Pith. sign in

REVIEW 4 major objections 5 minor 67 references

UD-Mamba: A pixel-level uncertainty-driven Mamba model for medical image segmentation

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read UD-Mamba redefines Mamba's pixel scan order using per-pixel channel uncertainty, reporting higher Dice and IoU than Mamba-UNet and Swin-UMamba on pathology, skin-lesion, and cardiac MRI datasets while using fewer FLOPs.

desk verdict Readable incremental Mamba segmentation paper with a promising uncertainty-sorting idea, but the core mechanism is never isolated and the ablation has internal contradictions. read the letter →

arxiv 2502.02024 v1 pith:TIADQT66 submitted 2025-02-04 eess.IV cs.CV

Weiren Zhao , Feng Wang , Yanran Wang , Yutong Xie , Qi Wu , Yuyin Zhou This is my paper · ORCID
classification eess.IVcs.CV
keywords medicalimagesegmentationMambastatespacemodelsuncertainty-drivenscanningpixel-leveluncertaintyselectivechannelDicescore
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

UD-Mamba claims that the order in which a Mamba state-space model scans image pixels can be driven by pixel-level channel uncertainty rather than by fixed spatial location. For each spatial position it computes the standard deviation across feature channels, sorts the feature map so high-uncertainty pixels (foreground and boundaries) come first, and then applies two scanning patterns, sequential and skip, in both high-to-low and low-to-high directions. Four learned scalars reweight the four scans, and a cosine consistency loss aligns features from opposite directions. On DigestPath, ISIC 2018, and ACDC, the resulting model reports higher Dice and IoU than Mamba-UNet and Swin-UMamba while using fewer FLOPs. The paper's argument matters because it suggests a cheap, parameter-light way to make long-range sequence models attend to diagnostically important pixels.

What carries the argument

The load-bearing object is UD-SSM, an Uncertainty-Driven Selective Scanning Module. It computes a per-pixel uncertainty map $U_{h,w} = \sqrt{\frac{1}{C}\sum_c (X_{c,h,w}-\mu_{h,w})^2}$ as the standard deviation across channels, sorts the feature map by $U$, and then performs four scans: sequential and skip, each in high-to-low and low-to-high order. Four learnable parameters $\alpha_1,\dots,\alpha_4$ reweight the four scan outputs before the S6 selective state-space block, and a recovery step restores spatial positions. A cosine consistency loss $L_{cos}$ then aligns the forward and backward scan features. This machinery is what replaces the fixed location-based scanning of prior Mamba vision models.

What would settle it

Train the full UD-Mamba pipeline on the same datasets but replace the uncertainty-sorted order with a random permutation while keeping the four scans, reweighting, and cosine loss; if the Dice score does not drop relative to the uncertainty-sorted model, then the uncertainty ordering is not what drives the reported gains.

Watch

Extended reading notes

Core claim

The central discovery is that reordering a Mamba scan by channel uncertainty, specifically by the standard deviation of each pixel's feature vector across channels, improves medical image segmentation over location-based scanning. The authors treat the channel-standard-deviation map as an uncertainty map in which high uncertainty marks boundaries and foreground objects and low uncertainty marks background. Sorting by this map and scanning the sorted sequence lets the selective state-space model process critical regions contiguously, while skip scanning at fixed uncertainty intervals mixes background context into the foreground-dense sequence. The reported gains, Dice of 80.89 on DigestPath, 89.15 on ISIC 2018, and 91.99 on ACDC with HD95 of 1.31 mm on ACDC, are attributed to this uncertainty-driven ordering together with the four-scan reweighting and the cosine consistency loss.

Load-bearing premise

The load-bearing premise is that a pixel's standard deviation across feature channels marks the pixels that matter for segmentation, namely boundaries and foreground, so sorting by it makes the scan more useful; the paper tests the complete pipeline, not this correlation directly.

Editorial extensions

If this is right

  • If the ordering is the cause of the gains, any sequence-model architecture that processes pixels in a one-dimensional order can adopt the same uncertainty-sorted scan without changing its parameters or loss.
  • Pixel-level uncertainty granularity is preferable to region-based blocks: the paper's ablation reports that $1\times1$ uncertainty blocks outperform $2\times2$, $4\times4$, and $8\times8$ blocks on DigestPath.
  • Standard deviation is a better uncertainty estimator than mean absolute deviation, variance, entropy, or range for this purpose, according to the paper's ablation on DigestPath.
  • The four-scan ensemble with reweighting and cosine consistency adds value beyond any single scan direction, since combining all four scans raises DSC from 78.15 (vanilla position-based) to 80.32 and the final pipeline to 80.89.
  • The model's computational advantage, 19.12M parameters and 5.91G FLOPs on ACDC, below all compared transformer and Mamba baselines, makes the approach practical for high-resolution medical images.

