Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

LMS-Net: A Learned Mumford-Shah Network For Few-Shot Medical Image Segmentation

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

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

Pith's one-line read A network built by unfolding the iterations of a learned Mumford-Shah model beats prototypical few-shot baselines on three public medical segmentation datasets, with mean Dice gains of up to 4.14 points.

desk verdict A reasonable deep-unfolding FSS network that overstates its variational credentials; the architecture is worth a look, the derivation claim is not established. read the letter →

arxiv 2502.05473 v1 pith:56QOVKAK submitted 2025-02-08 cs.CV

Shengdong Zhang , Fan Jia , Xiang Li , Hao Zhang , Jun Shi , Liyan Ma , Shihui Ying This is my paper · ORCID
classification cs.CV
keywords few-shotsemanticsegmentationmedicalimagedeepunfoldingnetworkMumford-Shahmodeldenoisingpriorprimal-dualalgorithmprototypicalinterpretable
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

LMS-Net claims that few-shot medical image segmentation improves when the network is built by 'unfolding' the iterations of a learned Mumford-Shah variational model instead of by stacking ad-hoc segmentation modules. The paper argues that pixel-to-prototype comparison, the core of existing prototypical few-shot methods, is captured by the data-fidelity term of a Mumford-Shah energy, and that the model's regularizer can be reinterpreted as a denoising subproblem handled by a small CNN. From this formulation it derives a prototype-update step and a mask-update primal-dual step, then maps each step to a network module, giving a network whose components have a stated optimization meaning. On three public medical datasets (Synapse-CT, CHAOS-T2, MS-CMRSeg) the paper reports mean Dice gains of up to 4.14 points over ADNet and up to 3.00 points over RPT. If correct, this shows that variational priors and deep priors can be combined in few-shot segmentation with both better accuracy and a clearer explanation of what each module does.

What carries the argument

The central object is the Learned Mumford-Shah (LMS) energy, equation (4): minimizing over masks u in the simplex and prototypes l the sum of a pixel-to-prototype cosine cost u_i * rho(l_i, x), an entropy regularization term (1/alpha) u_i ln u_i, and a deep regularizer R(u_i). The argument runs on the alternating split: fixing l makes the prototype update a masked average pooling operation, while fixing u makes the mask update a primal-dual saddle-point problem. The load-bearing identity is the Moreau decomposition in equation (13), which turns the dual update into a proximal step interpreted as denoising; that proximal operator is then replaced by a five-layer CNN (Mask Denoiser). The Momentum Update Transformer and PD-Net are the unfolded versions of these two subproblems, and this correspondence is what the authors mean by interpretability.

What would settle it

Train PD-Net with the order of its primal and dual updates reversed (or with an arbitrary same-capacity CNN replacing the Mask Denoiser) and check on Synapse-CT whether the Dice scores change; if performance is unchanged, the unfolded-algorithm structure carries no explanatory weight. A more direct check is to compute the energy E(u^k,v^k) from equation (9) on test images across the K=2 stages and verify that it decreases; if the iterates do not lower the LMS energy, the network is not minimizing the claimed model.

Watch

Extended reading notes

Core claim

The central claim is that a learned Mumford-Shah energy with a latent-space data term and an unspecified deep regularizer R can be solved by alternating a momentum-based prototype update with a primal-dual mask update, and that unfolding these iterations yields a network that segments unseen medical structures from one labeled example. The authors cast the classical two-phase piecewise-constant Mumford-Shah functional in latent feature space, replacing pixel intensities with cosine similarity between query features and prototypes and replacing total variation with a deep prior. They split the energy into a prototype subproblem with closed-form masked average pooling and a mask subproblem whose entropy-regularized data term has a softmax closed form while the prior subproblem becomes a Gaussian denoising task via Moreau decomposition. Each of these steps becomes a module--MAP, Momentum Update Transformer, and PD-Net with a Mask Denoiser--so the network is, by construction, an approximation of the alternating minimization. The paper's evidence is the reported DSC improvements and iteration visualizations showing masks getting cleaner as the unfolded iterations progress.

Load-bearing premise

The main load-bearing premise is that the unknown regularizer R in equation (4) actually has a Fenchel conjugate and obeys the Moreau decomposition used in equations (9) and (13), and that the shallow CNN Mask Denoiser really approximates the proximal operator of R; the paper neither specifies R nor proves that equivalence, so if the CNN is just a learned refinement net, the network is not literally solving the stated variational model even if it segments well.

