism.
  • Since the method shows a larger relative gain on the private PBGTH dataset (lesion-consistent basal ganglia/thalamic hemorrhages) than on the more varied BHSD, one might infer that the difference signal is most useful when the label distribution is relatively homogeneous.
  • The reported Dice improvement on BHSD is modest (51.81 vs 50.70 for LeFeD) and the confidence intervals overlap; a reader should interpret the 'state-of-the-art' claim on BHSD with caution.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes SWDL-Net, a semi-supervised 3D intracranial hemorrhage segmentation framework combining a CNN encoder, a deep convolutional decoder, and a deep Laplacian pyramid upsampling decoder. The key mechanism is stratum-wise difference learning: feature differences between the two decoders are fed back into the encoder across iterations (Eq. 2). The method is evaluated on a private 271-case PBGTH dataset with 2% labeled data and on the public BHSD 2024 dataset with 5% labeled data, reporting Dice scores of 89.32% and 51.81%, respectively, and claiming state-of-the-art performance over UAMT, URPC, and LeFeD.

Significance. If the reported improvements are unbiased, the paper makes a useful contribution: the dual-decoder difference-learning mechanism is well motivated, the Laplacian pyramid decoder is a principled way to sharpen boundaries, and the authors state that code is released. The component ablation (Table V) provides evidence that difference learning and deep supervision each contribute to performance. However, the evaluation protocol has a load-bearing flaw: hyperparameters are selected on the test split, which inflates the reported margins and undermines the central SOTA claim. The significance is therefore conditional on fixing the evaluation-validity issues.

major comments (4)
  1. [III-F and IV-C] No validation split is described; Sec. III-F specifies an 80/20 train/test split only, while Sec. IV-C selects mu=1.5, T=3, and xi=1e-3 by maximizing performance on that same 20% test split (Tables II-IV). The headline Dice of 89.32% is thus the best point of a small hyperparameter grid evaluated on the test cases, not an unbiased estimate. Baselines are not given the same selection advantage, so the claimed margins and the Wilcoxon significance tests in Fig. 3 are optimistically biased. The authors must introduce a validation split (or nested cross-validation), tune on validation only, and then report test performance for all methods under the same protocol.
  2. [Table VI] The supervised VNet row reports ASD = 7.00 with 95% CI [14.32, 26.42], which is identical to the HD95 CI and impossible for an ASD mean below the CI lower bound; this is a clear data-entry error. The UAMT Dice CI [41.92, 95.90] is also implausibly wide relative to the mean 48.70. Because the BHSD results are used to support the generalizability claim, these numbers must be corrected and the statistical summaries recomputed.
  3. [II-C] The paper describes difference learning as the core mechanism (Eq. 2, encoder feedback of inter-decoder feature differences), but the unlabeled loss in Eq. (8) is only an MSE consistency between the two decoder outputs. It is not clear how the difference signal in Eq. (2) is trained or whether the gradient path through the feedback loop is active during optimization. Without this clarification, the ablation SWDL vs. SWDL-DL (Table V) is difficult to interpret: the improvement could come from the extra consistency loss or from the difference feedback, rather than from the claimed mechanism.
  4. [IV-B] The comparison set contains only three SSL baselines (UAMT, URPC, LeFeD). Given the claim of state-of-the-art performance, the evaluation should include at least one or two recent strong SSL segmentation methods, such as MC-Net+, UniMatch, or comparable 3D methods cited in the introduction. The current comparison is too narrow to support the 'current state-of-the-art' wording.
minor comments (6)
  1. [Table I] In the URPC row, the Accuracy CI lower bound is 84.27 while the mean is 94.95; this appears to be a typo and should be corrected.
  2. [II-C, Eq. (7)] The deep supervision loss uses both L and S for the number of strata; the notation should be made consistent (S is defined in the text, but Eq. (7) sums over L).
  3. [II-C] The symbol f_theta_DL is mentioned in the paragraph before Eq. (8) but is not defined; it appears to refer to the DelPU decoder and should be denoted consistently.
  4. [V (References)] Reference [42] is cited for UA-MT but is titled 'Unsupervised domain adaptation based on source-guided discrepancy'; the correct UA-MT reference is [20], or the citation should be fixed.
  5. [Abstract] The abstract states that 'Code and data have been released', but the PBGTH dataset is a private clinical dataset; please clarify exactly what is released (code, trained models, or only the public benchmark usage).
  6. [II-B] The stratum-adaptive depth rule in Eq. (6) gives D=2 for the input size 16x64x64 (max dim 64), but the text does not state the actual D used in the experiments; please specify the effective pyramid depth for all reported results.

