Pith. sign in

REVIEW 4 major objections 7 minor 32 references

Medical Image Segmentation Using Advanced Unet: VMSE-Unet and VM-Unet CBAM+

T0 review · 4 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper argues that adding Squeeze-and-Excitation and CBAM attention modules to the VM-UNet architecture improves medical segmentation accuracy while making inference faster and lighter on GPU and CPU.

desk verdict Routine SE/CBAM extension of VM-UNet whose own Table II refutes the abstract's central claim; not ready for serious review. read the letter →

arxiv 2507.00511 v2 pith:6VEYEFL7 submitted 2025-07-01 eess.IV cs.CVcs.LG

Sayandeep Kanrar , Raja Piyush , Qaiser Razi , Debanshi Chakraborty , Vikas Hassija , GSS Chalapathi This is my paper
classification eess.IVcs.CVcs.LG
keywords medicalimagesegmentationVisionMambaU-NetSqueeze-and-ExcitationConvolutionalBlockAttentionModuleVMSE-UnetVM-UnetCBAM+efficientinferenceMICCAI2009
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 establish that two lightweight attention mechanisms can be added to the Vision Mamba U-Net (VM-UNet) to make medical image segmentation both more accurate and cheaper to run. The authors insert a Squeeze-and-Excitation block, which rescales feature channels, and a Convolutional Block Attention Module, which applies channel and spatial attention, producing two variants: VMSE-Unet and VM-Unet CBAM+. Across cardiac MRI, polyp, and breast-ultrasound datasets, they report higher accuracy, IoU, precision, and recall for the enhanced models in most comparisons, and they report VMSE-Unet as the fastest and most memory-efficient on GPU and CPU. If those results hold, the work offers a straightforward recipe for clinical-grade segmentation on ordinary hardware.

What carries the argument

The load-bearing mechanism is the pair of attention modules inserted into the VM-UNet encoder-decoder path. VM-UNet is a U-Net-style encoder-decoder that uses Visual State Space blocks to model long-range image dependencies. The SE block squeezes each channel to a scalar via global average pooling, passes it through two fully connected layers with ReLU and sigmoid, and rescales channels by element-wise multiplication. The CBAM block computes channel attention from average- and max-pooled spatial summaries, then spatial attention from channel-pooled maps through a convolution, and applies both as multiplicative gates. The paper places these modules after the VSS blocks at multiple resolutions, which it claims lets the model emphasize clinically relevant features without the computational cost of transformers.

What would settle it

Run the three models on identical fixed splits with the same optimizer, learning rate, batch size, epochs, and several random seeds, then check whether VMSE-Unet still shows the lowest loss and fastest inference while matching or beating the other models on IoU and accuracy. A cheaper observation: Table II already records VM-Unet CBAM+ ahead of VMSE-Unet on MICCAI 2009 IoU (0.5679 vs 0.5660) and on Kvasir-SEG accuracy (0.9795 vs 0.9678), so the abstract's blanket claim is not directly supported by the paper's own table.

Watch

Extended reading notes

Core claim

The central claim is that VM-UNet, a U-Net variant that uses Visual State Space blocks to capture long-range dependencies, can be improved by inserting two attention modules at each encoder scale: the SE block, which recalibrates channel responses via global pooling and two fully connected layers, and the CBAM block, which applies channel attention followed by spatial attention. The resulting models, VMSE-Unet and VM-Unet CBAM+, are evaluated on MICCAI 2009 cardiac MRI, Kvasir-SEG polyps, and a breast-ultrasound dataset. The authors report that both enhanced models beat the baseline VM-UNet on accuracy, IoU, precision, and recall in most comparisons while reducing loss, and that VMSE-Unet has the fastest GPU and CPU inference times (0.04212 s and 1.11716 s) and the lowest memory use (2.01 GB GPU, 2.13 GB CPU). The paper concludes that VMSE-Unet is the best overall architecture and proposes it for real-time, resource-constrained clinical settings.

Load-bearing premise

The comparison assumes the three models were trained and evaluated under identical, representative conditions, so the metric differences reflect the attention modules rather than different data splits, training settings, or lucky single runs.

Editorial extensions

If this is right

  • If reproducible, VMSE-Unet's 0.042 s GPU inference and roughly 2 GB memory footprint make it suitable for near-real-time clinical use on ordinary hardware.
  • Because SE and CBAM are inserted as modular blocks around the existing VSS layers, the same recipe can be applied to other Mamba-based U-Net variants without redesign.
  • The large reported gains on Kvasir-SEG and BUS suggest these attention mechanisms help most where object boundaries are irregular, such as polyps and ultrasound lesions.
  • With comparable or better accuracy and lower computational cost, VMSE-Unet would be a strong baseline for future medical segmentation experiments.

