Pith. sign in

REVIEW 4 major objections 4 minor 72 references

Co-Evidential Fusion with Information Volume for Medical Image Segmentation

T0 review · 4 major / 4 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 The paper argues that co-evidential fusion with information-volume weighting yields more precise voxel-level uncertainty, and reports the best Dice, Jaccard, and surface-distance scores on LA, pancreas, cardiac MRI, and brain tumour…

desk verdict Essentially a re-presentation of the authors' own Pattern Recognition 2025 paper, with an under-specified fusion rule and no statistical grounding for the headline gains. read the letter →

arxiv 2506.02492 v1 pith:KHTLT6HV submitted 2025-06-03 cs.CV

Yuanpeng He , Lijian Li , Tianxiang Zhan , Chi-Man Pun , Wenpin Jiao , Zhi Jin This is my paper
classification cs.CV
keywords semi-supervisedmedicalimagesegmentationevidentialdeeplearningDempster-Shaferevidencetheorypignisticfusioninformationvolumeofmassfunctionuncertainty-guidedbidirectionalcopy-paste
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a co-evidential fusion strategy can make semi-supervised medical image segmentation more effective when labels are scarce. The idea is to turn predictions from mixed copy-pasted images and from original labeled images into Dempster-Shafer mass functions, fuse them with a pignistic combination rule, and use the fused uncertainty to guide voxel-level learning. A second ingredient, the information volume of a mass function (IVUM), reweights the evidential loss so that high-information evidence counts more. On four public benchmarks (left atrium, pancreas, cardiac MRI, brain tumour) with 5%, 10% and 20% labeled data, the method reports the best Dice, Jaccard, 95% Hausdorff distance and average surface distance, including a 3.61 Dice point gain over the BCP baseline on the LA dataset at 5% labels. If these results hold up, clinicians could train accurate segmenters from a small fraction of today's annotations.

What carries the argument

The load-bearing object is the pignistic co-evidential fusion rule in Eq. (30), which combines two mass functions so that large-cardinality propositions are less perturbed, and the IVUM score in Eq. (33), which multiplies the fused uncertainty by the information volume of the mass function. IVUM is computed by iteratively splitting mass functions according to the maximum Deng entropy until convergence. These two pieces turn raw network predictions into per-voxel weighting signals, and the weighting is applied through three optimization objectives in both the pre-training and self-training stages.

What would settle it

Run the LA 5%-labeled experiment with Eq. (30) replaced by Dempster's combination rule or plain averaging, keeping all other losses fixed; if Dice does not drop, the co-evidential fusion rule is not what produces the reported gain.

Watch

Extended reading notes

Core claim

At its core, the paper proposes that voxel-level uncertainty in semi-supervised segmentation is better estimated by combining evidence from several views than by any single prediction. In the pre-training and self-training stages, the method generates evidence from mixed images, and separately from the original labeled images, then fuses these evidence pairs with the proposed pignistic rule. The fused mass functions are scored by IVUM, the product of the uncertainty measure and the information volume of the mass function, and these scores enter three evidential losses: an uncertainty-sorted loss, a generalized evidential deep learning loss weighted by IVUM, and a combination of the two. The author's central claim is that this co-evidential pipeline, built on the BCP bidirectional copy-paste framework, is what produces the reported improvements over BCP, Co-BioNet, and other SSL methods across four datasets and three label ratios.

Load-bearing premise

The method's gains rest on the assumption that the pignistic fusion rule and IVUM weighting produce a genuinely better voxel-level uncertainty signal than simpler alternatives; if they do not, the reported improvements could be carried by the BCP copy-paste backbone and the added loss terms.

Editorial extensions

If this is right

  • If the central claim is correct, semi-supervised segmentation can reach roughly the same Dice with 5% labeled data that earlier methods need 10% or more to obtain, as reported on the NIH-Pancreas dataset.
  • The co-evidential fusion and IVUM weighting can be plugged into the BCP-style bidirectional copy-paste training pipeline used in the paper, providing a recipe for transferring the gains to other SSL segmentation tasks.
  • The reported LA results would mean the method narrows the gap to fully-supervised V-Net (92.62 Dice) at just 5% labels (91.63 Dice), leaving less than one point of difference.
  • Across 5%, 10% and 20% labeled ratios, the paper reports steady gains over BCP on all four metrics, so the benefit is not confined to a single low-label setting.