Circularity Check

1 steps flagged · score 4.0 of 10

Test-set hyperparameter selection in Sec. IV-C makes the headline SOTA margin partly a selection artifact on the evaluation split, but the method derivation (Eqs. 1-9) is not circular and the difference-learning mechanism retains independent ablation support.

  1. fitted input called prediction [Sec. III-F split combined with Sec. IV-C (Tables II-IV) and Sec. IV-B/Table I headline Dice]
    "The dataset partitioning was performed using the KFold method from sklearn.model selection, allocating 80% of the data for training and 20% for testing. [...] Our comprehensive analysis of the consistency weight hyperparameter µ demonstrates a clear performance peak at µ=1.5, as evidenced by the quantitative results in Table II."

    The hyperparameters µ, T, and ξ are selected by maximizing performance on the same 20% test split that later yields the headline result (Dice 89.32%, Table I). Section III-F describes an 80/20 train/test split with no validation set, so the reported headline number is the best point of the evaluated grid on the evaluation set itself, while the baselines (UAMT, URPC, LeFeD) receive no comparable selection. The claimed margin over LeFeD (+1.79 Dice) and the Wilcoxon tests in Fig. 3 are therefore optimistically biased: the 'prediction' of SOTA generalization is the fitted grid maximum on the selection data, not an out-of-sample estimate.

full rationale

The core derivation chain is not circular: the architecture (Eqs. 1-9), including the stratum-wise difference between DC and DelPU decoder features fed back into the encoder via Eq. (2) and the MSE unsupervised consistency loss (Eq. 8), is defined independently of any benchmark number, and no reported metric is computed from a fitted parameter by the paper's own equations. The only self-citation, [29] (Wang, Chen and Mi, Frontiers in Neurology 2024), supports the skull-stripping preprocessing step and is not load-bearing for the central claim, so it does not raise the score by itself. The single partial circular step is the hyperparameter protocol: Sec. III-F fixes an 80/20 train/test split with no validation set, and Sec. IV-C selects µ=1.5, T=3, and ξ=1e-3 by best performance on that same 20% split (Tables II-IV); the headline Dice 89.32% is then reported in Table I as the SOTA estimate, while the baselines receive no comparable selection, making the SOTA margin partly a test-set selection artifact. This is weighted at 4 rather than higher because the difference-learning mechanism has independent support (Table V: SWDL 89.32 vs SWDL-DL 88.15), the hyperparameter grid is small, and the BHSD evaluation (Sec. IV-D) transfers the frozen hyperparameters to a different public dataset, providing a genuine out-of-sample check. Additional non-circularity concerns are noted for the record: an internal contradiction in the loss assignment (Sec. II-A says the DC decoder uses CE and the DelPU decoder uses Dice, while Sec. II-C and Eq. 7 state the opposite); data-entry errors in Table VI (supervised VNet's ASD confidence interval [14.32, 26.42] is identical to its HD95 CI, and UAMT's Dice CI [41.92, 95.90] is implausibly wide); and a mis-cited UAMT reference in Sec. V-B ([42] is a domain-adaptation paper, not UAMT). These are correctness and integrity risks, not circularity.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the empirical usefulness of cross-decoder differences and on standard CNN training assumptions. No new physical entities or formal axioms are introduced. The free parameters are mostly standard SSL hyperparameters, but they are tuned on the test set, which is a methodological concern.