Reading between the lines

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

  • The paper does not show that the channel-standard-deviation map correlates with ground-truth boundaries, so the reordering's unique contribution is untested on its own; a random-order control would isolate it from the gains of the four-scan ensemble.
  • The same uncertainty map could serve as a cheap boundary-emphasis signal outside Mamba, for example as an auxiliary supervision target or an attention prior in CNN and transformer segmentation heads.
  • Because the four directional scans are reweighted by learned scalars, the method may be robust to the direction choice; the observed smaller decay of $\alpha_3$ and $\alpha_4$ hints that the low-to-high scans carry more weight, but this is a post-hoc observation from Figure 6, not a tested claim.
  • A natural extension is to apply UD-SSM to 3D medical volumes by treating slices or patches as the scan sequence, though the paper only reports 2D experiments.
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 / 5 minor

Summary. The paper proposes UD-Mamba, a U-Net-like Mamba architecture for medical image segmentation in which the pixel scan order is determined by a per-pixel uncertainty score, defined as the standard deviation across feature channels (Eq. 4). The feature map is reordered in descending uncertainty (Eqs. 5-6) and processed by four scanning streams: sequential and skip scans in both high-to-low and low-to-high uncertainty directions (Eqs. 7-10), reweighted by four learnable scalars, plus a cosine consistency loss between opposite directions. Experiments on DigestPath, ISIC 2018, and ACDC report DSC values of 80.89, 89.15, and 91.99, respectively, outperforming the compared Mamba baselines with lower FLOPs. The paper also ablates the scanning components, the uncertainty metric, and the uncertainty region size.

Significance. The core idea of using an information-density or uncertainty criterion to define Mamba's scan order, rather than a fixed spatial raster, is a reasonable and potentially useful direction for selective state-space models in dense prediction, and the reported computational efficiency (19.12M parameters, 5.91G FLOPs on ACDC) is attractive. The manuscript is generally clearly written, gives the architecture and losses in sufficient detail, and provides code. If the mechanism were properly isolated, the work could be a useful contribution to the Mamba-for-medical-imaging literature. As it stands, however, the evidence does not show that the channel-standard-deviation ordering is what produces the gains; the individual ablation rows in Table 4 in fact suggest that uncertainty ordering alone hurts performance, and the authors' own conclusion in Section 5 states that performance depends heavily on accurate channel uncertainty estimation. The central mechanistic claim therefore needs substantially stronger experimental support.

major comments (4)
  1. [§4.4, Table 4 and Eqs. (4)-(6)] The paper's central claim is that sorting pixels by channel standard deviation places boundaries and foreground before background and that this reordering drives the reported gains. The experiments never isolate this mechanism. Compared with the position-scanning baseline (first row of Table 4, DSC 78.15), the full model changes the scan order, replaces the single bidirectional location scan with four uncertainty-ordered scans, adds four learnable reweighting scalars, and adds a cosine consistency loss. Please add controls that keep everything else fixed while varying only the ordering rule: a random permutation, an anti-correlated permutation, and an oracle ordering based on the ground-truth boundary map, and also compare a single uncertainty-ordered scan against a single bidirectional location scan. Without these controls, the observed improvement cannot be attributed to Eq. 4.
  2. [Table 4, rows 2-5] The individual uncertainty-ordered components underperform the vanilla baseline: y1 alone gives DSC 75.67 and y2 alone gives 76.87, both below 78.15; even y1+y3 (76.27) and y2+y4 (77.90) are below the baseline. Only the four-way sum (80.32) exceeds it. This pattern is in tension with the claim that prioritizing high-uncertainty pixels improves segmentation precision, and it points instead to the four-scan ensemble as the source of the gain. Please analyze the contribution of each factor and report the performance of each scan direction with the same reweighting and consistency mechanisms applied.
  3. [§4.4, 'Analysis for re-weighting values', Fig. 6, and Eqs. (7)-(10), (14)] The text says alpha3 and alpha4 correspond to scanning from high to low uncertainty, but Eqs. (7)-(10) define y1/y2 as high-to-low and y3/y4 as low-to-high. The surrounding text is also inconsistent about which direction is preferable: Section 3.3 says the method enhances the advantages of low-to-high uncertainty scanning (before Eq. 11), Fig. 2's caption says ascending (low-to-high) outperforms descending, and the conclusion in Section 5 says low-to-high typically yields greater benefits than the reverse, while the re-weighting analysis claims high-to-low contributes more. Please correct the labels and provide a consistent, quantitative account of the directionality result.
  4. [Tables 1, 2, 4, and 5] All results are point estimates without standard deviations, confidence intervals, or significance tests. The reported differences over the strongest baselines are small (e.g., 0.91% DSC on ACDC in Table 2), so run-to-run variance is a real concern. In addition, the choice of STD as the uncertainty metric in Table 5 is made by comparing alternatives on the same test sets used for the headline results; this is a form of test-set selection. Please report mean ± std over multiple seeds and choose the uncertainty metric on a validation split.
minor comments (5)
  1. [Table 4, y2 row] The ACC value is listed as 64.67, which is far outside the plausible range and inconsistent with the other rows; this is likely a typo for 94.67.
  2. [Eq. (14)] The expression '1− cos_sim(yr1, yr3) +cos_sim(yr2, yr4) 2' is mathematically ambiguous; it should be written as 1 - (cos_sim(y1,y3) + cos_sim(y2,y4)) / 2.
  3. [Table 5] The label 'Entroph' should be 'Entropy', and 'Range between the two highest values' should be defined precisely (e.g., max - min or the difference between the two largest channel activations).
  4. [§3.3, Skip scanning] The description of skip scanning as moving vertically across the image at consistent uncertainty intervals is vague; please specify whether the intervals are in sorted-index space or in pixel coordinates and how the fixed intervals are chosen.
  5. [Figure 2] Figure 2 reports no numeric values or error bars, so the claimed advantage of ascending over descending scanning cannot be assessed from the figure alone; please provide the underlying numbers or refer to Table 4 more precisely.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical segmentation results, not derivations that reduce to their inputs.