Reading between the lines

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

  • A reader could test whether the specific fusion rule is responsible for the gain by replacing Eq. (30) with Dempster's rule or simple evidence averaging while keeping every other loss fixed; if Dice does not drop, the gain would be attributable to the broader evidential training scheme rather than the pignistic rule.
  • The paper does not evaluate the fused uncertainty estimates themselves for calibration or Brier score, so a natural follow-up is to check whether IVUM-weighted uncertainties are actually better calibrated than unfused ones.
  • The co-evidential idea is not inherently tied to medical images or to the BCP backbone, so a plausible extension is to apply the same evidence fusion and IVUM weighting to other dense prediction tasks with sparse labels, although the paper does not test this.
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 / 4 minor

Summary. The paper proposes a semi-supervised medical image segmentation method built on the BCP bidirectional copy-paste framework. Its two claimed contributions are (i) a 'pignistic co-evidential fusion' rule (Eq. 30) that combines evidence derived from mixed and original labeled images, and (ii) an 'Information Volume of Mass Function' (IVUM) measure, defined as the fused uncertainty mass times a Deng-entropy-based information volume, which is used to reweight an evidential deep learning loss and to define an additional uncertainty-sorted loss. The method is evaluated on LA, NIH-Pancreas, ACDC, and BraTS datasets at 5%, 10%, and 20% labeled ratios, with reported Dice/Jaccard/95HD/ASD metrics. The tables consistently place the proposed method ahead of BCP and other baselines, with the headline result being a 3.61 Dice improvement over BCP on the LA dataset at 5% labels.

Significance. If the results hold, the method would be a meaningful advance for low-label medical image segmentation, as it combines a strong existing backbone (BCP) with an evidence-theoretic uncertainty weighting scheme and reports consistent gains across four public benchmarks and three label ratios. The paper makes a genuine effort to evaluate on multiple datasets and metrics, and the direction of modeling voxel-level uncertainty through mass-function information volume is interesting. However, the significance is currently limited by three issues: the fusion rule that is the central novelty is underspecified, no ablation isolates its contribution from simpler alternatives, and the evaluation lacks error bars or significance tests. The reported gains are therefore not yet convincingly attributed to the proposed machinery.

major comments (4)
  1. [3.3.2, Eqs. (30)-(33)] The co-evidential fusion rule is incomplete. In Eq. (25) the uncertainty measure U is identified with the mass ε1({P_{N+1}}), and Eq. (33) forms IVUM by multiplying the fused uncertainty mass ε1,2({P_{N+1}}) with IV(ε1,2). However, Eq. (30) gives a formula only for the singleton propositions {P_i} and never specifies ες({S}) (equivalently ες({P_{N+1}})), even though Eq. (31) normalizes over i = 1..N+1. An independent implementation must guess how to fuse the uncertainty masses, and different guesses change IVUM and therefore the loss in Eq. (34). This is an internal gap, not a matter of taste; the method as written is not fully reproducible.
  2. [4.2 and Table 3] There is no ablation that isolates the proposed fusion operator in Eq. (30) from simpler alternatives. Table 3 sweeps only the six λ weights of the combined losses; it never compares Eq. (30) against Dempster's rule (Eq. 9), against averaging, or against omitting the fusion entirely. Consequently, the reported margin over BCP (e.g., 3.61 Dice at LA 5%) could in principle be due to the BCP copy-paste backbone and the added loss terms even if the fusion mechanism itself is inert or misspecified. An ablation replacing Eq. (30) with Dempster's rule and with plain averaging, holding all other components fixed, is needed to support the central claim.
  3. [4.1 and Table 1/Table 3] The evaluation reports a single run per condition with no standard deviations and no significance tests. Moreover, the parameter study in Table 3 is conducted on the same LA 5% labeled setting whose Dice of 91.63 is then reported as the headline result in Table 1. If the best λ configuration is selected on the same evaluation split whose number is reported, that number is a grid-selected maximum rather than an unbiased estimate. The paper should clarify which split is used for the parameter study and report multiple-seed means and standard deviations (or at least significance tests) for the main comparisons.
  4. [Table 3] Several cells in Table 3 contain impossible values in which the reported Jaccard exceeds the reported Dice, e.g., the row for λ3=1.6, λ6=2.4 shows '2.4 89.85 90.31 7.61 2.04' and the row for λ3=4.8, λ6=4.8 shows '4.8 89.18 89.97 8.07 2.33'. Since Jaccard ≤ Dice always holds for any segmentation, these entries indicate transcription errors. The table must be corrected, and the parameter analysis should be re-verified.
minor comments (4)
  1. [3.3.2, Eq. (30)] The proposed rule is called 'pignistic co-evidential fusion,' but Eq. (30) does not follow from the pignistic transformation defined in Eq. (11). Please rename or provide a derivation that justifies the use of 'pignistic.'
  2. [3.3.2, Eq. (32) and following text] The sentence 'ε1,2 and ε1,3 represent fused results of ε1,ε2 and ε2,ε3' should presumably read 'ε1,2 and ε1,3 represent fused results of ε1,ε2 and ε1,ε3'; the later reference to 'IVUM ε2,3' should be 'IVUM ε1,3'.
  3. [3.1.3, Eq. (13)] The convergence threshold ρ in the IVUM computation is never given a value in the experimental setup of Section 4.1. Please specify how ρ is set.
  4. [1, Introduction] The introduction contains long passages that read as generic descriptions of D-S evidence theory applications (robotics, cybersecurity, environmental science) that are not connected to the proposed method. Consider tightening this text to focus on the method's actual contributions.