Reading between the lines

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

  • The paper does not report training epochs, optimizer, learning rate, batch size, or random seeds, and Tables II and III show single-run values without variance, so the claimed advantages could be run-to-run noise rather than architecture effects.
  • Table II shows VM-Unet CBAM+ beating VMSE-Unet on MICCAI 2009 IoU and on Kvasir-SEG accuracy, and with lower loss on Kvasir-SEG, so the abstract's blanket 'highest accuracy, IoU, precision, recall' is not consistent with every reported row.
  • The efficiency comparison omits input resolution, batch size, and hardware details, so the inference-time and memory numbers are hard to compare with other published models; adding FLOPs, parameter counts, and multi-run error bars would make the claim testable.
  • The same SE-plus-CBAM insertion could be tested on other state-space U-Net variants and on 3D medical volumes, which would show whether the gains are specific to these three datasets or general to the architecture.
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 / 7 minor

Summary. The paper proposes two variants of the Vision Mamba U-Net (VM-UNet) architecture: VMSE-Unet, which adds Squeeze-and-Excitation (SE) blocks, and VM-Unet CBAM+, which adds Convolutional Block Attention Modules (CBAM). The authors evaluate these models against a baseline VM-UNet on three datasets (MICCAI 2009, Kvasir-SEG, and a synthetic BUS dataset) and report accuracy, IoU, precision, recall, loss, inference time, and memory usage. The central claim, stated in the abstract and conclusion, is that VMSE-Unet achieves the highest accuracy, IoU, precision, and recall while maintaining low loss and exhibiting the best computational efficiency.

Significance. If the empirical claims were reliable, the contribution would be a practical accuracy-efficiency trade-off for medical image segmentation, which is a relevant goal for clinical deployment. The architecture modifications (adding SE and CBAM to VM-UNet) are plausible but incremental, and no comparison to existing state-of-the-art segmentation models is provided. More importantly, the paper's own results table contradicts the central claim, and the experimental protocol lacks the details and statistical rigor needed to support any performance comparison. The reported findings therefore do not currently constitute a reliable contribution to the field.

major comments (4)
  1. [Abstract and Section VI, Table II] The abstract and conclusion claim that VMSE-Unet 'achieves the highest accuracy, IoU, precision, and recall while maintaining low loss values.' Table II directly contradicts this. On MICCAI 2009, VM-Unet CBAM+ has a higher IoU (0.5679) than VMSE-Unet (0.5660), and the baseline VM-Unet has a far lower loss (0.0591) than VMSE-Unet (0.3616). On Kvasir-SEG, VM-Unet CBAM+ has a higher accuracy (0.9795) than VMSE-Unet (0.9678). These entries falsify the universal superlatives in the abstract and conclusion, so the central claim is not supported by the paper's own data.
  2. [Data Availability statement] The Data Availability statement at the end of the paper reads 'No datasets were generated or analyzed during the current study.' This directly contradicts Section IV, which states that the models were trained and evaluated on MICCAI 2009, Kvasir-SEG, and BUS datasets. This internal inconsistency makes the experiments impossible to inspect or reproduce and must be corrected.
  3. [Sections IV.A and IV.C] The description of the data is internally inconsistent. Section IV.A identifies MICCAI 2009 as a brain tumor segmentation dataset ('multimodal brain pictures with an emphasis on brain tumor segmentation'), but Section IV.C refers to 'cardiac MRI images,' 'cardiac image size heterogeneity,' and '3,000 cardiac MRI images with masks.' The MICCAI 2009 challenge did involve brain tumor segmentation, so the cardiac references are unexplained and cast doubt on which data were actually used. The reported results cannot be interpreted without resolving this contradiction.
  4. [Sections V.A and VI] The training protocol is under-specified: Section V.A lists callbacks but omits the number of epochs, optimizer, learning rate, batch size, random seeds, and whether identical data splits were used for all models. Table II reports a single run per model with no variance, error bars, or significance tests. Consequently, the observed differences among VM-Unet, VM-Unet CBAM+, and VMSE-Unet cannot be attributed to the architecture changes rather than to training artifacts. The computational efficiency claims in Table III also lack a measurement protocol (input resolution, batch size, hardware specifics, number of runs), so the efficiency advantage asserted in the abstract is not substantiated.
minor comments (7)
  1. [Section I.C] The organization paragraph says Section II covers 'recent studies on the quantization of VM-Unet,' but Section II is a general related-work review; 'quantization' appears to be a typo.
  2. [Section II and References] Multiple in-text author names do not match the reference list: [12] is attributed to 'M. Zhang et al.' but the reference is by Xu, Li, and Chen; [15] is attributed to 'J. Wang et al.' but the reference is by Li, Chen, and Zhou; [16] is attributed to 'Z. Wang et al.' but the reference is by Wang, Zhang, and Liu; [17] is attributed to 'C. Yuan et al.' but the reference is by Zhou, Wang, and Xu.
  3. [Section II, paragraph on MS-UNet] The text states 'Kushnure et al. proposed MS-UNet [18]', but reference [18] is Calzolari and Liu, 'Deep learning to replace, improve, or aid CFD analysis...' which is unrelated; the MS-UNet citation should be [13].
  4. [Section IV.C, Eq. (16)] The ROI cropping notation x_resized[x_ROI, y_ROI] is ambiguous because x_ROI and y_ROI are not defined as ranges or coordinates; this should be clarified.
  5. [References] References [22] and [25] are the same paper (Tang et al., 'RM-UNet'), and [13] and [18] are inconsistently cited in the text; duplicate entries should be removed or consolidated.
  6. [Section VI and Figure 3] The caption of Figure 3 says 'Some perfect predicted masks from VMSE Unet,' but the term 'perfect' is unsupported by any quantitative analysis and no failure cases are shown; a neutral caption would be more appropriate.
  7. [Section VI] The paper claims 'state-of-the-art' performance but compares only against the VM-Unet baseline; comparisons with existing segmentation methods on these datasets are needed to support that claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical architecture comparison whose claims rest on reported benchmark evaluations, not on a derivation that reduces to its inputs.