full rationale

UD-Mamba is an empirical computer-vision paper: its central claim is that pixel-level channel standard deviation (Eq. 4), used as a sorting key (Eq. 5), improves Mamba-based segmentation, and this is supported by measured Dice scores on three benchmarks rather than by a formal derivation. The load-bearing assumption that higher channel standard deviation 'typically corresponds to key regions, such as object boundaries or foreground regions' is stated as an observation, not derived from the definition of standard deviation, so it is not circular; it is an empirical premise that could be false. The four learnable reweighting parameters and the cosine consistency loss are trained and evaluated under standard supervision, and the paper does not relabel any fitted value as a prediction. The self-citations (e.g., Ref. 52 with overlapping authors, Ref. 61 by a co-author) appear only as background related-work references and are not load-bearing for the main architecture or results. The absence of controls such as random or oracle orderings in the ablation study (Table 5) weakens causal attribution but is a question of experimental design and correctness risk, not circularity. The internal inconsistency in Section 4.4 regarding which alpha values correspond to high-to-low versus low-to-high scanning is an error in interpretation, not a circular argument. The paper's conclusions are therefore self-contained in the sense that they report measured performance rather than deriving a result from its own assumptions.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a small set of modeling choices: channel standard deviation as the uncertainty signal, uncertainty-based pixel reordering, four scan paths, learnable scan weights, and a cosine consistency loss. The main unvalidated premise is that channel variance marks segmentation-relevant pixels. No new physical entities are introduced. The scan weights and loss weight are learned or hand-set, and the uncertainty metric was selected by ablation on the evaluation benchmark.

free parameters (4)
  • alpha_1 through alpha_4 (learnable scan weights) = learned during training; Figure 6 shows all four decreasing over time
    Four scalar multipliers combine the four scan sequences in Eq. 11; their learned values are part of the method and are dataset-dependent.
  • lambda (consistency loss weight) = 0.3
    Hand-set hyperparameter balancing the supervised loss and the cosine consistency loss in Eq. 15.
  • Uncertainty metric choice = standard deviation (STD)
    Selected after comparing MAD, range, entropy, and variance on DigestPath (Table 5); the best metric on the evaluation benchmark was adopted, making this a model-selection step.
  • Skip scanning interval = not specified
    The paper describes skip scanning as sampling at fixed intervals but never states the interval value, leaving an undocumented implementation choice that affects the method's behavior.
assumptions (5)
  • standard math Mamba's S6 selective scan provides linear-complexity long-range dependency modeling.
    Assumed from prior work [14]; the paper builds UD-SSM on top of this foundation.
  • domain assumption Channel standard deviation at a spatial position identifies foreground, boundary, and otherwise segmentation-critical pixels.
    This is the load-bearing premise of Eqs. 4 to 6 and Figure 1; the paper provides no direct validation beyond final segmentation metrics.
  • domain assumption Sorting and rearranging pixels by uncertainty, then recovering spatial positions, preserves the information needed for segmentation.
    Required for the Rearrange operation in Eq. 6 and the Recover operation in Eq. 12; residual connections may compensate, but the paper does not analyze this.
  • domain assumption The benchmark ground-truth labels are correct and representative.
    Standard public datasets are used for training and evaluation; no label-quality analysis is provided.
  • ad hoc to paper Cosine similarity between opposite scanning directions is a useful consistency objective.
    Introduced to align forward and backward scans; the only supporting evidence is the Table 4 ablation row with and without Lcos.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UD-Mamba: A pixel-level uncertainty-driven Mamba model for medical image segmentation." pith.science (2026). https://pith.science/paper/TIADQT66

