Pith. sign in

REVIEW 2 major objections 8 minor 38 references

CmFNet: Cross-modal Fusion Network for Weakly-supervised Segmentation of Medical Images

T0 review · 2 major / 8 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A triple-branch network segments tumors from CT and MR using only scribble labels, matching full supervision.

desk verdict A useful cross-modal fusion paper with strong NPC results, but the MR branch is trained on CT-only scribbles with no described registration, so the central claim is hard to verify as written. read the letter →

arxiv 2506.18042 v1 pith:UOPJYG4Z submitted 2025-06-22 cs.CV

Dongdong Meng , Sheng Li , Hao Wu , Suqing Tian , Wenjun Ma , Guoping Wang , Xueqing Yan This is my paper · ORCID
classification cs.CV
keywords weaklysupervisedsegmentationscribbleannotationcross-modalfusionmedicalimagenasopharyngealcarcinomaCT-MRimaging3Dabdominalorgan
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that thin scribble lines drawn on medical images—rather than dense voxel-by-voxel labels—are enough to train accurate 3D segmentation networks when two imaging modalities are available. It proposes CmFNet, a triple-branch network with one branch for CT, one for MR, and a shared cross-modal branch that fuses and enhances features from both. A hybrid loss combines partial cross-entropy on scribbled pixels, a multi-view conditional random field regularizer, and mean-squared-error consistency between branch predictions to stop the model overfitting to the few labelled pixels. On a clinical nasopharyngeal carcinoma dataset the scribble-trained model reports 76.75 Dice, above the best weakly supervised baseline and above a fully supervised V-Net; the authors claim it also beats fully supervised methods when given full masks. A sympathetic reader would take the paper's claim to be that cross-modal fusion is the ingredient that lets weak supervision approach full supervision.

What carries the argument

The load-bearing mechanism is the triple-branch architecture plus a three-term loss. Two V-Net-style branches process CT and MR separately with different down-sampling depths (three stages for CT, four for MR), preserving detail in one and context in the other. Their feature maps are combined by the Cross-Modal Feature Fusion (CFF) module, a concatenation-plus-convolution, and the Cross-Modal Feature Enhancement (CFE) module, which aligns the third-layer features and increases receptive field before the shared decoder. Supervision is $L_{\mathrm{total}} = L_{\mathrm{ssl}} + L_{\mathrm{imr}} + L_{\mathrm{imc}}$: partial cross-entropy restricted to scribbled pixels, a multi-view conditional random field loss applied in axial, sagittal, and coronal planes, and mean-squared-error consistency pulling the cross-modal prediction toward each modality branch. The CFF module is deliberately simple; the paper argues the architecture itself, not complex attention, makes fusion effective.

What would settle it

Overlay the CT-drawn scribble masks on the corresponding MR volumes and measure their overlap with the tumor boundary visible in MR; if the masks are systematically offset, the MR branch receives corrupted labels and the reported DSC would drop when MR-native scribbles are used instead. A simpler computation: retrain the MR-specific branch with scribbles drawn directly on MR and compare to the current CT-transferred scribbles.

Watch

Extended reading notes

Core claim

The paper's central claim is that cross-modal fusion, not a more elaborate attention mechanism, carries most of the benefit in scribble-supervised segmentation. On its clinical NPC dataset, CmFNet with scribble annotations reaches 76.75±5.78 DSC and 2.69±0.89 mm ASD, exceeding the best weakly supervised baseline DMSPS (71.58±7.73 DSC, 4.70±1.81 mm ASD) and the fully supervised V-Net (73.98±7.28 DSC). With full masks the same network reports 78.95±5.99 DSC, ahead of all compared fully supervised methods including DA-VNet. On the single-modal WORD dataset, where the same CT volume is fed into both branches, the method reaches a mean DSC of 84.19, second to the task-specific DMSPS baseline but with stronger surface-distance results on several organs. The authors interpret these results as evidence that modality-specific branches preserve each image type's details while the shared branch aligns them, and that the three-term loss prevents the overfitting that sparse scribble labels usually cause.

Load-bearing premise

All three branches are trained with the same scribble mask, but the paper says that mask was drawn on CT images only; for the MR branch to receive correct supervision, the CT and MR volumes must be aligned pixel-for-pixel, and the paper does not describe any registration or alignment check.