Circularity Check

1 steps flagged · score 3.0 of 10

One grid-selected lambda maximum is reported as the LA 5% result; otherwise the method is evaluated against external benchmarks and the derivation is not circular.

  1. fitted input called prediction [Section 4.4 (Parameters Study), Table 3; Table 1 (LA 5% results)]
    "When λ1 =1.6 and λ2 =2.4, our model achieves the best performance (91.63%, 83.58%, 5.26, and 1.67 on the four given metrics). ... Ours91.63 83.58 5.26 1.67"

    The 'Ours' entry for LA 5% in Table 1 is identical to the best Dice found by the lambda grid search in Table 3, which is reported on the same LA 5% label-ratio configuration. The loss-balancing weights λ1-λ6 are therefore selected to maximize the exact metric that is then presented as the method's result and used to claim a 3.61 Dice margin over BCP. Unless a separate validation split was used (the paper does not say so), this headline number is a grid-selected maximum, so the comparison is partially forced by the selection procedure rather than being an independent prediction of a fixed method.

full rationale

The central segmentation claims are evaluated on four external benchmarks against published methods, so the main performance result is not equivalent to the paper's inputs. The evidential fusion and IVUM components are specified as new training objectives, and their formulas are either standard (D-S combination, pignistic transform, Deng entropy) or stated in the paper; the many self-citations ([16]-[24], [28], [31]) are related-work pointers rather than load-bearing derivations. The principal circularity risk is confined to the LA 5% parameter study: the best lambda configuration is chosen on the same benchmark configuration and then reported as 'Ours', making that headline Dice a selected maximum (fitted-input-called-prediction). Two serious non-circular gaps also weaken the derivation chain: Eq. (30) defines the fused mass only for singleton propositions, leaving ε1,2({S}) undefined even though Eq. (33) multiplies exactly that quantity into IVUM; and no ablation isolates the co-evidential fusion rule from Dempster's rule or simple averaging, so the causal contribution of the fusion mechanism is not established. These are missing-proof and validation issues, not reductions by construction. Overall, the method has independent external content, so the score is 3.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central claim rests on a stack of imported evidential constructs (Dirichlet-based EDL, Deng entropy, information volume) plus two new ad hoc components (the pignistic fusion rule and the IVUM weighting). Six loss weights are tuned on the LA 5% dataset and then reported as the method's performance, and both phi and rho are unspecified. There are no invented physical entities, but the new IVUM metric and fusion operator are not independently validated.

free parameters (5)
  • lambda1, lambda2, lambda3 (pre-training loss weights) = lambda1=1.6, lambda2=2.4, lambda3=0.8
    Selected in Table 3 to maximize Dice on LA dataset with 5% labeled data; the same best configuration is reported as the method's result, without nested validation.
  • lambda4, lambda5, lambda6 (self-training loss weights) = lambda4=1.6, lambda5=2.4, lambda6=0.8
    Selected in Table 3 on LA 5% with the same protocol as above.
  • phi (amplitude in dynamic weight function, Eq. 22) = not reported
    Controls the range of the per-epoch/voxel weight; a value is needed to implement the loss but is not specified.
  • rho (convergence threshold in IVUM algorithm, Eq. 13) = not reported
    Stopping criterion for the Deng-entropy iterative splitting; not specified, which affects the computed information volume.
  • eta (copy-paste mask zero-region fraction) = varies per dataset (e.g., 74x74x53 of 112x112x80 for LA)
    Set by hand following BCP/SS-Net; a design choice that influences how much context is mixed.
assumptions (5)
  • domain assumption The Dirichlet-based evidential deep learning formulation of Sensoy et al. [47] correctly captures classification uncertainty.
    Section 3.1.1 adopts EDL equations (1)-(4) without independent validation in this medical segmentation setting.
  • domain assumption Deng entropy and the iterative information-volume algorithm of [56,57] are valid measures of evidence information content.
    Used in Eqs. (12)-(13) and (33); the paper does not verify these measures against other uncertainty metrics.
  • ad hoc to paper The pignistic co-evidential fusion rule in Eq. (30) is a valid and useful combination operator.
    Introduced heuristically with the statement that it is 'designed to obtain more obvious uncertainty indicators'; no derivation or comparison to Dempster's rule.
  • ad hoc to paper Weighting the evidential loss by (1 - IVUM) in Eq. (34) improves learning.
    The specific functional form is postulated; its effect is tested only indirectly through final segmentation accuracy.
  • domain assumption The two-sub-network mutual training strategy of BCP [58] works as described.
    Section 3.2 states 'the training strategy is the same as BCP [58]', so the paper inherits BCP's assumptions about pseudo-labeling and copy-paste.
invented entities (2)
  • IVUM (Information Volume of Mass Function)
    purpose: A voxel-level weighting score that combines the uncertainty measure U with the information volume of the fused mass function; used to reweight the evidential loss and to re-rank voxels for uncertainty-guided learning.
    Defined in Eq. (33) for the first time; there is no external validation of IVUM as a calibrated uncertainty or information measure, only downstream segmentation accuracy.
  • Pignistic co-evidential fusion operator (epsilon_sigma in Eq. 30)
    purpose: Combines evidence from mixed images and original labeled images to produce fused mass functions intended to highlight uncertainty.
    A new ad hoc operator; its behavior is not characterized theoretically or compared against existing fusion rules in an ablation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Co-Evidential Fusion with Information Volume for Medical Image Segmentation." pith.science (2026). https://pith.science/paper/KHTLT6HV