free parameters (5)
  • mu (edge sharpening coefficient) = 1.5
    Selected by ablation on the test set (Table II); controls Laplacian pyramid reconstruction strength in Equation 5.
  • xi (difference learning influence) = 1e-3
    Selected by ablation on the test set (Table IV); scales the feedback difference added to encoder features in Equation 2.
  • T (difference learning iterations) = 3
    Selected by ablation on the test set (Table III); iteration period for the discrepancy loops in Section II-A.
  • omega_s (deep supervision weights) = {0.8, 0.6, 0.4, 0.2, 0.1}
    Set by hand in Equation 7; used for deeply supervised losses from low-stratum to high-stratum features.
  • ROI intensity threshold = 20-40 HU
    Empirically derived from training data in preprocessing; defines the hemorrhage search region and affects all methods.
assumptions (3)
  • domain assumption The difference between DC and DelPU decoder features at each stratum is a useful semi-supervised signal.
    Core premise of SWDL; only supported by the DL ablation in Table V on one dataset.
  • domain assumption Laplacian pyramid upsampling with a fixed Gaussian kernel preserves high-frequency boundary detail better than learned upsampling alone.
    Asserted in Section II-B and Discussion; no direct measurement beyond overall metrics.
  • domain assumption The VNet encoder/backbone is adequate for the 3D ICH segmentation task.
    Borrowed from prior SSL segmentation work [14], [20], [38].

how reviews work

0 comments
Cite this review

Pith. "Pith review of SWDL: Stratum-Wise Difference Learning with Deep Laplacian Pyramid for Semi-Supervised 3D Intracranial Hemorrhage Segmentation." pith.science (2026). https://pith.science/paper/Y6NPTH3G

@misc{pith2026250610325,
  author       = {Pith},
  title        = {Pith review of: SWDL: Stratum-Wise Difference Learning with Deep Laplacian Pyramid for Semi-Supervised 3D Intracranial Hemorrhage Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y6NPTH3G}},
  note         = {Machine review of arXiv:2506.10325}
}
read the original abstract

Recent advances in medical imaging have established deep learning-based segmentation as the predominant approach, though it typically requires large amounts of manually annotated data. However, obtaining annotations for intracranial hemorrhage (ICH) remains particularly challenging due to the tedious and costly labeling process. Semi-supervised learning (SSL) has emerged as a promising solution to address the scarcity of labeled data, especially in volumetric medical image segmentation. Unlike conventional SSL methods that primarily focus on high-confidence pseudo-labels or consistency regularization, we propose SWDL-Net, a novel SSL framework that exploits the complementary advantages of Laplacian pyramid and deep convolutional upsampling. The Laplacian pyramid excels at edge sharpening, while deep convolutions enhance detail precision through flexible feature mapping. Our framework achieves superior segmentation of lesion details and boundaries through a difference learning mechanism that effectively integrates these complementary approaches. Extensive experiments on a 271-case ICH dataset and public benchmarks demonstrate that SWDL-Net outperforms current state-of-the-art methods in scenarios with only 2% labeled data. Additional evaluations on the publicly available Brain Hemorrhage Segmentation Dataset (BHSD) with 5% labeled data further confirm the superiority of our approach. Code and data have been released at https://github.com/SIAT-CT-LAB/SWDL.

Figures

Figures reproduced from arXiv: 2506.10325 by the authors.

Figure 1
Figure 1. (a) Network architecture of SWDL-Net, highlighting its hybrid design integrating Laplacian pyramid processing and deep convolutional pathways. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Comparative visualization of segmentation performance across different methods for Primary Basal Ganglia and Thalamic Hemorrhages (PBGTH). [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison of SWDL versus baseline methods across six segmentation metrics. Boxplots show distributions (median ± IQR), with [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparative visualization of segmentation performance across different methods for the Brain Hemorrhage Segmentation Dataset (BHSD) 2024. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 34 canonical work pages

  1. [1]

    Multiclass semantic segmentation and quantification of traumatic brain injury lesions on head ct using deep learning: an algorithm development and multicentre validation study,

    M. Monteiro, V . F. Newcombe, F. Mathieu, K. Adatia, K. Kamnitsas, E. Ferrante, T. Das, D. Whitehouse, D. Rueckert, D. K. Menonet al., “Multiclass semantic segmentation and quantification of traumatic brain injury lesions on head ct using deep learning: an algorithm development and multicentre validation study,”The Lancet Digital Health, vol. 2, no. 6, pp...

    work page 2020

  2. [2]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer, 2015, pp. 234–241

    2015

  3. [3]

    Optimized u-net for brain tumor segmenta- tion,

    A. Arabo and S. Prasath, “Optimized u-net for brain tumor segmenta- tion,”Computers in Biology and Medicine, vol. 150, p. 106131, 2022

    work page 2022

  4. [4]

    Intracerebral haemorrhage,

    L. Puy, A. R. Parry-Jones, E. C. Sandset, D. Dowlatshahi, W. Ziai, and C. Cordonnier, “Intracerebral haemorrhage,”Nature Reviews Disease Primers, vol. 9, no. 1, p. 14, 2023

    work page 2023

  5. [5]