@misc{pith2026250202024,
  author       = {Pith},
  title        = {Pith review of: UD-Mamba: A pixel-level uncertainty-driven Mamba model for medical image segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TIADQT66}},
  note         = {Machine review of arXiv:2502.02024}
}
read the original abstract

Recent advancements have highlighted the Mamba framework, a state-space model known for its efficiency in capturing long-range dependencies with linear computational complexity. While Mamba has shown competitive performance in medical image segmentation, it encounters difficulties in modeling local features due to the sporadic nature of traditional location-based scanning methods and the complex, ambiguous boundaries often present in medical images. To overcome these challenges, we propose Uncertainty-Driven Mamba (UD-Mamba), which redefines the pixel-order scanning process by incorporating channel uncertainty into the scanning mechanism. UD-Mamba introduces two key scanning techniques: 1) sequential scanning, which prioritizes regions with high uncertainty by scanning in a row-by-row fashion, and 2) skip scanning, which processes columns vertically, moving from high-to-low or low-to-high uncertainty at fixed intervals. Sequential scanning efficiently clusters high-uncertainty regions, such as boundaries and foreground objects, to improve segmentation precision, while skip scanning enhances the interaction between background and foreground regions, allowing for timely integration of background information to support more accurate foreground inference. Recognizing the advantages of scanning from certain to uncertain areas, we introduce four learnable parameters to balance the importance of features extracted from different scanning methods. Additionally, a cosine consistency loss is employed to mitigate the drawbacks of transitioning between uncertain and certain regions during the scanning process. Our method demonstrates robust segmentation performance, validated across three distinct medical imaging datasets involving pathology, dermatological lesions, and cardiac tasks.

Figures

Figures reproduced from arXiv: 2502.02024 by the authors.

Figure 1
Figure 1. Pixel-level channel uncertainty-based scanning mechanism. (a) Input im [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Illustration of the UD-Mamba ar￾chitecture, which includes a patch em￾bedding layer, an encoder-decoder with Uncertainty-Driven (UD) Blocks, and a segmentation head. Each UD Block fea￾tures the Uncertainty-Driven Selective Scanning Model (UD-SSM) for process￾ing input. on imaging methods such as CT [65,64,49,13] and MRI [23,57]. Among them, Convolutional Neural Networks (CNNs) and Transformers dominate as leading fr… view at source ↗
Figure 4
Figure 4. Detailed description of the UD-SSM. I. Describes the uncertainty calcu [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Visual comparisons of segmentation results from UD-Mamba and various [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Analysis of recorded values for four learnable reweighting parameters. [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 34 canonical work pages

  1. [1]

    In: ICLR (2024)

    Bao, Y., Sivanandan, S., Karaletsos, T.: Channel vision transformers: An image is worth 1 x 16 x 16 words. In: ICLR (2024)

    work page 2024

  2. [2]

    In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2019

    Baumgartner, C.F., Tezcan, K.C., Chaitanya, K., Hötker, A.M., Muehlematter, U.J., Schawkat, K., Becker, A.S., Donati, O., Konukoglu, E.: Phiseg: Capturing uncertainty in medical image segmentation. In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2019. pp. 119–127. Springer (2019)

    work page 2019

  3. [3]

    Zhao et al

    Bernard, O., Lalande, A., Zotti, C., Cervenansky, F., Yang, X., Heng, P.A., Cetin, I., Lekadir, K., Camara, O., Ballester, M.A.G., et al.: Deep learning techniques for 16 W. Zhao et al. automatic mri cardiac multi-structures segmentation and diagnosis: is the problem solved? IEEE transactions on medical imaging37(11), 2514–2525 (2018)

    work page 2018

  4. [4]

    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)

    2022

  5. [5]

    Medical Image Analysis p

    Chen, J., Mei, J., Li, X., Lu, Y., Yu, Q., Wei, Q., Luo, X., Xie, Y., Adeli, E., Wang, Y., et al.: Transunet: Rethinking the u-net architecture design for medical image segmentation through the lens of transformers. Medical Image Analysis p. 103280 (2024)

    work page 2024

  6. [6]

    arXiv preprint arXiv:1902.03368 (2019)