@misc{pith2026250602492,
  author       = {Pith},
  title        = {Pith review of: Co-Evidential Fusion with Information Volume for Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KHTLT6HV}},
  note         = {Machine review of arXiv:2506.02492}
}
read the original abstract

Although existing semi-supervised image segmentation methods have achieved good performance, they cannot effectively utilize multiple sources of voxel-level uncertainty for targeted learning. Therefore, we propose two main improvements. First, we introduce a novel pignistic co-evidential fusion strategy using generalized evidential deep learning, extended by traditional D-S evidence theory, to obtain a more precise uncertainty measure for each voxel in medical samples. This assists the model in learning mixed labeled information and establishing semantic associations between labeled and unlabeled data. Second, we introduce the concept of information volume of mass function (IVUM) to evaluate the constructed evidence, implementing two evidential learning schemes. One optimizes evidential deep learning by combining the information volume of the mass function with original uncertainty measures. The other integrates the learning pattern based on the co-evidential fusion strategy, using IVUM to design a new optimization objective. Experiments on four datasets demonstrate the competitive performance of our method.

Figures

Figures reproduced from arXiv: 2506.02492 by the authors.

Figure 1
Figure 1. Details of the Pre-training Stage. L–LM : Labeled and Labeled Mixed Information. to regions where mass functions exhibit concentrated belief (e.g., clear tumor boundaries) while deprioritizing areas with diffuse or conflicting evidence (e.g., ambiguous lesions). This approach reduces noise in pseudo-labels and strengthens consistency regularization, particularly in low-data regimes. Moreover, IVMF-driven learning fr… view at source ↗
Figure 2
Figure 2. Details of the Self-training Stage. L–ULM :Labeled and Unlabeled & Labeled Mixed Information. where λ1, λ2 and λ3 are balancing parameters. The evidential learning process that produces λ1Lsort,I + λ2Lgedl,I + λ3Ls−gl,I is denoted by E. For the acquisition of parameter Is2, the corresponding optimization objective can be defined as: L f inal Is2 = LIs1 + EIs1 (38) where L f inal Is2 represents the final objective fu… view at source ↗
Figure 4
Figure 4. The Jaccard Results of the Comparative Models on LA [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figures from the paper (8 more)
Figure 6
Figure 6. Figure 6: The ASD Results of the Comparative Models on LA Dataset [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Comparative Results of Four Metrics on BraTS Dataset [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Average Metrics on BraTS Dataset: Comparative Results [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Performance Comparisons with Variation of Parameter [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: Performance Comparisons with Variation of Parameter [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Performance Comparisons with Variation of Parameter [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Visualized Comparisons on LA, BraTS and Pancreas dataset. The first row represents the ground truth. In the second row, the first [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]
Figure 13
Figure 13. Figure 13: Visualized Comparisons with ground truth and BCP on the ACDC Dataset.The first row represents the ground truth, the second [PITH_FULL_IMAGE:figures/full_fig_p022_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

72 extracted references · 39 canonical work pages

  1. [24]

    Y. He, L. Li, T. Zhan, C.-M. Pun, W. Jiao, Z. Jin, Co-evidential fusion with information volume for semi- supervised medical image segmentation, Pattern Recognition 166 (2025) 111639

    2025

  2. [22]

    Y. He, L. Li, Uncertainty-aware evidential fusion-based learning for semi-supervised medical image segmen- tation, arXiv preprint arXiv:2404.06177 (2024)

    arXiv 2024

  3. [23]

    He, Epl: Evidential prototype learning for semi-supervised medical image segmentation, arXiv preprint arXiv:2404.06181 (2024)

    Y. He, Epl: Evidential prototype learning for semi-supervised medical image segmentation, arXiv preprint arXiv:2404.06181 (2024)

    arXiv 2024

  4. [1]

    Tragakis, C

    A. Tragakis, C. Kaul, R. Murray-Smith, D. Husmeier, The fully convolutional transformer for medical image segmentation, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023, pp. 3660–3669

    work page 2023

  5. [2]

    Z. Li, Y. Li, Q. Li, P . Wang, D. Guo, L. Lu, D. Jin, Y. Zhang, Q. Hong, Lvit: language meets vision transformer in medical image segmentation, IEEE Transactions on Medical Imaging (2023)

    work page 2023

  6. [3]