    Swin-unet: Unet-like pure transformer for medical image segmenta- tion,

    H. Cao, Y . Wang, J. Chen, D. Jiang, X. Zhang, Q. Tian, and M. Wang, “Swin-unet: Unet-like pure transformer for medical image segmenta- tion,”ECCV, 2022

    work page 2022

  6. [6]

    nnformer: In- terleaved transformer for volumetric segmentation,

    H.-Y . Zhou, J. Guo, X. Zhang, L. Yu, L. Wang, and Y . Yu, “nnformer: In- terleaved transformer for volumetric segmentation,”IEEE TMI, vol. 42, no. 4, pp. 941–952, 2023

    work page 2023

  7. [7]

    Med-former: A transformer based archi- tecture for medical image classification,

    G. J. Chowdary and Z. Yin, “Med-former: A transformer based archi- tecture for medical image classification,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2024, pp. 448–457

    work page 2024

  8. [8]

    Hdc-net: Hierarchical decoupled convolution network for brain tumor segmentation,

    Y . Wang, Y . Zhang, J. Tian, C. Zhong, Z. Shi, Y . Zhang, and Z. He, “Hdc-net: Hierarchical decoupled convolution network for brain tumor segmentation,”IEEE TMI, vol. 40, no. 10, pp. 2796–2807, 2021

    work page 2021

Show all 43 references
  1. [9]

    Dense-unet: a novel multiphoton in vivo cellular image segmentation model based on a convolutional neural network,

    S. Cai, Y . Tian, H. Lui, H. Zeng, Y . Wu, and G. Chen, “Dense-unet: a novel multiphoton in vivo cellular image segmentation model based on a convolutional neural network,”Quantitative imaging in medicine and surgery, vol. 10, no. 6, p. 1275, 2020

    2020

  2. [10]

    Self-supervised learning for medical image analysis using image context restoration,

    L. Chen, P. Bentley, K. Mori, K. Misawa, M. Fujiwara, and D. Rueck- ert, “Self-supervised learning for medical image analysis using image context restoration,”Medical image analysis, vol. 58, p. 101539, 2019

    2019

  3. [11]

    Inter-slice context residual learning for 3d medical image segmentation,

    J. Zhang, Y . Xie, Y . Wang, and Y . Xia, “Inter-slice context residual learning for 3d medical image segmentation,”IEEE Transactions on Medical Imaging, vol. 40, no. 2, pp. 661–672, 2020

    2020

  4. [12]

    Learning with limited annotations: a survey on deep semi-supervised learning for medical image segmentation,

    R. Jiao, Y . Zhang, L. Ding, B. Xue, J. Zhang, R. Cai, and C. Jin, “Learning with limited annotations: a survey on deep semi-supervised learning for medical image segmentation,”Computers in Biology and Medicine, vol. 169, p. 107840, 2024

    2024

  5. [13]

    Intra-and inter-pair consistency for semi-supervised gland segmentation,

    Y . Xie, J. Zhang, Z. Liao, J. Verjans, C. Shen, and Y . Xia, “Intra-and inter-pair consistency for semi-supervised gland segmentation,”IEEE Transactions on Image Processing, vol. 31, pp. 894–905, 2021

    2021

  6. [14]

    Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency,

    X. Luo, G. Wang, W. Liao, J. Chen, T. Song, Y . Chen, S. Zhang, D. N. Metaxas, and S. Zhang, “Semi-supervised medical image segmentation via uncertainty rectified pyramid consistency,”Medical Image Analysis, vol. 80, p. 102517, 2022

    2022

  7. [15]

    Mutual consistency learning for semi-supervised medical image segmentation,

    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, vol. 81, p. 102530, 2022

    2022

  8. [16]

    Diverse cotraining makes strong semi-supervised segmentor,

    Y . Li, X. Wang, L. Yang, L. Feng, W. Zhang, and Y . Gao, “Diverse cotraining makes strong semi-supervised segmentor,”arXiv preprint arXiv:2308.09281, 2023

    2023 arXiv

  9. [17]

    Semi-supervised semantic segmentation under label noise via diverse learning groups,

    P. Li, P. Purkait, T. Ajanthan, M. Abdolshah, R. Garg, H. Husain, C. Xu, S. Gould, W. Ouyang, and A. Van Den Hengel, “Semi-supervised semantic segmentation under label noise via diverse learning groups,” inProceedings of the IEEE/CVF International Conference on Computer Vision...