    Codella, N., Rotemberg, V., Tschandl, P., Celebi, M.E., Dusza, S., Gutman, D., Helba,B.,Kalloo,A.,Liopyris,K., Marchetti,M., et al.:Skinlesionanalysis toward melanoma detection 2018: A challenge hosted by the international skin imaging collaboration (isic). arXiv preprint arXiv:1902.03368 (2019)

    arXiv 2019

  7. [7]

    Medical Image Analysis 80, 102485 (2022)

    Da, Q., Huang, X., Li, Z., Zuo, Y., Zhang, C., Liu, J., Chen, W., Li, J., Xu, D., Hu, Z., et al.: Digestpath: A benchmark dataset with challenge review for the pathological detection and segmentation of digestive-system. Medical Image Analysis 80, 102485 (2022)

    work page 2022

  8. [8]

    arXiv preprint arXiv:2405.21060 (2024)

    Dao, T., Gu, A.: Transformers are ssms: Generalized models and efficient algo- rithms through structured state space duality. arXiv preprint arXiv:2405.21060 (2024)

    arXiv 2024

Show all 67 references
  1. [9]

    In: 2009 IEEE conference on computer vision and pattern recognition

    Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large- scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition. pp. 248–255. Ieee (2009)

    2009

  2. [10]

    ICLR (2021)

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An image is worth 16x16 words: Transformers for image recognition at scale. ICLR (2021)

    2021

  3. [11]

    arXiv preprint arXiv:2407.08481 (2024)

    Fan, C., Yu, H., Wang, L., Huang, Y., Wang, L., Jia, X.: Slicemamba for medical image segmentation. arXiv preprint arXiv:2407.08481 (2024)

    2024 arXiv

  4. [12]

    In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition

    Fan, J., Gao, B., Jin, H., Jiang, L.: Ucc: Uncertainty guided cross-head co-training for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition. pp. 9947–9956 (2022)

    2022

  5. [13]

    Physica Medica85, 107–122 (2021)

    Fu,Y.,Lei,Y.,Wang,T.,Curran,W.J.,Liu,T.,Yang,X.:Areviewofdeeplearning based methods for medical image multi-organ segmentation. Physica Medica85, 107–122 (2021)

    2021

  6. [14]

    arXiv preprint arXiv:2312.00752 (2023)

    Gu, A., Dao, T.: Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752 (2023)

    2023 arXiv

  7. [15]

    arXiv preprint arXiv:2111.00396 (2021)

    Gu, A., Goel, K., Ré, C.: Efficiently modeling long sequences with structured state spaces. arXiv preprint arXiv:2111.00396 (2021)

    2021 arXiv

  8. [16]

    In: Proceedings of the IEEE/CVF winter conference on applications of computer vi- sion

    Hatamizadeh, A., Tang, Y., Nath, V., Yang, D., Myronenko, A., Landman, B., Roth, H.R., Xu, D.: Unetr: Transformers for 3d medical image segmentation. In: Proceedings of the IEEE/CVF winter conference on applications of computer vi- sion. pp. 574–584 (2022)

    2022

  9. [17]

    IEEE Transactions on Medical Imaging 42(9), 2763–2775 (2023)

    He, A., Wang, K., Li, T., Du, C., Xia, S., Fu, H.: H2former: An efficient hierarchical hybrid transformer for medical image segmentation. IEEE Transactions on Medical Imaging 42(9), 2763–2775 (2023)

    2023

  10. [18]

    arXiv preprint arXiv:2404.06564 (2024)

    He, H., Bai, Y., Zhang, J., He, Q., Chen, H., Gan, Z., Wang, C., Li, X., Tian, G., Xie, L.: Mambaad: Exploring state space models for multi-class unsupervised anomaly detection. arXiv preprint arXiv:2404.06564 (2024)

    2024 arXiv

  11. [19]

    ECCV (2024) Title Suppressed Due to Excessive Length 17

    Hu, V.T., Baumann, S.A., Gui, M., Grebenkova, O., Ma, P., Fischer, J., Ommer, B.: Zigma: Zigzag mamba diffusion model. ECCV (2024) Title Suppressed Due to Excessive Length 17

    2024

  12. [20]

    arXiv preprint arXiv:2109.07162 (2021)

    Huang, X., Deng, Z., Li, D., Yuan, X.: Missformer: An effective medical image segmentation transformer. arXiv preprint arXiv:2109.07162 (2021)

    2021 arXiv

  13. [21]