    T. Xu, K. Yan, Y. He, S. Gao, K. Yang, J. Wang, J. Liu, Z. Liu, Spatio-temporal variability analysis of vege- tation dynamics in china from 2000 to 2022 based on leaf area index: A multi-temporal image classification perspective, Remote Sensing 15 (12) (2023) 2975

    2023

  7. [4]

    W. Feng, L. Ju, L. Wang, K. Song, X. Zhao, Z. Ge, Unsupervised domain adaptation for medical image segmen- tation by selective entropy constraints and adaptive semantic alignment, in: B. Williams, Y. Chen, J. Neville (Eds.), Thirty-Seventh AAAI Conference on Artificial Intelligence, AAAI 2023, Thirty-Fifth Conference on In- novative Applications of Artifi...

    work page 2023

  8. [5]

    Y. Cui, W. Ren, S. Yang, X. Cao, A. Knoll, Irnext: Rethinking convolutional network design for image restora- tion, in: A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, J. Scarlett (Eds.), International Conference 24 on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, Vol. 202 of Proceedings of Ma- chine Learning Research, PM...

    work page 2023

Show all 72 references
  1. [6]

    Balaji, Image augmentation based on variational autoencoder for breast tumor segmentation, Academic Radiology (2023)

    K. Balaji, Image augmentation based on variational autoencoder for breast tumor segmentation, Academic Radiology (2023)

    2023

  2. [7]

    Chaitanya, E

    K. Chaitanya, E. Erdil, N. Karani, E. Konukoglu, Local contrastive loss with pseudo-label based self-training for semi-supervised medical image segmentation, Medical Image Analysis 87 (2023) 102792

    2023

  3. [8]

    Zhang, S

    J. Zhang, S. Zhang, X. Shen, T. Lukasiewicz, Z. Xu, Multi-condos: Multimodal contrastive domain sharing generative adversarial networks for self-supervised medical image segmentation, IEEE Transactions on Med- ical Imaging (2023)

    2023

  4. [9]

    Y. Meng, H. Zhang, Y. Zhao, D. Gao, B. Hamill, G. Patri, T. Peto, S. Madhusudhan, Y. Zheng, Dual consistency enabled weakly and semi-supervised optic disc and cup segmentation with dual adaptive graph convolu- tional networks, IEEE Transactions on Medical Imaging 42 (2) (2022) 416–429

    2022

  5. [10]

    Jiang, Z

    H. Jiang, Z. Diao, T. Shi, Y. Zhou, F. Wang, W. Hu, X. Zhu, S. Luo, G. Tong, Y.-D. Yao, A review of deep learning-based multiple-lesion recognition from medical images: classification, detection and segmentation, Computers in Biology and Medicine (2023) 106726

    2023

  6. [11]

    A. B. Abdusalomov, R. Nasimov, N. Nasimova, B. Muminov, T. K. Whangbo, Evaluating synthetic medical images using artificial intelligence with the gan algorithm, Sensors 23 (7) (2023) 3440

    2023

  7. [12]

    H. Wu, X. Li, K.-T. Cheng, Exploring feature representation learning for semi-supervised medical image seg- mentation, IEEE Transactions on Neural Networks and Learning Systems (2023)

    2023

  8. [13]

    L. Qiu, J. Cheng, H. Gao, W. Xiong, H. Ren, Federated semi-supervised learning for medical image segmen- tation via pseudo-label denoising, IEEE Journal of Biomedical and Health Informatics (2023)

    2023

  9. [14]

    Y. Wang, B. Xiao, X. Bi, W. Li, X. Gao, Mcf: Mutual correction framework for semi-supervised medical image segmentation, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 15651–15660

    2023

  10. [15]

    Rangnekar, C

    A. Rangnekar, C. Kanan, M. Hoffman, Semantic segmentation with active semi-supervised learning, in: Pro- ceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023, pp. 5966–5977

    2023

  11. [16]

    Y. He, F. Xiao, Conflicting management of evidence combination from the point of improvement of basic probability assignment, International Journal of Intelligent Systems 36 (5) (2021) 1914–1942

    2021

  12. [17]

    Y. He, F. Xiao, A new base function in basic probability assignment for conflict management, Applied Intelli- gence 52 (4) (2022) 4473–4487

    2022

  13. [18]

    Y. He, Y. Deng, Mmget: a markov model for generalized evidence theory, Computational and Applied Math- ematics 41 (2022) 1–41

    2022

  14. [19]

    Y. He, Y. Deng, Ordinal fuzzy entropy, Iranian Journal of Fuzzy Systems 19 (3) (2022) 171–186. 25

    2022

  15. [20]

    Y. He, Y. Deng, Ordinal belief entropy, Soft Computing 27 (11) (2023) 6973–6981

    2023

  16. [21]

    Y. He, Y. Deng, Tdqmf: Two-dimensional quantum mass function, Information Sciences 621 (2023) 749–765

    2023

  17. [25]

    Y. He, L. Li, T. Zhan, W. Jiao, C.-M. Pun, Generalized uncertainty-based evidential fusion with hybrid multi- head attention for weak-supervised temporal action localization, in: ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)...