Editorial extensions

If this is right

  • If the central claim is correct, few-shot medical segmentation improves with a variational-model-derived architecture: fLMS-Net (fixed prototypes) beats ADNet by 1.00 to 4.14 mean DSC, and LMS-Net beats RPT by 0.28 to 3.00 mean DSC across the three datasets.
  • The deep-prior idea transfers from image reconstruction to segmentation: the same proximal-operator-as-denoiser trick used in learned primal-dual reconstruction becomes a mask-refinement module, so future FSS methods can plug a denoiser into their mask-refinement loop.
  • Because fLMS-Net (prototype update disabled) already outperforms ADNet, the mask-refinement block alone drives much of the gain, suggesting the data-fidelity/denoiser split is useful even when prototypes are held fixed.
  • The LMS model framework is not restricted to few-shot segmentation: the paper states it can be unfolded for broader semantic segmentation with task-specific prototype initialization.

Reading between the lines

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

  • The interpretability claim is stronger than what the mathematics supports: because R is unspecified and the Mask Denoiser is a non-convex CNN, the network modules are best read as algorithm-inspired skeletons rather than exact discretizations of the model; I would test the equivalence directly before relying on it in a clinical safety case.
  • MUT is an architectural addition beyond the alternating minimization--equations (6)-(13) contain no transformer--so the 'every module corresponds to an iteration' narrative is approximate; the momentum mechanism is a learned heuristic layered on top of the derived steps.
  • A cheap testable extension of the paper's own logic is to use the PD-Net mask-refinement block as a drop-in decoder for other prototypical FSS baselines, which would isolate how much of the gain comes from the denoiser rather than from prototype updates.
  • The reported margins are established against only two baselines on three datasets; extending the comparison to recent transformer- and self-supervision-based FSS methods would tell whether the gains persist.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes LMS-Net, a deep unfolding network for few-shot medical image segmentation. It introduces a Learned Mumford-Shah (LMS) model with a cosine data fidelity term, entropy regularization, and an unspecified deep prior R, and claims to solve it via alternating prototype and mask updates. The iterative steps are unfolded into network modules: MAP for prototype update, MUT for momentum-based prototype refinement, and PD-Net for primal-dual mask refinement. Experiments on Synapse-CT, CHAOS-MRI, and CMR compare with ADNet and RPT, reporting mean DSC gains of 0.28 to 4.14 points and claiming superior accuracy and robustness.

Significance. If the derivation were rigorous, this would be a valuable example of an interpretable deep unfolding network for medical segmentation, connecting variational models with few-shot learning. The paper offers a complete architecture, ablations, and a promise of released code, which are positive elements. However, the mathematical claims underlying the 'derived from' narrative are not established, and the empirical evidence is weaker than the text asserts. The core interpretability contribution is therefore not yet supported, although the architecture may still have standalone value.