    IEEE Transactions on Medical Imaging42(5), 1484–1494 (2022)

    Huang, X., Deng, Z., Li, D., Yuan, X., Fu, Y.: Missformer: An effective transformer for 2d medical image segmentation. IEEE Transactions on Medical Imaging42(5), 1484–1494 (2022)

    2022

  14. [22]

    Nature methods 18(2), 203–211 (2021)

    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 segmentation. Nature methods 18(2), 203–211 (2021)

    2021

  15. [23]

    Advances in neural information processing systems 35, 36722–36732 (2022)

    Ji, Y., Bai, H., Ge, C., Yang, J., Zhu, Y., Zhang, R., Li, Z., Zhanng, L., Ma, W., Wan, X., et al.: Amos: A large-scale abdominal multi-organ benchmark for versatile medical image segmentation. Advances in neural information processing systems 35, 36722–36732 (2022)

    2022

  16. [24]

    In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2019

    Jungo,A.,Reyes,M.:Assessingreliabilityandchallengesofuncertaintyestimations for medical image segmentation. In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2019. pp. 48–56. Springer (2019)

    2019

  17. [25]

    Kalman, R.E.: A new approach to linear filtering and prediction problems (1960)

    1960

  18. [26]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Le, V.L., Saut, O.: Rrc-unet 3d for lung tumor segmentation from ct scans of non- small cell lung cancer patients. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 2316–2325 (2023)

    2023

  19. [27]

    Neural Computing and Applications35(30), 22071–22085 (2023)

    Li, H., Nan, Y., Del Ser, J., Yang, G.: Region-based evidential deep learning to quantify uncertainty and improve robustness of brain tumor segmentation. Neural Computing and Applications35(30), 22071–22085 (2023)

    2023

  20. [28]

    ECCV (2024)

    Li, K., Li, X., Wang, Y., He, Y., Wang, Y., Wang, L., Qiao, Y.: Videomamba: State space model for efficient video understanding. ECCV (2024)

    2024

  21. [29]

    IEEE transactions on medical imaging37(12), 2663–2674 (2018)

    Li, X., Chen, H., Qi, X., Dou, Q., Fu, C.W., Heng, P.A.: H-denseunet: hybrid densely connected unet for liver and tumor segmentation from ct volumes. IEEE transactions on medical imaging37(12), 2663–2674 (2018)

    2018

  22. [30]

    IEEE Transactions on Instru- mentation and Measurement71, 1–15 (2022)

    Lin, A., Chen, B., Xu, J., Zhang, Z., Lu, G., Zhang, D.: Ds-transunet: Dual swin transformer u-net for medical image segmentation. IEEE Transactions on Instru- mentation and Measurement71, 1–15 (2022)

    2022

  23. [31]

    In: International Conference on Medical Image Computing and Computer- Assisted Intervention

    Liu, J., Yang, H., Zhou, H.Y., Xi, Y., Yu, L., Li, C., Liang, Y., Shi, G., Yu, Y., Zhang, S., et al.: Swin-umamba: Mamba-based unet with imagenet-based pretrain- ing. In: International Conference on Medical Image Computing and Computer- Assisted Intervention. pp. 615–625. Spri...

    2024

  24. [32]

    arXiv preprint arXiv:2401.10166 (2024)

    Liu, Y., Tian, Y., Zhao, Y., Yu, H., Xie, L., Wang, Y., Ye, Q., Liu, Y.: Vmamba: Visual state space model. arXiv preprint arXiv:2401.10166 (2024)

    2024 arXiv

  25. [33]

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer:Hierarchical visiontransformerusingshiftedwindows.In:Proceedings of the IEEE/CVF international conference on computer vision. pp. 10012–10022 (2021)

    2021

  26. [34]

    Biomedical Signal Processing and Control 79, 104203 (2023)

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

    2023

  27. [35]

    arXiv preprint arXiv:2401.04722 (2024)

    Ma, J., Li, F., Wang, B.: U-mamba: Enhancing long-range dependency for biomed- ical image segmentation. arXiv preprint arXiv:2401.04722 (2024)

    2024 arXiv

  28. [36]

    IEEE transactions on medical imaging39(12), 3868–3878 (2020)

    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 (2020)

    2020

  29. [37]

    Zhao et al

    Monteiro, M., Le Folgoc, L., Coelho de Castro, D., Pawlowski, N., Marques, B., Kamnitsas, K., van der Wilk, M., Glocker, B.: Stochastic segmentation networks: 18 W. Zhao et al. Modelling spatially correlated aleatoric uncertainty. Advances in neural informa- tion processing sy...