    2023

  10. [18]

    Revisiting weak- to-strong consistency in semi-supervised semantic segmentation,

    L. Yang, L. Qi, L. Feng, W. Zhang, and Y . Shi, “Revisiting weak- to-strong consistency in semi-supervised semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 7236–7246

    2023

  11. [19]

    Boostmis: Boosting medical image semi-supervised learning with adaptive pseudo labeling and informative active annotation,

    W. Zhang, L. Zhu, J. Hallinan, S. Zhang, A. Makmur, Q. Cai, and B. C. Ooi, “Boostmis: Boosting medical image semi-supervised learning with adaptive pseudo labeling and informative active annotation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern reco...

    2022

  12. [20]

    Uncertainty-aware self-ensembling model for semi-supervised 3d left atrium segmenta- tion,

    L. Yu, S. Wang, X. Li, C.-W. Fu, and P.-A. Heng, “Uncertainty-aware self-ensembling model for semi-supervised 3d left atrium segmenta- tion,” inMedical image computing and computer assisted intervention– MICCAI 2019: 22nd international conference, Shenzhen, China, October 13–1...

    2019

  13. [21]

    Pefat: Boosting semi-supervised medical image classification via pseudo-loss estimation and feature adversarial training,

    Q. Zeng, Y . Xie, Z. Lu, and Y . Xia, “Pefat: Boosting semi-supervised medical image classification via pseudo-loss estimation and feature adversarial training,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 15 671–15 680

    2023

  14. [22]

    Compete to win: Enhancing pseudo labels for barely-supervised medical image segmentation,

    H. Wu, X. Li, Y . Lin, and K.-T. Cheng, “Compete to win: Enhancing pseudo labels for barely-supervised medical image segmentation,”IEEE Transactions on Medical Imaging, vol. 42, no. 11, pp. 3244–3255, 2023

    2023

  15. [23]

    Uncertainty-guided semi- supervised (ugss) mean teacher framework for brain hemorrhage seg- mentation and volume quantification,

    S. H. Emon, T.-L. B. Tseng, M. Pokojovy, S. Moen, P. McCaffrey, E. Walser, A. V o, and M. F. Rahman, “Uncertainty-guided semi- supervised (ugss) mean teacher framework for brain hemorrhage seg- mentation and volume quantification,”Biomedical Signal Processing and Control, vol....

    2025

  16. [24]

    Label-efficient sequential model- based weakly supervised intracranial hemorrhage segmentation in low- 11 data non-contrast ct imaging,

    S. H. Ramananda and V . Sundaresan, “Label-efficient sequential model- based weakly supervised intracranial hemorrhage segmentation in low- 11 data non-contrast ct imaging,”Medical Physics, vol. 52, no. 4, pp. 2123– 2144, 2025

    2025

  17. [25]

    Cut-paste consistency learning for semi- supervised lesion segmentation,

    B. P. Yap and B. K. Ng, “Cut-paste consistency learning for semi- supervised lesion segmentation,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, 2023, pp. 6160–6169

    2023

  18. [26]

    Weakly supervised intracranial hemorrhage segmentation with yolo and an uncertainty rectified segment anything model,

    P. Spiegler, A. Rasoulian, and Y . Xiao, “Weakly supervised intracranial hemorrhage segmentation with yolo and an uncertainty rectified segment anything model,” inMICCAI Challenge on Ischemic Stroke Lesion Segmentation. Springer, 2024, pp. 12–21

    2024

  19. [27]

    Bhsd: A 3d multi-class brain hemorrhage segmentation dataset,

    B. Wu, Y . Xie, Z. Zhang, J. Ge, K. Yaxley, S. Bahadir, Q. Wu, Y . Liu, and M.-S. To, “Bhsd: A 3d multi-class brain hemorrhage segmentation dataset,” inInternational Workshop on Machine Learning in Medical Imaging. Springer, 2023, pp. 147–156

    2023

  20. [28]

    3d slicer,

    S. Pieper, M. Halle, and R. Kikinis, “3d slicer,” in2004 2nd IEEE international symposium on biomedical imaging: nano to macro (IEEE Cat No. 04EX821). IEEE, 2004, pp. 632–635