major comments (4)
  1. [Section III-A3, Eq. (8)] The claim that Eq. (8) solves the prototype subproblem (6) is not correct for the cosine data term (5). Minimizing the integral of u_i(x) rho(l_i, x) with rho(l_i, x) = -F_q(x) . l_i / (||F_q(x)|| ||l_i||) over l_i yields a minimizer proportional to the integral of u_i F_q / ||F_q||, not the integral of u_i F_q, unless the query features are L2-normalized. The paper never states that F_q is normalized, and ResNet-101 features are not unit-norm by default. Thus the MAP operator is not the closed-form solution of the prototype update step, and the claimed correspondence between the module and the model is broken.
  2. [Section III-A3, Eqs. (9)-(13)] The saddle-point reformulation (9) and the Moreau decomposition (13) require R to be a proper, convex, lower-semicontinuous functional. The regularizer R is never specified; the Mask Denoiser is a non-convex 5-layer ReLU CNN with a Sigmoid output. No argument or experiment is provided to show that the CNN approximates the proximal operator of any convex functional. Moreover, the training loss (22)-(23) does not include the LMS energy (4), so the network is not even optimizing the proposed model. Consequently, the statement that LMS-Net is 'derived from' and 'solves' the LMS model is unsupported; at best the architecture is inspired by the model.
  3. [Section IV-D, Table I] The text says 'fLMS-Net consistently outperforms ADNet' and 'LMS-Net consistently surpasses RPT,' but Table I contradicts this. On Abd-CT, fLMS-Net's right-kidney DSC is 68.15 versus ADNet's 79.06 (-10.91), and LMS-Net's liver DSC is 78.57 versus RPT's 82.57 (-4.00). On Abd-MRI, LMS-Net's spleen DSC is 74.97 versus RPT's 76.37 (-1.40). No error bars, confidence intervals, or statistical significance tests are reported, so the mean improvements of 0.28 to 4.14 points may not be reliable. The robustness claim is therefore overstated and should be reworded or supported with proper statistics.
  4. [Section IV-E3, Table II] The ablation claims that PD-Net improves performance by 0.56% and 1.27% on the mean DSC of Abd-CT and Abd-MRI, but per-organ results show that PD-Net degrades liver DSC on Abd-CT from 82.04 to 78.57 and slightly reduces liver DSC on Abd-MRI from 83.9 to 83.71. Reporting only mean improvements obscures these organ-specific negative effects. The discussion should acknowledge this mixed behavior and explain why the overall improvement is still meaningful.
minor comments (5)
  1. [References, [35]] Reference [35] contains a typo: 'Computer Cision and Pattern Recognition' should be 'Computer Vision and Pattern Recognition.'
  2. [Eq. (14)] The notation in Eq. (14) is confusing: the proximal operator prox_{1/δ_k R} is applied to u_i^k + (1/δ_k) v_i^{k-1}, but the definition writes the arg min over v_i; please clarify the variable substitution.
  3. [Section III-A3, Eq. (8)] The sentence 'Since scaling l does not affect the solution of this problem' is used to justify Eq. (8), but then β_i is introduced as a scaling factor; please clarify whether β_i is learned, fixed, or normalized away.
  4. [Section IV-B] The paper says it follows the 'self-supervised training approach' of [32], [34], but it does not explain how self-supervision is applied in this work; please provide details or remove the claim.
  5. [Fig. 6] Figure 6 reports the Dice Score for different numbers of stages K, but the plot lacks numerical labels or error bars; please add exact values and variance information.

Circularity Check

2 steps flagged · score 4.0 of 10

Empirical gains are independently benchmarked, but the 'derived from LMS model' claim is partly circular: R is defined by the CNN denoiser, and Eq. (8) relabels masked average pooling as a cosine-distance minimizer.

  1. self definitional [Section III-B.3, 'PD-Net' (Eqs. (21) and (13)); model definition in Section III-A.2, Eq. (4)]
    "The key issue of unfolding the algorithm of (13) is how to represent the proximal operator proxR(.). Motivated by the denoising interpretation of the proximal operator and the powerful performance of CNNs in image denoising [29], we replace proxR(.) with a shallow CNN."

    In model (4), R is introduced only as 'the deep prior term'; no functional form, convexity, or lower-semicontinuity is stated. The saddle-point reformulation (9) and Moreau decomposition (13) require R to be a proper convex l.s.c. functional. The paper then declares the Mask Denoiser, a non-convex 5-layer CNN, to be the proximal operator of R. Since any trained CNN is not guaranteed to be a proximal map of a convex energy, the regularizer R is effectively defined after training as the function whose proximal operator is the network. Thus the statement that PD-Net is derived from the LMS model and has clear interpretability is a restatement of the design choice 'use this CNN as a denoiser', not a derived correspondence.

  2. renaming known result [Section III-A.3, prototype update (Eq. (8)) with data term (Eq. (5))]
    "Since scaling l does not affect the solution of this problem, the solution for l can be expressed as: l_i^k = β_i ∫ F_q ⊙ u_i^{k−1} dx / ∫ u_i^{k−1} dx, β_i > 0, while ρ is defined by ρ(li, x) = − F_q(x) · li / (∥F_q(x)∥∥li∥)."