Editorial extensions

If this is right

  • Scribble-only training on paired CT and MR can match or beat a fully supervised V-Net on NPC tumor segmentation, so annotation cost could drop substantially in radiotherapy workflows.
  • The same network works on single-modal CT data by feeding the same volume into both branches, giving competitive abdominal organ segmentation on WORD (mean DSC 84.19), though below the task-specific DMSPS baseline.
  • Ablations show each added component contributes: CFE raises DSC from 73.76 to 75.96, adding CFF reaches 76.75, and the hybrid loss raises scribble-only training from 68.20 to 76.75, so the gains are attributed to the fusion and regularization together.
  • When full masks are available, the same architecture reports 78.95 DSC, ahead of the compared fully supervised methods, suggesting the cross-modal design does not depend on weak supervision to help.

Reading between the lines

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

  • If the CT/MR alignment assumption holds, the same triple-branch recipe should transfer to other paired modalities such as PET/CT or T1/T2 MRI; the paper only claims future work, this is an extension.
  • The WORD experiment suggests the cross-modal branch behaves as a feature-sharing regularizer rather than a true modality encoder, since identical CT content is fed into both branches; this interpretation goes beyond the authors' framing.
  • A testable extension the paper does not report: compare CT-transferred scribbles against MR-native scribbles for the MR branch; a large gap would indicate alignment noise rather than cross-modal learning drives part of the gain.
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

2 major / 8 minor

Summary. The manuscript proposes CmFNet, a 3D weakly supervised segmentation network for cross-modal medical images. The architecture consists of two modality-specific branches (CT and MR) and a shared cross-modal branch, combined with CFF and CFE modules. Training uses a hybrid loss: partial cross-entropy on scribbles, a multi-view CRF loss for intra-modal regularization, and an MSE inter-modal consistency loss. Experiments on a clinical NPC dataset (CT+MR) and the WORD CT dataset compare with weakly and fully supervised baselines. The paper reports that CmFNet achieves 76.75±5.78 DSC and 2.69±0.89 mm ASD on NPC with scribbles, outperforming the listed weakly supervised baselines, and 84.19 mean DSC on WORD, ranking second behind DMSPS; with full annotations on NPC, it reports 78.95 DSC, above the fully supervised baselines listed.

Significance. If the results are reproducible, CmFNet would be a useful contribution to scribble-based cross-modal segmentation. The paper has practical strengths: it reports standard deviations, runs ablations of the fusion modules and loss terms, and includes a hyperparameter sensitivity analysis. The architectural motivation of separating modality-specific and shared features is reasonable, and the loss decomposition is clear. However, the current manuscript leaves a critical spatial-correspondence issue unresolved: scribbles are drawn only on CT, yet the same mask supervises the MR branch, and the geometry between CT and MR is not described. This, together with an imported fully-supervised comparison from earlier work, currently limits confidence in the headline numbers.