    2020

  30. [38]

    arXiv preprint arXiv:1804.03999 (2018)

    Oktay, O., Schlemper, J., Folgoc, L.L., Lee, M., Heinrich, M., Misawa, K., Mori, K., McDonagh, S., Hammerla, N.Y., Kainz, B., et al.: Attention u-net: Learning where to look for the pancreas. arXiv preprint arXiv:1804.03999 (2018)

    2018 arXiv

  31. [39]

    In: MICCAI

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomed- ical image segmentation. In: MICCAI. pp. 234–241. Springer (2015)

    2015

  32. [40]

    arXiv preprint arXiv:2402.02491 (2024)

    Ruan, J., Xiang, S.: Vm-unet: Vision mamba unet for medical image segmentation. arXiv preprint arXiv:2402.02491 (2024)

    2024 arXiv

  33. [41]

    IEEE transactions on medical imaging41(3), 608–620 (2021)

    Shi, Y., Zhang, J., Ling, T., Lu, J., Zheng, Y., Yu, Q., Qi, L., Gao, Y.: Inconsistency-aware uncertainty estimation for semi-supervised medical image seg- mentation. IEEE transactions on medical imaging41(3), 608–620 (2021)

    2021

  34. [42]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Tang, H., Liu, X., Sun, S., Yan, X., Xie, X.: Recurrent mask refinement for few- shot medical image segmentation. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 3918–3928 (2021)

    2021

  35. [43]

    In: NeurIPS

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, L.u., Polosukhin, I.: Attention is all you need. In: NeurIPS. vol. 30 (2017)

    2017

  36. [44]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, F., Zheng, K., Lu, L., Xiao, J., Wu, M., Miao, S.: Automatic vertebra local- ization and identification in ct by spine rectification and anatomically-constrained optimization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5280...

    2021

  37. [45]

    arXiv preprint arXiv:2405.14858 (2024)

    Wang, F., Wang, J., Ren, S., Wei, G., Mei, J., Shao, W., Zhou, Y., Yuille, A., Xie, C.: Mamba-r: Vision mamba also needs registers. arXiv preprint arXiv:2405.14858 (2024)

    2024 arXiv

  38. [46]

    Neurocomputing338, 34–45 (2019)

    Wang, G., Li, W., Aertsen, M., Deprest, J., Ourselin, S., Vercauteren, T.: Aleatoric uncertainty estimation with test-time augmentation for medical image segmenta- tion with convolutional neural networks. Neurocomputing338, 34–45 (2019)

    2019

  39. [47]

    In: ICASSP 2022-2022 IEEE international conference on acoustics, speech and signal processing (ICASSP)

    Wang, H., Xie, S., Lin, L., Iwamoto, Y., Han, X.H., Chen, Y.W., Tong, R.: Mixed transformer u-net for medical image segmentation. In: ICASSP 2022-2022 IEEE international conference on acoustics, speech and signal processing (ICASSP). pp. 2390–2394. IEEE (2022)

    2022

  40. [48]

    arXiv preprint arXiv:2403.07332 (2024)

    Wang, J., Chen, J., Chen, D., Wu, J.: Large window-based mamba unet for med- ical image segmentation: Beyond convolution and self-attention. arXiv preprint arXiv:2403.07332 (2024)

    2024 arXiv

  41. [49]

    Medical image analysis55, 88–102 (2019)

    Wang, Y., Zhou, Y., Shen, W., Park, S., Fishman, E.K., Yuille, A.L.: Abdomi- nal multi-organ segmentation with organ-attention networks and statistical fusion. Medical image analysis55, 88–102 (2019)

    2019

  42. [50]

    In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV)

    Wang, Y., Peng, J., Zhang, Z.: Uncertainty-aware pseudo label refinery for domain adaptive semantic segmentation. In: 2021 IEEE/CVF International Conference on Computer Vision (ICCV). pp. 9072–9081 (2021).https://doi.org/10.1109/ ICCV48922.2021.00896

    2021

  43. [51]

    arXiv preprint arXiv:2402.05079 (2024)

    Wang, Z., Zheng, J.Q., Zhang, Y., Cui, G., Li, L.: Mamba-unet: Unet-like pure visual mamba for medical image segmentation. arXiv preprint arXiv:2402.05079 (2024)

    2024 arXiv

  44. [52]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Wei, Q., Yu, L., Li, X., Shao, W., Xie, C., Xing, L., Zhou, Y.: Consistency-guided meta-learning for bootstrapping semi-supervised medical image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 183–193. Springer (2023)

    2023

  45. [53]

    Medical image analysis65, 101766 (2020) Title Suppressed Due to Excessive Length 19

    Xia, Y., Yang, D., Yu, Z., Liu, F., Cai, J., Yu, L., Zhu, Z., Xu, D., Yuille, A., Roth, H.: Uncertainty-aware multi-view co-training for semi-supervised medical image segmentation and domain adaptation. Medical image analysis65, 101766 (2020) Title Suppressed Due to Excessive ...