    With ρ as the cosine distance (5), the minimizer of ∫ u_i ρ(l,x) dx over l (up to scale) is proportional to ∫ u_i F_q/∥F_q∥ dx, not to ∫ u_i F_q dx. Equation (8) is exactly the unnormalized masked average pooling used by prototypical FSS baselines such as ADNet. The paper never states that ResNet-101 query features are L2-normalized, so (8) does not follow from subproblem (6) with the stated cosine data term. The prototype-update module is therefore not derived from the Mumford-Shah data term; instead, the known MAP operator is relabeled as the closed-form variational solution. This makes the variational derivation of the MAP stage circular: the model's data term is (incorrectly) solved by assuming the desired MAP formula.

full rationale

The empirical performance claims are not circular: Table I reports 5-fold held-out comparisons on public datasets against external baselines ADNet and RPT, and the DSC gains are independently measurable. There is also no load-bearing self-citation chain: references to prior work by co-authors (e.g., [16]) are background, and the optimization tools ([12], [19], [27], [28]) and DnCNN ([29]) are standard external results. The circularity is confined to the central interpretability/derivation narrative. The LMS model (4) leaves the regularizer R unspecified, and the network replaces prox_R by a CNN; after training, R is whatever the CNN implements, so the claim that PD-Net 'solves' the model is true only by definition. Separately, the prototype-update formula (8) is presented as the closed-form minimizer of a cosine data term, but it is actually the standard masked average pooling recipe; the derivation is a renaming of a known FSS component rather than a consequence of the stated variational model. These issues undermine the 'clear interpretability' and 'derived from the LMS model' claims without invalidating the held-out segmentation results, giving a moderate partial circularity score.

Assumptions & free parameters 6 free parameters · 5 assumptions · 2 invented entities

The central claim leans on a stack of hand choices and unproved equivalences: an unspecified regularizer R assumed convex, a CNN treated as its proximal operator, a hand-set entropy coefficient and step sizes, and a fixed two-stage unfolding. The free parameters listed above are what the reader must accept to get from the LMS energy to the reported numbers.

free parameters (6)
  • alpha (entropy regularization coefficient) = not reported
    Controls softmax sharpness in equations (12), (16), (20), and (21); set by hand or tuning and not given, so the derivation's only explicit temperature scale is a free choice.
  • delta_k (primal-dual step size) = not reported
    Appears in the dual update equations (11), (13), and (21); no value or learning rule is reported.
  • K (number of unfolded stages) = 2 for LMS-Net
    Chosen by validation on Synapse-CT (Fig. 6), not derived from the model.
  • beta_i (prototype scaling) = 1 for i=1,2
    Set by default in equation (8); an arbitrary scaling choice.
  • epsilon threshold in MUT = defined as (min+mean)/2
    Equation (19) uses a heuristic gating threshold that determines which support prototypes contribute; it is not learned or derived from the variational model.
  • number of representative prototypes Np = not reported
    Voronoi superpixel initialization in Section III-B.2 requires a value for Np, which is not stated.
assumptions (5)
  • ad hoc to paper The regularizer R is convex, or at least has a well-defined Fenchel conjugate and Moreau decomposition.
    Equations (9) through (14) use R* and prox_R, which require convexity; R is never specified, and the CNN used later is non-convex.
  • ad hoc to paper The proximal operator of R can be replaced by a CNN denoiser trained with the segmentation loss.
    Section III-B.3 states 'we replace proxR(.) with a shallow CNN'; this is an unproved equivalence and is the core of the unfolding claim.
  • domain assumption Support and query features of the same class are close enough that setting the background prototype to the negative of the foreground prototype is valid.
    The initialization in Section III-B.2 sets l0_2 = -l0_1; this is a strong assumption about the geometry of the learned latent space.
  • ad hoc to paper Two unfolded iterations (K=2) are enough for the alternating algorithm to approach the solution of (4).
    Section IV-B sets K=2 and Fig. 6 shows a performance peak, but no convergence analysis is provided.
  • domain assumption A pretrained ResNet-101 on MS-COCO provides a feature space in which cosine pixel-to-prototype comparison is a valid fidelity term.
    Section IV-B uses ResNet-101 pretrained on MS-COCO; the fidelity term in (5) presumes this feature space transfers meaningfully to medical images.
invented entities (2)
  • Deep prior regularizer R(u_i)
    purpose: Regularizer in the LMS energy (4) that is supposed to capture the spatial structure of masks; never explicitly defined, with only its proximal operator replaced by a CNN.
    The paper introduces R as the mathematical foundation but provides no functional form or independent measurement; the only handle is the trained CNN, which is fit to the same segmentation data.
  • Mask Denoiser (MD) as proximal operator
    purpose: CNN that approximates prox_R in the unfolded network; trained with cross-entropy loss rather than a denoising objective, so there is no check that it corresponds to any fixed R.
    Section III-B.3 and equation (21) treat MD as the proximal operator, but the paper provides no evidence that the learned network is the proximity operator of a well-defined regularizer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LMS-Net: A Learned Mumford-Shah Network For Few-Shot Medical Image Segmentation." pith.science (2026). https://pith.science/paper/56QOVKAK