major comments (2)
  1. [Section IV-A3 / Section III-D1] Scribble supervision for the MR branch requires spatial correspondence that is not established. Section IV-A3 states that scribbles were manually drawn on CT images only, and the CT and MR volumes have different matrices and pixel spacings (512x512, 1.27 mm vs 384x384, 1.30 mm). Eq. (4) nevertheless applies the same mask s to y_mr, and Eq. (8) uses y_mr as a consistency target for the cross-modal output. If the MR volumes are not registered or resampled to the CT grid, the MR branch is trained on misaligned labels and the IMC loss propagates that noise. The Resampler shown in Fig. 2 is never described in the method or implementation sections. The authors must specify the registration/resampling protocol, describe how the scribble mask is transformed, and provide an alignment check or quantitative evidence that the CT-drawn scribbles are valid for the MR branch. Without this, the reported NPC gains in Table I are not reproducible from the paper as written.
  2. [Section IV-C3 / Table III] The fully-supervised comparison is not self-contained. The manuscript says, 'we referred to the experimental results from [25] on the NPC dataset,' but it does not establish that the data splits, preprocessing, training settings, and evaluation metric computations are identical to those used for CmFNet. Since Table III is the basis for the claim that CmFNet outperforms fully supervised methods with full annotations, the authors need to retrain all baselines under the same protocol or provide a detailed equivalence argument. Without this, the 78.95±5.99 DSC result is not directly comparable to the DA-VNet, V-Net, UNet++, and other listed numbers.
minor comments (8)
  1. [Section III-D2] Eq. (7) lists lambda_ct twice instead of lambda_ct and lambda_mr; since the sentence explicitly refers to both weights, this is a typo, but the equation should be corrected.
  2. [Abstract / Section IV-C2] The abstract states that the approach outperforms state-of-the-art weakly supervised methods without qualification, but on WORD Table II shows CmFNet ranks second overall behind DMSPS in mean DSC (84.19 vs 86.72). Please revise the abstract and conclusion to say that it is competitive with, or second-best to, DMSPS on WORD.
  3. [Section IV-C3] The text says Table III presents a qualitative comparison, but Table III reports quantitative DSC/ASD values; use quantitative instead.
  4. [Tables I-VII] Several table cells contain duplicated numbers in the text (for example, 76.75±5.7876.75±5.78 in Table I and 1.44±2.911.44±2.91 in Table II); clean these formatting artifacts before resubmission.
  5. [Section IV-B] The implementation section names hyperparameters lambda_1, lambda_2, alpha_1, alpha_2, while Eq. (7) and Fig. 6 use lambda_ct and lambda_mr; align the notation.
  6. [Section IV-C1 / Table I] The row labeled FullySup is described as an upper bound, yet Ours with scribbles (76.75 DSC) exceeds it (73.98 DSC). Clarify what architecture and supervision this row uses and avoid the term upper bound unless the model is identical to the one used for the weakly supervised comparisons.
  7. [Section III-C / Eq. (1)] Eq. (1) has an unclosed parenthesis after f_i_mr, and the text defines W as weight when width is intended; correct these presentation issues.
  8. [Section IV-C2] On WORD, CT images are fed into both modality-specific branches, so the method is not cross-modal in that setting; state this explicitly and explain how the conclusions transfer to the cross-modal NPC scenario.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CmFNet is an empirical architecture-and-loss paper whose reported gains rest on external baselines and ablation experiments, not on a fitted constant or a self-referential derivation.

full rationale

No step in the paper derives a prediction from a quantity that already contains that prediction. The scalar losses in Eqs. (3)-(9) are standard weak-supervision terms: partial cross-entropy over scribbled pixels, multi-view CRF regularization, and MSE consistency among branches; none of these terms is fitted to the evaluation metric or to the test set. The hyper-parameters λ_ct, λ_mr, α_1, and α_2 are tuned via the sensitivity analysis in Fig. 6 rather than solved for from the reported DSC/ASD values, so the numbers in Tables I-III are not reproductions of a fitted curve. The only self-citation is to the authors' earlier DA-VNet [25] as the source of fully-supervised baseline numbers in Table III; that citation is a benchmark-comparison choice and does not argue for CmFNet's architecture, so it is not load-bearing. The CT/MR scribble-alignment concern and the duplicated λ_ct in Eq. (7) are correctness or reproducibility risks, not circularity. The weakly supervised comparisons on NPC and WORD are against external methods (DMSPS, TV, EM, MS, USTM, S2L, and pCE) and are described as re-run from public source code, and the ablation study isolates the contribution of each proposed module. The derivation chain is therefore self-contained: the claimed improvement is an empirical result, not a consequence of the method's own definitions.

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

The central empirical claim rests on standard deep-learning assumptions (valid labels, aligned multimodal data, comparable baselines) plus four tuned loss weights. No new physical entities or theoretical constructs are introduced.

free parameters (4)
  • lambda_ct (weight of CT-image CRF regularization, L_imr) = 0.2
    Tuned on the NPC validation split; sensitivity shown in Fig. 6; chosen for balanced DSC/ASD.
  • lambda_mr (weight of MR-image CRF regularization, L_imr) = 0.2
    Tuned like lambda_ct; Eq. (7) mistakenly prints lambda_ct twice, but text and implementation state lambda_ct and lambda_mr.
  • alpha1 (weight of MSE consistency between cross-modal and CT branch) = 0.8
    Tuned in (0.0, 1.0); sensitivity shown in Fig. 6; optimal at 0.8.
  • alpha2 (weight of MSE consistency between cross-modal and MR branch) = 0.8
    Same as alpha1.
assumptions (5)
  • ad hoc to paper CT and MR volumes share the same voxel grid so that one scribble mask supervises both branches.
    Scribbles drawn only on CT (Sec. IV-A3) are used in Eq. (4) for y_ct, y_mr and y_mm; no registration or separate MR scribbles are described.
  • domain assumption Scribble annotations on the annotated pixels are correct.
    Partial cross-entropy loss assumes label accuracy on the scribble pixels.
  • domain assumption The public WORD dataset's scribbles are valid and follow the same protocol.
    Used as ground truth for weak supervision; no verification described.
  • ad hoc to paper Fully-supervised baseline numbers from [25] (DA-VNet, etc.) are comparable because identical NPC data split and evaluation protocol were used.
    Section IV-C3 states results were 'referred to' from [25] rather than re-run; protocol matching is assumed.
  • domain assumption The V-Net backbone with chosen downsampling depths is a sufficient feature extractor for both modalities.
    Architecture choice not derived; standard assumption in empirical segmentation papers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CmFNet: Cross-modal Fusion Network for Weakly-supervised Segmentation of Medical Images." pith.science (2026). https://pith.science/paper/UOPJYG4Z