    2024

  18. [26]

    Y. He, W. Song, L. Li, T. Zhan, W. Jiao, Residual feature-reutilization inception network, Pattern Recognition (2024) 110439

    2024

  19. [27]

    Y. Bi, E. Che, Y. Chen, Y. He, J. Qu, Multi-prototype-based embedding refinement for medical image seg- mentation, in: ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2025, pp. 1–5

    2025

  20. [28]

    L. Li, Y. He, C.-M. Pun, Efficient prototype consistency learning in semi-supervised medical image segmen- tation via joint uncertainty and data augmentation, in: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), IEEE, 2024, pp. 2114–2121

    2024

  21. [29]

    Y. He, L. Li, T. Zhan, A matrix-based distance of pythagorean fuzzy set and its application in medical diag- nosis, arXiv preprint arXiv:2102.01538 (2021)

    2021 arXiv

  22. [30]

    Huang, Y

    C.-j. Huang, Y. He, X. Han, W. Jiao, Z. Jin, L. Wang, Unitrans: A unified vertical federated knowledge transfer framework for enhancing cross-hospital collaboration, arXiv preprint arXiv:2501.11388 (2025)

    2025 arXiv

  23. [31]

    Y. He, Y. Bi, L. Li, C.-M. Pun, W. Jiao, Z. Jin, Mutual evidential deep learning for semi-supervised medical im- age segmentation, in: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), IEEE, 2024, pp. 2010–2017

    2024

  24. [32]

    L. Li, Y. He, L. Li, Nndf: A new neural detection network for aspect-category sentiment analysis, in: Interna- tional Conference on Knowledge Science, Engineering and Management, Springer International Publishing Cham, 2022, pp. 339–355

    2022

  25. [33]

    T. Zhan, Y. He, Y. Deng, Z. Li, Differential convolutional fuzzy time series forecasting, IEEE Transactions on Fuzzy Systems 32 (3) (2023) 831–845

    2023

  26. [34]

    T. Zhan, Y. He, Y. Deng, Z. Li, W. Du, Q. Wen, Time evidence fusion network: Multi-source view in long-term time series forecasting, arXiv preprint arXiv:2405.06419 (2024). 26

    2024 arXiv

  27. [35]

    L. Li, Z. Jin, Y. He, D. Jin, H. Duan, Z. Tao, X. Zhang, J. Li, Rethinking regularization methods for knowledge graph completion, arXiv preprint arXiv:2505.23442 (2025)

    2025 arXiv

  28. [36]

    L. Li, Z. Jin, Y. Zhang, D. Jin, C. Dou, Y. He, X. Zhang, H. Zhao, Towards structure-aware model for multi- modal knowledge graph completion, arXiv preprint arXiv:2505.21973 (2025)

    2025 arXiv

  29. [37]

    X. Chen, Z. Tao, K. Zhang, C. Zhou, W. Gu, Y. He, M. Zhang, X. Cai, H. Zhao, Z. Jin, Revisit self-debugging with self-generated tests for code generation, arXiv preprint arXiv:2501.12793 (2025)

    2025 arXiv

  30. [38]

    L. Li, Y. He, C.-M. Pun, An adaptive framework for multi-view clustering leveraging conditional entropy op- timization, in: ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

    2025

  31. [39]

    T. Wang, J. Lu, Z. Lai, J. Wen, H. Kong, Uncertainty-guided pixel contrastive learning for semi-supervised medical image segmentation, in: Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI, 2022, pp. 1444–1450

    2022

  32. [40]

    L. Lu, M. Yin, L. Fu, F. Yang, Uncertainty-aware pseudo-label and consistency for semi-supervised medical image segmentation, Biomedical Signal Processing and Control 79 (2023) 104203

    2023

  33. [41]

    Peiris, M

    H. Peiris, M. Hayat, Z. Chen, G. Egan, M. Harandi, Uncertainty-guided dual-views for semi-supervised vol- umetric medical image segmentation, Nature Machine Intelligence 5 (7) (2023) 724–738

    2023

  34. [42]

    K. Wang, B. Zhan, C. Zu, X. Wu, J. Zhou, L. Zhou, Y. Wang, Semi-supervised medical image segmentation via a tripled-uncertainty guided mean teacher model with contrastive learning, Medical Image Analysis 79 (2022) 102447

    2022

  35. [43]

    Y. Shi, J. Zhang, T. Ling, J. Lu, Y. Zheng, Q. Yu, L. Qi, Y. Gao, Inconsistency-aware uncertainty estimation for semi-supervised medical image segmentation, IEEE Trans. Medical Imaging 41 (3) (2022) 608–620.doi: 10.1109/TMI.2021.3117888. URLhttps://doi.org/10.1109/TMI.2021.3117888

    2022

  36. [44]

    Y. Li, J. Chen, X. Xie, K. Ma, Y. Zheng, Self-loop uncertainty: A novel pseudo-label for semi-supervised medical image segmentation, in: A. L. Martel, P . Abolmaesumi, D. Stoyanov, D. Mateus, M. A. Zuluaga, S. K. Zhou, D. Racoceanu, L. Joskowicz (Eds.), Medical Image Computing...