full rationale

This manuscript makes no first-principles derivation and offers no prediction that is restated from a fitted value. The central claims—that VMSE-Unet and VM-Unet CBAM+ improve segmentation accuracy and efficiency over baseline VM-Unet—are supported by measured metrics in Tables II and III obtained on external datasets (MICCAI 2009, Kvasir-SEG, BUS). The attention modules (SE and CBAM) are standard, previously published components, and the baseline VM-Unet is cited from external prior work ([19]), so there is no load-bearing self-citation chain. Equations (1)-(13) simply restate the standard SE and CBAM definitions; they are architectural descriptions, not outputs derived from the paper's own results. The claimed efficiency advantage is an experimental measurement, not a consequence constructed from the inputs. Discrepancies between the abstract and Table II (e.g., VM-Unet CBAM+ having higher IoU on MICCAI 2009, and baseline VM-Unet having lower loss) and the Data Availability statement contradicting Section IV are serious correctness and reproducibility concerns, but they are not circularity: they do not show that a prediction is equivalent to its inputs by construction. No fitted parameter is renamed as a prediction, and no uniqueness theorem or prior result by the same authors is invoked to force the conclusion. The score is therefore 0.

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

The central claim rests on an empirical comparison rather than a mathematical derivation. The listed missing training details, unspecified architecture choices, and the contradictory data availability statement mean the comparison cannot be reconstructed or trusted from the manuscript alone.

free parameters (3)
  • SE reduction ratio = not reported
    The SE block's bottleneck dimension is a hand-chosen hyperparameter that affects capacity and parameter count, but the paper does not state it.
  • CBAM kernel size and block placement = not reported
    The number, location, and kernel size of CBAM blocks shown in Figure 1 are not specified precisely enough to reproduce the architecture.
  • Training hyperparameters = not reported
    Epochs, learning rate, batch size, optimizer, and random seeds are omitted from Section V, so the reported metrics cannot be reproduced or compared fairly.
assumptions (3)
  • domain assumption All three models were trained and evaluated under identical protocols and representative conditions.
    Section V lists callbacks but no hyperparameters, seeds, or confirmation that data splits were identical across models.
  • domain assumption The reported single-run metrics are statistically representative.
    Tables II and III give single numbers with no variance, confidence intervals, or significance tests.
  • ad hoc to paper The datasets were actually used as described.
    The Data Availability section says 'No datasets were generated or analyzed during the current study,' which directly contradicts the reported experiments on MICCAI 2009, Kvasir-SEG, and BUS.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Medical Image Segmentation Using Advanced Unet: VMSE-Unet and VM-Unet CBAM+." pith.science (2026). https://pith.science/paper/6VEYEFL7

@misc{pith2026250700511,
  author       = {Pith},
  title        = {Pith review of: Medical Image Segmentation Using Advanced Unet: VMSE-Unet and VM-Unet CBAM+},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6VEYEFL7}},
  note         = {Machine review of arXiv:2507.00511}
}
read the original abstract