@misc{pith2026250618042,
  author       = {Pith},
  title        = {Pith review of: CmFNet: Cross-modal Fusion Network for Weakly-supervised Segmentation of Medical Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UOPJYG4Z}},
  note         = {Machine review of arXiv:2506.18042}
}
read the original abstract

Accurate automatic medical image segmentation relies on high-quality, dense annotations, which are costly and time-consuming. Weakly supervised learning provides a more efficient alternative by leveraging sparse and coarse annotations instead of dense, precise ones. However, segmentation performance degradation and overfitting caused by sparse annotations remain key challenges. To address these issues, we propose CmFNet, a novel 3D weakly supervised cross-modal medical image segmentation approach. CmFNet consists of three main components: a modality-specific feature learning network, a cross-modal feature learning network, and a hybrid-supervised learning strategy. Specifically, the modality-specific feature learning network and the cross-modal feature learning network effectively integrate complementary information from multi-modal images, enhancing shared features across modalities to improve segmentation performance. Additionally, the hybrid-supervised learning strategy guides segmentation through scribble supervision, intra-modal regularization, and inter-modal consistency, modeling spatial and contextual relationships while promoting feature alignment. Our approach effectively mitigates overfitting, delivering robust segmentation results. It excels in segmenting both challenging small tumor regions and common anatomical structures. Extensive experiments on a clinical cross-modal nasopharyngeal carcinoma (NPC) dataset (including CT and MR imaging) and the publicly available CT Whole Abdominal Organ dataset (WORD) show that our approach outperforms state-of-the-art weakly supervised methods. In addition, our approach also outperforms fully supervised methods when full annotation is used. Our approach can facilitate clinical therapy and benefit various specialists, including physicists, radiologists, pathologists, and oncologists.

Figures

Figures reproduced from arXiv: 2506.18042 by the authors.

Figure 1
Figure 1. Examples of scribble annotations on NPC and WORD datasets. The abbreviations Liv, Spl, Lkid, RKid, Sto, Gal, and Pan are used to denote the liver, spleen, left kidney, right kidney, stomach, gallbladder, and pancreas, respectively. F ULLY-SUPERVISED deep learning methods are gener￾ally limited by the time-consuming and costly process of manual annotations [1], [2]. To address this, many techniques have been develope… view at source ↗
Figure 2
Figure 2. Overview of our approach. (a) Our network consists of triple branches, i.e., two modality-specific feature learning branches from individual modalities, along with a shared cross-modal feature learning branch. The prediction results of the three branches, yct, ymr and ymm are guided by a hybrid-supervised learning strategy that comprises scribble-supervised loss Lssl, intra-modal regularization Limr, and inter-modal… view at source ↗
Figure 3
Figure 3. Qualitative comparison between our method and SOTA weakly-supervised methods on the NPC dataset (corresponding to Table I). Our approach achieves the best performance and is very close to the ground truth. TABLE II QUANTITATIVE COMPARISON BETWEEN OUR METHOD AND SOTA WEAKLY-SUPERVISED METHODS ON WORD DATASET. OUR APPROACH GENERALLY RANKS THE SECOND BEST, ONLY BEHIND THE TASK-SPECIFIC DMSPS [5]. Metric Methods Data Li… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison between our method and SOTA weakly-supervised methods on the WORD dataset (corresponding to Table II). dispersed tumors, outperforming other approaches that fail to effectively capture multi-modal features, resulting in detail loss and imprecise …
Figure 5
Figure 5. Figure 5: Qualitative comparison between our method and SOTA fully-supervised methods on NPC dataset (corresponding to Table III). TABLE III QUANTITATIVE COMPARISON BETWEEN OUR METHOD AND SOTA FULLY SUPERVISED METHODS ON NPC DATASET. Methods Data DSC (%) ↑ ASD (mm) ↓ M-CNN [37] …
Figure 6
Figure 6. Figure 6: Sensitivity analysis of hyper-parameter λct, λmr, α1 and α2 on the NPC dataset, respectively. challenges of accuracy degradation and overfitting in weakly supervised learning, particularly in cases involving small tu￾mor regions and complex anatomical structures. The m…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 31 canonical work pages