@misc{pith2026250205473,
  author       = {Pith},
  title        = {Pith review of: LMS-Net: A Learned Mumford-Shah Network For Few-Shot Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/56QOVKAK}},
  note         = {Machine review of arXiv:2502.05473}
}
read the original abstract

Few-shot semantic segmentation (FSS) methods have shown great promise in handling data-scarce scenarios, particularly in medical image segmentation tasks. However, most existing FSS architectures lack sufficient interpretability and fail to fully incorporate the underlying physical structures of semantic regions. To address these issues, in this paper, we propose a novel deep unfolding network, called the Learned Mumford-Shah Network (LMS-Net), for the FSS task. Specifically, motivated by the effectiveness of pixel-to-prototype comparison in prototypical FSS methods and the capability of deep priors to model complex spatial structures, we leverage our learned Mumford-Shah model (LMS model) as a mathematical foundation to integrate these insights into a unified framework. By reformulating the LMS model into prototype update and mask update tasks, we propose an alternating optimization algorithm to solve it efficiently. Further, the iterative steps of this algorithm are unfolded into corresponding network modules, resulting in LMS-Net with clear interpretability. Comprehensive experiments on three publicly available medical segmentation datasets verify the effectiveness of our method, demonstrating superior accuracy and robustness in handling complex structures and adapting to challenging segmentation scenarios. These results highlight the potential of LMS-Net to advance FSS in medical imaging applications. Our code will be available at: https://github.com/SDZhang01/LMSNet

Figures

Figures reproduced from arXiv: 2502.05473 by the authors.

Figure 1
Figure 1. Illustration of the pixel-to-prototype comparison mechanism in prototypical FSS methods and how our LMS model integrates pixel-to￾prototype comparison and deep prior techniques into a unified framework. straightforward prototype initialization within the variational framework. Furthermore, this integration paves the way for adopting deep prior techniques within the realm of the FSS task. Originally developed for ima… view at source ↗
Figure 2
Figure 2. The overall structure of the proposed Learned Mumford-Shah Network (LMS-Net) for the FSS task. for solving various image reconstruction problems can also be separated into distinct data subproblems and identical prior subproblems. Significant success has been achieved in solving the denoising subproblem using powerful CNN-based denoisers. We adopt this idea to unfold our optimization algorithm into a network archite… view at source ↗
Figure 3
Figure 3. The structure of the proposed Mask Denoiser and Momentum Update Transformer. further divided into Np regions using a superpixel algorithm based on Voronoi partitioning. p 0 can then be obtained by applying the MAP operator to each of the Np regions individ￾ually. 3) LMS Iteration Module: As shown in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison of our proposed LMS-Net with other medical FSS methods on the Synapse-CT and CHAOS-T2 datasets. GT is the ground truth [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 6
Figure 6. Figure 6: The Dice Score curve on Synapse-CT dataset with a different number of stages K. performance with two cascades, while fLMS-Net performs best with three cascades. 2) Iteration results: To better understand how the spatial prior of semantic regions is captured, we visuali…
Figure 7
Figure 7. Figure 7: Iterates 0,1,2,3 in the fLMS-Net when refining the prediction mask. Left: predicted mask uk 1 after binarization. Middle: predicted noise v k 1 . Right: Entropy of predicted mask E(uk). TABLE II QUANTITATIVE RESULTS OF THE LMS-NET WITH AND WITHOUT PD-NET ON SYNAPSE-CT …
Figure 8
Figure 8. Figure 8: Schematic illustrations of the variants of Mask Denoiser, where (c) is our proposed Mask Denoiser. [9] K. Wang, J. H. Liew, Y. Zou, D. Zhou, and J. Feng, “Panet: Few-shot image semantic segmentation with prototype alignment,” in Proceedings of the IEEE/CVF Internationa…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Medical Image Segmentation based on Deep Active Contour and Mean Curvature Loss Function

    eess.IV 2026-07 conditional novelty 4.5 of 10

    DACMC loss (active-contour region term plus convolution-approximated mean curvature) yields higher Dice than CE/DC/AC/ACE on small liver CT and spleen MRI test sets with U-Net.