In this paper, we present the VMSE U-Net and VM-Unet CBAM+ model, two cutting-edge deep learning architectures designed to enhance medical image segmentation. Our approach integrates Squeeze-and-Excitation (SE) and Convolutional Block Attention Module (CBAM) techniques into the traditional VM U-Net framework, significantly improving segmentation accuracy, feature localization, and computational efficiency. Both models show superior performance compared to the baseline VM-Unet across multiple datasets. Notably, VMSEUnet achieves the highest accuracy, IoU, precision, and recall while maintaining low loss values. It also exhibits exceptional computational efficiency with faster inference times and lower memory usage on both GPU and CPU. Overall, the study suggests that the enhanced architecture VMSE-Unet is a valuable tool for medical image analysis. These findings highlight its potential for real-world clinical applications, emphasizing the importance of further research to optimize accuracy, robustness, and computational efficiency.

Figures

Figures reproduced from arXiv: 2507.00511 by the authors.

Figure 1
Figure 1. Working diagram of our proposed model. medical images [27]. The SE block starts with a global pooling operation to generate channel-wise statistics: zc = 1 H × W X H i=1 X W j=1 Fc,i,j (1) where zc represents the aggregated global context for channel c. Next, these channel-wise statistics are passed through two fully connected layers with a non-linear activation function to produce recalibration weights: sc = σ(W2 ·… view at source ↗
Figure 2
Figure 2. Diagram of Squeeze Excitation Block and Convolutional Block Attention Module. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Some perfect predicted masks from VMSE Unet. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Loss Comparison. MICCAI 2009 Kvasir-SEG BUS Dataset Datasets 0 0.2 0.4 0.6 0.8 1 IoU Performance Comparison: IoU VM-Unet VM-Unet CBAM+ VMSE-Unet [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Intersection Over Union Comparison. Performance Comparison: Accuracy 0.991 0.840 0.956 0.991 0.992 0.980 0.990 0.968 0.992 MICCAI 2009 Kvasir-SEG BUS Dataset Datasets 0 0.2 0.4 0.6 0.8 1 1.2 Accuracy VM-Unet VM-Unet CBAM+ VMSE-Unet [PITH_FULL_IMAGE:figures/full_fig_p0…
Figure 6
Figure 6. Figure 6: Accuracy Comparison [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Precision Comparison. MICCAI 2009 KVASIR-SEG BUS Dataset Datasets 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Recall Comparison of Recall Metrics Across Datasets VM-Unet VM-Unet CBAM+ VMSE-Unet [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Recall Evaluation. underscore the effectiveness of our proposed enhancements, establishing VMSE-Unet as the optimal model for medical image segmentation tasks. The findings further highlight the potential of VMSE-Unet for real-world clinical applications, where accurac…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 26 canonical work pages

  1. [1]

    Deep learning applications in medical image analysis,

    A. Singha, R. S. Thakur, and T. Patel, “Deep learning applications in medical image analysis,”Biomedical Data Mining for Information Retrieval: Methodologies, Tech- niques and Applications, pp. 293–350, 2021

    work page 2021

  2. [2]

    U-net: Con- volutional networks for biomedical image segmenta- tion,

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

    work page 2015

  3. [3]

    VM-DDPM: Vision Mamba Diffusion for Medical Image Synthesis

    Z. Ju and W. Zhou, “Vm-ddpm: Vision mamba dif- fusion for medical image synthesis,”arXiv preprint arXiv:2405.05667, 2024

    work page Pith review arXiv 2024

  4. [4]

    Sonka, V

    M. Sonka, V . Hlavac, and R. Boyle,Image processing, analysis and machine vision. Springer, 2013

    work page 2013

  5. [5]

    State-of- the-art and challenges in pancreatic ct segmentation: A systematic review of u-net and its variants,

    C. Zhang, A. Achuthan, and G. M. S. Himel, “State-of- the-art and challenges in pancreatic ct segmentation: A systematic review of u-net and its variants,”IEEE Access, 2024

    work page 2024

  6. [6]

    Learning and transferring mid-level image representations using convolutional neural networks,

    M. Oquab, L. Bottou, I. Laptev, and J. Sivic, “Learning and transferring mid-level image representations using convolutional neural networks,” inProceedings of the IEEE conference on computer vision and pattern recog- nition, 2014, pp. 1717–1724

    work page 2014

  7. [7]

    Physics-inspired generative models in medical imaging: A review,

    D. Hein, A. Bozorgpour, D. Merhof, and G. Wang, “Physics-inspired generative models in medical imaging: A review,”arXiv preprint arXiv:2407.10856, 2024

    arXiv 2024

  8. [8]