  1. [25]

    3d reconstruction-oriented fully automatic multi-modal tumor segmentation by dual attention-guided vnet,

    D. Meng, S. Li, B. Sheng, H. Wu, S. Tian, W. Ma, G. Wang, and X. Yan, “3d reconstruction-oriented fully automatic multi-modal tumor segmentation by dual attention-guided vnet,”The Visual Computer, vol. 39, no. 8, pp. 3183–3196, 2023

    work page 2023

  2. [1]

    Scribble-supervised medical image segmentation via dual-branch net- work and dynamically mixed pseudo labels supervision,

    X. Luo, M. Hu, W. Liao, S. Zhai, T. Song, G. Wang, and S. Zhang, “Scribble-supervised medical image segmentation via dual-branch net- work and dynamically mixed pseudo labels supervision,” inInterna- tional Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 528–538, Springer, 2022

    work page 2022

  3. [2]

    Image segmentation using deep learning: A survey,

    S. Minaee, Y . Boykov, F. Porikli, A. Plaza, N. Kehtarnavaz, and D. Terzopoulos, “Image segmentation using deep learning: A survey,” IEEE transactions on pattern analysis and machine intelligence, vol. 44, no. 7, pp. 3523–3542, 2021

    2021

  4. [3]

    Learning deep features for discriminative localization,

    B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, “Learning deep features for discriminative localization,” inProceedings of the IEEE conference on computer vision and pattern recognition, pp. 2921–2929, 2016

    2016

  5. [4]

    Weakly supervised segmentation with cross- modality equivariant constraints,

    G. Patel and J. Dolz, “Weakly supervised segmentation with cross- modality equivariant constraints,”Medical image analysis, vol. 77, p. 102374, 2022

    work page 2022

  6. [5]

    Dmsps: Dynamically mixed soft pseudo-label supervision for scribble-supervised medical image segmentation,

    M. Han, X. Luo, X. Xie, W. Liao, S. Zhang, T. Song, G. Wang, and S. Zhang, “Dmsps: Dynamically mixed soft pseudo-label supervision for scribble-supervised medical image segmentation,”Medical Image Analysis, vol. 97, p. 103274, 2024

    work page 2024

  7. [6]

    Scribble2label: Scribble-supervised cell segmentation via self-generating pseudo-labels with consistency,

    H. Lee and W.-K. Jeong, “Scribble2label: Scribble-supervised cell segmentation via self-generating pseudo-labels with consistency,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part I 23, pp. 14–23, Springer, 2020

    work page 2020

  8. [7]

    Deepcut: Object segmentation from bounding box annotations using convolutional neural networks,

    M. Rajchl, M. C. Lee, O. Oktay, K. Kamnitsas, J. Passerat-Palmbach, W. Bai, M. Damodaram, M. A. Rutherford, J. V . Hajnal, B. Kainz,et al., “Deepcut: Object segmentation from bounding box annotations using convolutional neural networks,”IEEE transactions on medical imaging, vol. 36, no. 2, pp. 674–683, 2016

    work page 2016

Show all 38 references
  1. [8]

    Pa-seg: learning from point annotations for 3d medical image segmentation using contextual regularization and cross knowledge distillation,

    S. Zhai, G. Wang, X. Luo, Q. Yue, K. Li, and S. Zhang, “Pa-seg: learning from point annotations for 3d medical image segmentation using contextual regularization and cross knowledge distillation,”IEEE Transactions on Medical Imaging, 2023

    2023

  2. [9]

    Scribformer: Transformer makes cnn work better for scribble-based medical image segmentation,

    Z. Li, Y . Zheng, D. Shan, S. Yang, Q. Li, B. Wang, Y . Zhang, Q. Hong, and D. Shen, “Scribformer: Transformer makes cnn work better for scribble-based medical image segmentation,”IEEE Transactions on Medical Imaging, 2024