Reference graph

Works this paper leans on

43 extracted references · 36 canonical work pages · cited by 1 Pith paper

  1. [1]

    A survey on medical image segmentation,

    S. Masood, M. Sharif, A. Masood, M. Yasmin, and M. Raza, “A survey on medical image segmentation,” Current Medical Imaging , vol. 11, no. 1, pp. 3–14, 2015

    work page 2015

  2. [2]

    A deep learning-based auto-segmentation system for organs-at-risk on whole-body computed tomography images for radiation therapy,

    X. Chen, S. Sun, N. Bai, K. Han, Q. Liu, S. Yao, H. Tang, C. Zhang, Z. Lu, Q. Huang et al., “A deep learning-based auto-segmentation system for organs-at-risk on whole-body computed tomography images for radiation therapy,” Radiotherapy and Oncology , vol. 160, pp. 175–184, 2021

    work page 2021

  3. [3]

    Concurrent multimodality image segmentation by active contours for radiotherapy treatment planning a,

    I. El Naqa, D. Yang, A. Apte, D. Khullar, S. Mutic, J. Zheng, J. D. Bradley, P. Grigsby, and J. O. Deasy, “Concurrent multimodality image segmentation by active contours for radiotherapy treatment planning a,” Medical Physics , vol. 34, no. 12, pp. 4738–4749, 2007

    work page 2007

  4. [4]

    Segnet: A deep con- volutional encoder-decoder architecture for image segmentation,

    V . Badrinarayanan, A. Kendall, and R. Cipolla, “Segnet: A deep con- volutional encoder-decoder architecture for image segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 12, pp. 2481–2495, 2017

    work page 2017

  5. [5]

    Transunet: Transformers make strong encoders for medical image segmentation,

    J. Chen, Y . Lu, Q. Yu, X. Luo, E. Adeli, Y . Wang, L. Lu, A. L. Yuille, and Y . Zhou, “Transunet: Transformers make strong encoders for medical image segmentation,” arXiv preprint arXiv:2102.04306 , 2021

    arXiv 2021

  6. [6]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, proceedings, part III

    work page 2015

  7. [7]

    Active contours without edges,

    T. F. Chan and L. A. Vese, “Active contours without edges,” IEEE Transactions on Image Processing , vol. 10, no. 2, pp. 266–277, 2001

    work page 2001

  8. [8]

    A multiphase level set framework for image segmentation using the mumford and shah model,

    L. A. Vese and T. F. Chan, “A multiphase level set framework for image segmentation using the mumford and shah model,” International Journal of Computer Vision , vol. 50, pp. 271–293, 2002. 10 IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. XX, NO. XX, XXXX 2020 Fig. 8. Schematic illustrations of the variants of Mask Denoiser, where (c) is our proposed Mask Denoiser

    work page 2002

Show all 43 references
  1. [9]

    Panet: Few-shot image semantic segmentation with prototype alignment,

    K. Wang, J. H. Liew, Y . Zou, D. Zhou, and J. Feng, “Panet: Few-shot image semantic segmentation with prototype alignment,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 9197–9206

    2019

  2. [10]

    Flexisp: A flexible camera image processing framework,

    F. Heide, M. Steinberger, Y .-T. Tsai, M. Rouf, D. Paj ˛ ak, D. Reddy, O. Gallo, J. Liu, W. Heidrich, K. Egiazarian et al. , “Flexisp: A flexible camera image processing framework,” ACM Transactions on Graphics (ToG), vol. 33, no. 6, pp. 1–13, 2014

    2014

  3. [11]

    Learned primal-dual reconstruction,

    J. Adler and O. Öktem, “Learned primal-dual reconstruction,” IEEE Transactions on Medical Imaging , vol. 37, no. 6, pp. 1322–1332, 2018

    2018

  4. [12]

    A first-order primal-dual algorithm for convex problems with applications to imaging,

    A. Chambolle and T. Pock, “A first-order primal-dual algorithm for convex problems with applications to imaging,”Journal of Mathematical Imaging and Vision , vol. 40, pp. 120–145, 2011