    2004

  21. [29]

    A task-driven cerebral angiographic imaging based on ct perfusion,

    C. Wang, S. Chen, and D. Mi, “A task-driven cerebral angiographic imaging based on ct perfusion,”Frontiers in Neurology, vol. 14, p. 1328184, 2024

    2024

  22. [30]

    Individual comparisons by ranking methods,

    F. Wilcoxon, “Individual comparisons by ranking methods,”Break- throughs in statistics, pp. 196–202, 1992

    1992

  23. [31]

    W. J. Conover,Practical nonparametric statistics. John Wiley & Sons, 1999, vol. 350

    1999

  24. [32]

    Measures of the amount of ecologic association between species,

    L. R. Dice, “Measures of the amount of ecologic association between species,”Ecology, vol. 26, no. 3, pp. 297–302, 1945

    1945

  25. [33]

    An efficient algorithm for calculating the exact hausdorff distance,

    A. A. Taha and A. Hanbury, “An efficient algorithm for calculating the exact hausdorff distance,”IEEE transactions on pattern analysis and machine intelligence, vol. 37, no. 11, pp. 2153–2163, 2015

    2015

  26. [34]

    Comparison and evaluation of methods for liver segmentation from ct datasets,

    T. Heimann, B. Van Ginneken, M. A. Styner, Y . Arzhaeva, V . Aurich, C. Bauer, A. Beck, C. Becker, R. Beichel, G. Bekeset al., “Comparison and evaluation of methods for liver segmentation from ct datasets,”IEEE transactions on medical imaging, vol. 28, no. 8, pp. 1251–1265, 2009

    2009

  27. [35]

    Evaluation: from precision, recall and f-measure to roc, informedness, markedness and correlation,

    D. M. W. Powers, “Evaluation: from precision, recall and f-measure to roc, informedness, markedness and correlation,”arXiv preprint arXiv:2010.16061, 2020

    2010 arXiv

  28. [36]

    Sch ¨utze, C

    H. Sch ¨utze, C. D. Manning, and P. Raghavan,Introduction to informa- tion retrieval. Cambridge University Press Cambridge, 2008, vol. 39

    2008

  29. [37]

    The distribution of the flora in the alpine zone,

    P. Jaccard, “The distribution of the flora in the alpine zone,”New Phytologist, vol. 11, no. 2, pp. 37–50, 1912

    1912

  30. [38]

    Consistency-guided differential decoding for enhancing semi- supervised medical image segmentation,

    Q. Zeng, Y . Xie, Z. Lu, M. Lu, J. Zhang, Y . Zhou, and Y . Xia, “Consistency-guided differential decoding for enhancing semi- supervised medical image segmentation,”IEEE Transactions on Medical Imaging, 2024

    2024

  31. [39]

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

    F. Milletari, N. Navab, and S.-A. Ahmadi, “V-net: Fully convolutional neural networks for volumetric medical image segmentation,” in2016 fourth international conference on 3D vision (3DV). Ieee, 2016, pp. 565–571

    2016

  32. [40]

    Pytorch: An imperative style, high-performance deep learning library. arxiv 2019,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antigaet al., “Pytorch: An imperative style, high-performance deep learning library. arxiv 2019,” arXiv preprint arXiv:1912.01703, vol. 10, 1912

    2019 arXiv

  33. [41]

    On the importance of initialization and momentum in deep learning,

    I. Sutskever, J. Martens, G. Dahl, and G. Hinton, “On the importance of initialization and momentum in deep learning,” inInternational conference on machine learning. PMLR, 2013, pp. 1139–1147

    2013

  34. [42]

    Unsupervised domain adapta- tion based on source-guided discrepancy,

    T. Yao, Y . Pan, C.-W. Ngo, and T. Mei, “Unsupervised domain adapta- tion based on source-guided discrepancy,”AAAI, vol. 33, pp. 9022–9029, 2019

    2019

  35. [43]

    Semi-supervised brain lesion segmentation with an adapted mean teacher model,

    C. Chen, Q. Dou, H. Chen, J. Qin, and P.-A. Heng, “Semi-supervised brain lesion segmentation with an adapted mean teacher model,”IPMI, pp. 554–565, 2019

    2019

Pith tools

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