    2024

  3. [10]

    Scribblesup: Scribble- supervised convolutional networks for semantic segmentation,

    D. Lin, J. Dai, J. Jia, K. He, and J. Sun, “Scribblesup: Scribble- supervised convolutional networks for semantic segmentation,” inPro- ceedings of the IEEE conference on computer vision and pattern recognition, pp. 3159–3167, 2016

    2016

  4. [11]

    Semi-supervised learning by entropy minimization,

    Y . Grandvalet and Y . Bengio, “Semi-supervised learning by entropy minimization,”Advances in neural information processing systems, vol. 17, 2004

    2004

  5. [12]

    Unsupervised total variation loss for semi-supervised deep learning of semantic segmenta- tion,

    M. Javanmardi, M. Sajjadi, T. Liu, and T. Tasdizen, “Unsupervised total variation loss for semi-supervised deep learning of semantic segmenta- tion,”arXiv preprint arXiv:1605.01368, 2016

    2016 arXiv

  6. [13]

    Weakly supervised segmentation of covid19 infection with scribble annotation on ct images,

    X. Liu, Q. Yuan, Y . Gao, K. He, S. Wang, X. Tang, J. Tang, and D. Shen, “Weakly supervised segmentation of covid19 infection with scribble annotation on ct images,”Pattern recognition, vol. 122, p. 108341, 2022

    2022

  7. [14]

    A review: Deep learning for medical im- age segmentation using multi-modality fusion,

    T. Zhou, S. Ruan, and S. Canu, “A review: Deep learning for medical im- age segmentation using multi-modality fusion,”Array, vol. 3, p. 100004, 2019

    2019

  8. [15]

    Comparing different ct, pet and mri multi-modality image combinations for deep learning- based head and neck tumor segmentation,

    J. Ren, J. Eriksen, J. Nijkamp, and S. Korreman, “Comparing different ct, pet and mri multi-modality image combinations for deep learning- based head and neck tumor segmentation,”Acta Oncologica, vol. 60, pp. 1–8, 07 2021

    2021

  9. [16]

    Hyperdense-net: a hyper-densely connected cnn for multi-modal image segmentation,

    J. Dolz, K. Gopinath, J. Yuan, H. Lombaert, C. Desrosiers, and I. B. Ayed, “Hyperdense-net: a hyper-densely connected cnn for multi-modal image segmentation,”IEEE transactions on medical imaging, vol. 38, no. 5, pp. 1116–1126, 2018

    2018

  10. [17]

    Flexible fusion network for multi-modal brain tumor segmentation,

    H. Yang, T. Zhou, Y . Zhou, Y . Zhang, and H. Fu, “Flexible fusion network for multi-modal brain tumor segmentation,”IEEE Journal of Biomedical and Health Informatics, vol. 27, no. 7, pp. 3349–3359, 2023

    2023

  11. [18]

    Modality-aware mutual learning for multi-modal medical image seg- mentation,

    Y . Zhang, J. Yang, J. Tian, Z. Shi, C. Zhong, Y . Zhang, and Z. He, “Modality-aware mutual learning for multi-modal medical image seg- mentation,” inMMICCAI 2021: 24th International Conference, Stras- bourg, France, September 27–October 1, 2021, Proceedings, Part I 24, pp. 58...

    2021

  12. [19]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical Image Computing and Computer-Assisted Intervention – MICCAI 2015, pp. 234–241, Springer, 2015

    2015

  13. [20]

    3d u-net: learning dense volumetric segmentation from sparse annota- tion,

    ¨O. C ¸ ic ¸ek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger, “3d u-net: learning dense volumetric segmentation from sparse annota- tion,” inMedical Image Computing and Computer-Assisted Intervention – MICCAI 2016, pp. 424–432, Springer, 2016

    2016

  14. [21]

    Medical image segmentation review: The success of u-net,

    R. Azad, E. K. Aghdam, A. Rauland, Y . Jia, A. H. Avval, A. Bo- zorgpour, S. Karimijafarbigloo, J. P. Cohen, E. Adeli, and D. Merhof, “Medical image segmentation review: The success of u-net,”ArXiv, vol. abs/2211.14830, 2022

    2022 arXiv

  15. [22]

    Unet++: Re- designing skip connections to exploit multiscale features in image segmentation,