    2011

  5. [13]

    Optimal approximations by piecewise smooth functions and associated variational problems,

    D. B. Mumford and J. Shah, “Optimal approximations by piecewise smooth functions and associated variational problems,” Communications on Pure and Applied Mathematics , 1989

    1989

  6. [14]

    Ista-net: Interpretable optimization-inspired deep network for image compressive sensing,

    J. Zhang and B. Ghanem, “Ista-net: Interpretable optimization-inspired deep network for image compressive sensing,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2018, pp. 1828–1837

    2018

  7. [15]

    Admm-csnet: A deep learning approach for image compressive sensing,

    Y . Yang, J. Sun, H. Li, and Z. Xu, “Admm-csnet: A deep learning approach for image compressive sensing,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 42, no. 3, pp. 521–538, 2018

    2018

  8. [16]

    Nonlocal regularized cnn for image segmentation,

    F. Jia, X.-C. Tai, and J. Liu, “Nonlocal regularized cnn for image segmentation,” Inverse Problems & Imaging, vol. 14, no. 5, pp. 891–911, 2020

    2020

  9. [17]

    Deep convolutional neural networks with spatial regularization, volume and star-shape priors for image segmentation,

    J. Liu, X. Wang, and X.-C. Tai, “Deep convolutional neural networks with spatial regularization, volume and star-shape priors for image segmentation,” Journal of Mathematical Imaging and Vision , vol. 64, no. 6, pp. 625–645, 2022

    2022

  10. [18]

    Springer, 2015, pp. 234–241

    2015

  11. [19]

    Assembling a learnable mumford–shah type model with multigrid technique for image segmen- tation,

    J. Meng, W. Guo, J. Liu, and M. Yang, “Assembling a learnable mumford–shah type model with multigrid technique for image segmen- tation,” SIAM Journal on Imaging Sciences , vol. 17, no. 2, pp. 1007– 1039, 2024

    2024

  12. [20]

    Global minimization for continuous multiphase partitioning problems using a dual approach,

    E. Bae, J. Yuan, and X.-C. Tai, “Global minimization for continuous multiphase partitioning problems using a dual approach,” International Journal of Computer Vision , vol. 92, no. 1, pp. 112–129, 2011

    2011

  13. [21]

    Denoising prior driven deep neural network for image restoration,

    W. Dong, P. Wang, W. Yin, G. Shi, F. Wu, and X. Lu, “Denoising prior driven deep neural network for image restoration,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 41, no. 10, pp. 2305– 2318, 2018

    2018

  14. [22]

    Deep unfolding network for image super-resolution,

    K. Zhang, L. V . Gool, and R. Timofte, “Deep unfolding network for image super-resolution,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 3217–3226

    2020

  15. [23]

    Cpp-net: Embracing multi-scale feature fusion into deep unfolding cp-ppa network for compressive sensing,

    Z. Guo and H. Gan, “Cpp-net: Embracing multi-scale feature fusion into deep unfolding cp-ppa network for compressive sensing,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 25 086–25 095

    2024

  16. [24]

    A multiphase image segmentation method based on fuzzy region competition,

    F. Li, M. K. Ng, T. Y . Zeng, and C. Shen, “A multiphase image segmentation method based on fuzzy region competition,” SIAM Journal on Imaging Sciences , vol. 3, no. 3, pp. 277–299, 2010

    2010

  17. [25]

    Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration,

    Y . Chen and T. Pock, “Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1256– 1272, 2016

    2016

  18. [26]

    Unfolded proximal neural networks for robust image gaussian denoising,

    H. T. V . Le, A. Repetti, and N. Pustelnik, “Unfolded proximal neural networks for robust image gaussian denoising,” IEEE Transactions on Image Processing, 2024

    2024

  19. [27]

    Some generalized order-disorder transformations,

    R. B. Potts, “Some generalized order-disorder transformations,” in Math- ematical proceedings of the cambridge philosophical society , vol. 48, no. 1. Cambridge University Press, 1952, pp. 106–109

    1952

  20. [28]

    Signal recovery by proximal forward- backward splitting,

    P. L. Combettes and V . R. Wajs, “Signal recovery by proximal forward- backward splitting,” Multiscale Modeling & Simulation , vol. 4, no. 4, pp. 1168–1200, 2005