    2020

  46. [54]

    arXiv preprint arXiv:2401.13560 (2024)

    Xing,Z.,Ye,T.,Yang,Y.,Liu,G.,Zhu,L.:Segmamba:Long-rangesequentialmod- eling mamba for 3d medical image segmentation. arXiv preprint arXiv:2401.13560 (2024)

    2024 arXiv

  47. [55]

    arXiv preprint arXiv:2403.06800 (2024)

    Yang, S., Wang, Y., Chen, H.: Mambamil: Enhancing long sequence modeling with sequence reordering in computational pathology. arXiv preprint arXiv:2403.06800 (2024)

    2024 arXiv

  48. [56]

    arXiv preprint arXiv:2402.08506 (2024)

    Ye, Z., Chen, T.: P-mamba: Marrying perona malik diffusion with mamba for efficient pediatric echocardiographic left ventricular segmentation. arXiv preprint arXiv:2402.08506 (2024)

    2024 arXiv

  49. [57]

    Frontiers in Neuroinformatics 14, 610967 (2020)

    Zeng, C., Gu, L., Liu, Z., Zhao, S.: Review of deep learning approaches for the seg- mentation of multiple sclerosis lesions on brain mri. Frontiers in Neuroinformatics 14, 610967 (2020)

    2020

  50. [58]

    arXiv preprint arXiv:2406.10700 (2024)

    Zhang, G., Fan, L., He, C., Lei, Z., Zhang, Z., Zhang, L.: Voxel mamba: Group- free state space models for point cloud based 3d object detection. arXiv preprint arXiv:2406.10700 (2024)

    2024 arXiv

  51. [59]

    Artificial Intel- ligence in Medicine138, 102476 (2023)

    Zhang, Y., Jiao, R., Liao, Q., Li, D., Zhang, J.: Uncertainty-guided mutual con- sistency learning for semi-supervised medical image segmentation. Artificial Intel- ligence in Medicine138, 102476 (2023)

    2023

  52. [60]

    In: ICLR (2024)

    Zhang, Y., Wei, L., Freris, N.: Synergistic patch pruning for vision transformer: Unifying intra-& inter-layer patch importance. In: ICLR (2024)

    2024

  53. [61]

    In: 2024 IEEE International Symposium on Biomedical Imaging (ISBI)

    Zhao, W., Zhong, L., Wang, G.: Semi-contrans: Semi-supervised medical image segmentation via multi-scale feature fusion and cross teaching of cnn and trans- former. In: 2024 IEEE International Symposium on Biomedical Imaging (ISBI). pp. 1–5. IEEE (2024)

    2024

  54. [62]

    International Journal of Computer Vision 129(4), 1106–1120 (2021)

    Zheng, Z., Yang, Y.: Rectifying pseudo label learning via uncertainty estimation for domain adaptive semantic segmentation. International Journal of Computer Vision 129(4), 1106–1120 (2021)

    2021

  55. [63]

    IEEE Transactions on Image Processing (2023)

    Zhou, H.Y., Guo, J., Zhang, Y., Han, X., Yu, L., Wang, L., Yu, Y.: nnformer: Volumetric medical image segmentation via a 3d transformer. IEEE Transactions on Image Processing (2023)

    2023

  56. [64]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Zhou, Y., Li, Z., Bai, S., Wang, C., Chen, X., Han, M., Fishman, E., Yuille, A.L.: Prior-aware neural network for partially-supervised multi-organ segmentation. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 10672–10681 (2019)

    2019

  57. [65]

    In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV)

    Zhou, Y., Wang, Y., Tang, P., Bai, S., Shen, W., Fishman, E., Yuille, A.: Semi- supervised 3d abdominal multi-organ segmentation via deep multi-planar co- training. In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV). pp. 121–140. IEEE (2019)

    2019

  58. [66]

    IEEE TMI39(6), 1856–1867 (2019)

    Zhou, Z., Siddiquee, M.M.R., Tajbakhsh, N., Liang, J.: Unet++: Redesigning skip connections to exploit multiscale features in image segmentation. IEEE TMI39(6), 1856–1867 (2019)

    2019

  59. [67]

    ICML (2024)

    Zhu, L., Liao, B., Zhang, Q., Wang, X., Liu, W., Wang, X.: Vision mamba: Efficient visual representation learning with bidirectional state space model. ICML (2024)

    2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.