    Z. Zhou, M. Siddiquee, N. Tajbakhsh, and J. Liang, “Unet++: Re- designing skip connections to exploit multiscale features in image segmentation,”IEEE Transactions on Medical Imaging, vol. 39, no. 6, pp. 1856–1867, 2020

    2020

  16. [23]

    Densely connected convolutional networks,

    G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolutional networks,” in2017 IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), pp. 2261–2269, 2017

    2017

  17. [24]

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

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

    2016

  18. [26]

    Unetr: Transformers for 3d medical image segmentation,

    A. Hatamizadeh, Y . Tang, V . Nath, D. Yang, A. Myronenko, B. Land- man, H. R. Roth, and D. Xu, “Unetr: Transformers for 3d medical image segmentation,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, pp. 574–584, 2022

    2022

  19. [27]

    Mamba: Linear-time sequence modeling with selective state spaces,

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

    2023 arXiv

  20. [28]

    Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality,

    T. Dao and A. Gu, “Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality,” inInterna- tional Conference on Machine Learning (ICML), 2024

    2024

  21. [29]

    Vision mamba: Efficient visual representation learning with bidirectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vision mamba: Efficient visual representation learning with bidirectional state space model,”arXiv preprint arXiv:2401.09417, 2024

    2024 arXiv

  22. [30]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,

    K. He, X. Zhang, S. Ren, and J. Sun, “Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,”2015 IEEE International Conference on Computer Vision (ICCV), pp. 1026– 1034, 2015

    2015

  23. [31]

    Word: A large scale dataset, bench- mark and clinical applicable study for abdominal organ segmentation from ct image,

    X. Luo, W. Liao, J. Xiao, J. Chen, T. Song, X. Zhang, K. Li, D. N. Metaxas, G. Wang, and S. Zhang, “Word: A large scale dataset, bench- mark and clinical applicable study for abdominal organ segmentation from ct image,”Medical Image Analysis, vol. 82, p. 102642, 2022

    2022

  24. [32]

    Scribble- based 3d multiple abdominal organ segmentation via triple-branch multi- dilated network with pixel-and class-wise consistency,

    M. Han, X. Luo, W. Liao, S. Zhang, S. Zhang, and G. Wang, “Scribble- based 3d multiple abdominal organ segmentation via triple-branch multi- dilated network with pixel-and class-wise consistency,” inInternational Conference on Medical Image Computing and Computer-Assisted Inte...

    2023

  25. [33]

    User-guided 3d active contour segmentation of anatomical structures: significantly improved efficiency and reliability,

    P. A. Yushkevich, J. Piven, H. C. Hazlett, R. G. Smith, S. Ho, J. C. Gee, and G. Gerig, “User-guided 3d active contour segmentation of anatomical structures: significantly improved efficiency and reliability,” Neuroimage, vol. 31, no. 3, pp. 1116–1128, 2006

    2006

  26. [34]

    Learning to segment from scribbles using multi-scale adversarial attention gates,

    G. Valvano, A. Leo, and S. A. Tsaftaris, “Learning to segment from scribbles using multi-scale adversarial attention gates,”IEEE Transac- tions on Medical Imaging, vol. 40, no. 8, pp. 1990–2001, 2021

    1990

  27. [35]

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

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga,et al., “Pytorch: An imperative style, high-performance deep learning library,”Advances in neural information processing systems, vol. 32, 2019

    2019

  28. [36]

    Mumford–shah loss functional for image seg- mentation with deep learning,

    B. Kim and J. C. Ye, “Mumford–shah loss functional for image seg- mentation with deep learning,”IEEE Transactions on Image Processing, vol. PP, pp. 1–1, 09 2019

    2019

  29. [37]

    Na- sopharyngeal carcinoma segmentation based on enhanced convolutional neural networks using multi-modal metric learning,

    Z. Ma, S. Zhou, X. Wu, H. Zhang, W. Yan, S. Sun, and J. Zhou, “Na- sopharyngeal carcinoma segmentation based on enhanced convolutional neural networks using multi-modal metric learning,”Physics in Medicine & Biology, vol. 64, no. 2, p. 025005, 2019

    2019

  30. [38]

    Road extraction by deep residual u- net,

    Z. Zhang, Q. Liu, and Y . Wang, “Road extraction by deep residual u- net,”IEEE Geoscience and Remote Sensing Letters, vol. PP, no. 99, pp. 1–5, 2017

    2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.