    2005

  21. [29]

    Proximité et dualité dans un espace hilbertien,

    J.-J. Moreau, “Proximité et dualité dans un espace hilbertien,” Bulletin de la Société mathématique de France , vol. 93, pp. 273–299, 1965

    1965

  22. [30]

    Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising,

    K. Zhang, W. Zuo, Y . Chen, D. Meng, and L. Zhang, “Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising,” IEEE Transactions on Image Processing , vol. 26, no. 7, pp. 3142–3155, 2017

    2017

  23. [31]

    Few-shot semantic segmentation with proto- type learning

    N. Dong and E. P. Xing, “Few-shot semantic segmentation with proto- type learning.”in Proceedings of the British Machine Vision Conference , vol. 3, no. 4, 2018, p. 4

    2018

  24. [32]

    Few-shot medical image segmentation via a region-enhanced prototypical transformer,

    Y . Zhu, S. Wang, T. Xin, and H. Zhang, “Few-shot medical image segmentation via a region-enhanced prototypical transformer,” in In- ternational Conference on Medical Image Computing and Computer- Assisted Intervention . Springer, 2023, pp. 271–280

    2023

  25. [33]

    Anomaly detection-inspired few-shot medical image segmentation through self- supervision with supervoxels,

    S. Hansen, S. Gautam, R. Jenssen, and M. Kampffmeyer, “Anomaly detection-inspired few-shot medical image segmentation through self- supervision with supervoxels,” Medical Image Analysis , vol. 78, p. 102385, 2022

    2022

  26. [34]

    Intermediate prototype mining transformer for few-shot semantic segmentation,

    Y . Liu, N. Liu, X. Yao, and J. Han, “Intermediate prototype mining transformer for few-shot semantic segmentation,” Advances in Neural Information Processing Systems , vol. 35, pp. 38 020–38 031, 2022

    2022

  27. [35]

    Self- supervision with superpixels: Training few-shot medical image seg- mentation without annotation,

    C. Ouyang, C. Biffi, C. Chen, T. Kart, H. Qiu, and D. Rueckert, “Self- supervision with superpixels: Training few-shot medical image seg- mentation without annotation,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part X...

    2020

  28. [36]

    Rethinking few-shot medical segmentation: a vector quantization view,

    S. Huang, T. Xu, N. Shen, F. Mu, and J. Li, “Rethinking few-shot medical segmentation: a vector quantization view,” in Proceedings of the IEEE/CVF Conference on Computer Cision and Pattern Recognition , 2023, pp. 3072–3081

    2023

  29. [37]

    Recurrent mask refinement for few-shot medical image segmentation,

    H. Tang, X. Liu, S. Sun, X. Yan, and X. Xie, “Recurrent mask refinement for few-shot medical image segmentation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 3918–3928

    2021

  30. [38]

    Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge,

    B. Landman, Z. Xu, J. Igelsias, M. Styner, T. Langerak, and A. Klein, “Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge,” in Proc. MICCAI Multi-Atlas Labeling Beyond Cranial V ault—Workshop Challenge, vol. 5, 2015, p. 12

    2015

  31. [39]

    Chaos challenge- combined (ct-mr) healthy abdominal organ segmentation,

    A. E. Kavur, N. S. Gezer, M. Barı¸ s, S. Aslan, P.-H. Conze, V . Groza, D. D. Pham, S. Chatterjee, P. Ernst, S. Özkan et al. , “Chaos challenge- combined (ct-mr) healthy abdominal organ segmentation,” Medical Image Analysis , vol. 69, p. 101950, 2021

    2021

  32. [40]

    Multivariate mixture model for myocardial segmentation combining multi-source images,

    X. Zhuang, “Multivariate mixture model for myocardial segmentation combining multi-source images,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 41, no. 12, pp. 2933–2946, 2018

    2018

  33. [41]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2016, pp. 770–778

    2016

  34. [42]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 . Springer, 2...

    2014

  35. [43]

    Large-scale machine learning with stochastic gradient de- scent,

    L. Bottou, “Large-scale machine learning with stochastic gradient de- scent,” in Proceedings of COMPSTAT’2010: 19th International Con- ference on Computational StatisticsParis France, August 22-27, 2010 Keynote, Invited and Contributed Papers . Springer, 2010, pp. 177– 186

    2010

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.