    2020

  37. [45]

    W. Wu, C. Zhang, P . C. Woodland, Estimating the uncertainty in emotion attributes using deep evidential regression, in: A. Rogers, J. L. Boyd-Graber, N. Okazaki (Eds.), Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pap...

    2023 doi

  38. [47]

    Sensoy, L

    M. Sensoy, L. Kaplan, M. Kandemir, Evidential deep learning to quantify classification uncertainty, Advances in neural information processing systems 31 (2018)

    2018

  39. [48]

    Amini, W

    A. Amini, W. Schwarting, A. Soleimany, D. Rus, Deep evidential regression, Advances in Neural Information Processing Systems 33 (2020) 14927–14937

    2020

  40. [49]

    Z. Han, C. Zhang, H. Fu, J. T. Zhou, Trusted multi-view classification with dynamic evidential fusion, IEEE Trans. Pattern Anal. Mach. Intell. 45 (2) (2023) 2551–2566.doi:10.1109/TPAMI.2022.3171983. URLhttps://doi.org/10.1109/TPAMI.2022.3171983

    2023

  41. [50]

    D. Deng, G. Chen, Y. Yu, F. Liu, P .-A. Heng, Uncertainty estimation by fisher information-based eviden- tial deep learning, in: Proceedings of the 40th International Conference on Machine Learning, ICML’23, JMLR.org, 2023

    2023

  42. [51]

    Jøsang, Subjective Logic - A Formalism for Reasoning Under Uncertainty, Artificial Intelligence: Founda- tions, Theory, and Algorithms, Springer, 2016.doi:10.1007/978-3-319-42337-1

    A. Jøsang, Subjective Logic - A Formalism for Reasoning Under Uncertainty, Artificial Intelligence: Founda- tions, Theory, and Algorithms, Springer, 2016.doi:10.1007/978-3-319-42337-1. URLhttps://doi.org/10.1007/978-3-319-42337-1

    2016 doi

  43. [52]

    A. P . Dempster, Upper and lower probabilities induced by a multivalued mapping, in: R. R. Yager, L. Liu (Eds.), Classic Works of the Dempster-Shafer Theory of Belief Functions, Vol. 219 of Studies in Fuzziness and Soft Computing, Springer, 2008, pp. 57–72.doi:10.1007/978-3-54...

    2008 doi

  44. [53]

    Shafer, A mathematical theory of evidence turns 40, Int

    G. Shafer, A mathematical theory of evidence turns 40, Int. J. Approx. Reason. 79 (2016) 7–25.doi:10.1016/ j.ijar.2016.07.009. URLhttps://doi.org/10.1016/j.ijar.2016.07.009

    2016 doi

  45. [54]

    Smets, Constructing the pignistic probability function in a context of uncertainty, in: M

    P . Smets, Constructing the pignistic probability function in a context of uncertainty, in: M. Henrion, R. D. Shachter, L. N. Kanal, J. F. Lemmer (Eds.), UAI ’89: Proceedings of the Fifth Annual Conference on Uncer- tainty in Artificial Intelligence, Windsor, Ontario, Canada, ...

    1989

  46. [55]

    C. E. Shannon, A mathematical theory of communication, Bell Syst. Tech. J. 27 (3) (1948) 379–423.doi: 10.1002/J.1538-7305.1948.TB01338.X. URLhttps://doi.org/10.1002/j.1538-7305.1948.tb01338.x

    1948

  47. [56]

    Deng, Uncertainty measure in evidence theory, Sci

    Y. Deng, Uncertainty measure in evidence theory, Sci. China Inf. Sci. 63 (11) (2020).doi:10.1007/ S11432-020-3006-9. URLhttps://doi.org/10.1007/s11432-020-3006-9

    2020 doi

  48. [57]

    Yong, Information volume of mass function, International Journal of Computers Communications & Con- trol 15 (6) (2020)

    D. Yong, Information volume of mass function, International Journal of Computers Communications & Con- trol 15 (6) (2020). 28

    2020

  49. [59]

    M. Chen, J. Gao, C. Xu, Cascade evidential learning for open-world weakly-supervised temporal action local- ization, in: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023, IEEE, 2023, pp. 14741–14750.doi:10.1109/...

    2023

  50. [60]

    Xiong, Q

    Z. Xiong, Q. Xia, Z. Hu, N. Huang, C. Bian, Y. Zheng, S. Vesal, N. Ravikumar, A. Maier, X. Yang, et al., A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging, Medical image analysis 67 (2021) 101832

    2021

  51. [61]

    Y. Wu, Z. Wu, Q. Wu, Z. Ge, J. Cai, Exploring smoothness and class-separation for semi-supervised medical image segmentation, in: L. Wang, Q. Dou, P . T. Fletcher, S. Speidel, S. Li (Eds.), Medical Image Computing and Computer Assisted Intervention - MICCAI 2022 - 25th Interna...

    2022 doi

  52. [62]

    H. R. Roth, L. Lu, A. Farag, H.-C. Shin, J. Liu, E. B. Turkbey, R. M. Summers, Deeporgan: Multi-level deep convolutional networks for automated pancreas segmentation, in: Medical Image Computing and Computer- Assisted Intervention–MICCAI 2015: 18th International Conference, Mu...

    2015

  53. [63]

    Bernard, A

    O. Bernard, A. Lalande, C. Zotti, F. Cervenansky, X. Yang, P .-A. Heng, I. Cetin, K. Lekadir, O. Camara, M. A. G. Ballester, et al., Deep learning techniques for automatic mri cardiac multi-structures segmentation and diag- nosis: is the problem solved?, IEEE transactions on m...

    2018

  54. [64]

    Antonelli, A

    M. Antonelli, A. Reinke, S. Bakas, K. Farahani, A. Kopp-Schneider, B. A. Landman, G. Litjens, B. Menze, O. Ronneberger, R. M. Summers, et al., The medical segmentation decathlon, Nature communications 13 (1) (2022) 4128

    2022

  55. [65]

    L. Yu, S. Wang, X. Li, C. Fu, P . Heng, Uncertainty-aware self-ensembling model for semi-supervised 3d left atrium segmentation, in: D. Shen, T. Liu, T. M. Peters, L. H. Staib, C. Essert, S. Zhou, P . Yap, A. R. Khan (Eds.), Medical Image Computing and Computer Assisted Interv...

    2019 doi

  56. [66]

    S. Li, C. Zhang, X. He, Shape-aware semi-supervised 3d semantic segmentation for medical images, in: A. L. Martel, P . Abolmaesumi, D. Stoyanov, D. Mateus, M. A. Zuluaga, S. K. Zhou, D. Racoceanu, L. Joskowicz (Eds.), Medical Image Computing and Computer Assisted Intervention ...

    2020 doi

  57. [67]

    X. Luo, J. Chen, T. Song, G. Wang, Semi-supervised medical image segmentation through dual-task con- sistency, in: Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applications of Artificial Intelligence, IAAI 2021, The ...

    2021 doi

  58. [68]

    X. Luo, W. Liao, J. Chen, T. Song, Y. Chen, S. Zhang, N. Chen, G. Wang, S. Zhang, Efficient semi-supervised gross target volume of nasopharyngeal carcinoma segmentation via uncertainty rectified pyramid consis- tency, in: M. de Bruijne, P . C. Cattin, S. Cotin, N. Padoy, S. Sp...

    2021

  59. [69]

    Y. Wu, M. Xu, Z. Ge, J. Cai, L. Zhang, Semi-supervised left atrium segmentation with mutual consistency training, in: M. de Bruijne, P . C. Cattin, S. Cotin, N. Padoy, S. Speidel, Y. Zheng, C. Essert (Eds.), Medical Image Computing and Computer Assisted Intervention - MICCAI 2...

    2021 doi

  60. [70]

    Zhang, H

    Z. Zhang, H. Zhou, X. Shi, R. Ran, C. Tian, F. Zhou, An evidential-enhanced tri-branch consistency learning method for semi-supervised medical image segmentation, arXiv preprint arXiv:2404.07032 (2024)

    2024 arXiv

  61. [71]

    Y. Chen, Z. Yang, C. Shen, Z. Wang, Y. Qin, Y. Zhang, Evil: Evidential inference learning for trustworthy semi-supervised medical image segmentation, in: 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI), IEEE, 2023, pp. 1–5

    2023

  62. [72]

    Huang, S

    L. Huang, S. Ruan, T. Denoeux, Belief function-based semi-supervised learning for brain tumor segmentation, in: 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), IEEE, 2021, pp. 160–164

    2021

  63. [73]

    K. Wang, C. Zhang, Y. Geng, H. Ma, Evidential pseudo-label ensemble for semi-supervised classification, Pattern Recognition Letters 177 (2024) 135–141

    2024

  64. [74]

    Y. Chen, Z. Yang, C. Shen, Z. Wang, Z. Zhang, Y. Qin, X. Wei, J. Lu, Y. Liu, Y. Zhang, Evidence-based uncertainty-aware semi-supervised medical image segmentation, Computers in Biology and Medicine 170 (2024) 108004. 30

    2024

Pith tools

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