    A survey on visual mamba,

    H. Zhang, Y . Zhu, D. Wang, L. Zhang, T. Chen, Z. Wang, and Z. Ye, “A survey on visual mamba,”Applied Sci- ences, vol. 14, no. 13, p. 5683, 2024

    work page 2024

Show all 32 references
  1. [9]

    Aggn: Attention-based glioma grading network 10 with multi-scale feature extraction and multi-modal infor- mation fusion,

    P. Wu, Z. Wang, B. Zheng, H. Li, F. E. Alsaadi, and N. Zeng, “Aggn: Attention-based glioma grading network 10 with multi-scale feature extraction and multi-modal infor- mation fusion,”Computers in biology and medicine, vol. 152, p. 106457, 2023

    2023

  2. [10]

    Trans- formers in medical image segmentation: A review,

    H. Xiao, L. Li, Q. Liu, X. Zhu, and Q. Zhang, “Trans- formers in medical image segmentation: A review,” Biomedical Signal Processing and Control, vol. 84, p. 104791, 2023

    2023

  3. [11]

    Image segmentation us- ing text and image prompts,

    T. L ¨uddecke and A. Ecker, “Image segmentation us- ing text and image prompts,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 7086–7096

    2022

  4. [12]

    Vm-unetv2: Rethinking vision mamba unet for medical image segmentation,

    X. Xu, X. Li, and K. Chen, “Vm-unetv2: Rethinking vision mamba unet for medical image segmentation,” arXiv preprint arXiv:2403.09157, 2023

    2023 arXiv

  5. [13]

    Ms-unet: A multi-scale unet with feature recalibration approach for automatic liver and tumor segmentation in ct images,

    D. T. Kushnure and S. N. Talbar, “Ms-unet: A multi-scale unet with feature recalibration approach for automatic liver and tumor segmentation in ct images,”Computer- ized Medical Imaging and Graphics, vol. 89, p. 101885, 2021

    2021

  6. [14]

    After-unet: Axial fusion transformer unet for medical image segmentation,

    X. Yan, H. Tang, S. Sun, H. Ma, D. Kong, and X. Xie, “After-unet: Axial fusion transformer unet for medical image segmentation,” inProceedings of the IEEE/CVF winter conference on applications of computer vision, 2022, pp. 3971–3981

    2022

  7. [15]

    Large window-based mamba unet for medical image segmentation: Be- yond convolution and self-attention,

    X. Li, K. Chen, and Z. Zhou, “Large window-based mamba unet for medical image segmentation: Be- yond convolution and self-attention,”arXiv preprint arXiv:2403.07332, 2023

    2023 arXiv

  8. [16]

    Efficient mamba u-net for robust medical image segmentation,

    Y . Wang, W. Zhang, and C. Liu, “Efficient mamba u-net for robust medical image segmentation,”IEEE Transac- tions on Medical Imaging, 2022

    2022

  9. [17]

    Adaptive mamba u-net for multi-modal medical image segmentation,

    Z. Zhou, Y . Wang, and X. Xu, “Adaptive mamba u-net for multi-modal medical image segmentation,”Medical Image Analysis, vol. 75, p. 102342, 2021

    2021

  10. [18]

    Deep learning to replace, improve, or aid cfd analysis in built environment appli- cations: A review,

    G. Calzolari and W. Liu, “Deep learning to replace, improve, or aid cfd analysis in built environment appli- cations: A review,”Building and Environment, vol. 206, p. 108315, 2021

    2021

  11. [19]

    Vm-unet: Vision mamba unet for medical image segmentation,

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

    2024 arXiv

  12. [21]

    Understanding deep learning techniques for image segmentation,

    S. Ghosh, N. Das, I. Das, and U. Maulik, “Understanding deep learning techniques for image segmentation,”ACM computing surveys (CSUR), vol. 52, no. 4, pp. 1–35, 2019

    2019

  13. [23]

    Transformers in medical imaging: A survey,

    F. Shamshad, S. Khan, S. W. Zamir, M. H. Khan, M. Hayat, F. S. Khan, and H. Fu, “Transformers in medical imaging: A survey,”Medical Image Analysis, vol. 88, p. 102802, 2023

    2023

  14. [24]

    Computation-efficient era: A compre- hensive survey of state space models in medical image analysis,

    M. Heidari, S. G. Kolahi, S. Karimijafarbigloo, B. Azad, A. Bozorgpour, S. Hatami, R. Azad, A. Diba, U. Bagci, D. Merhofet al., “Computation-efficient era: A compre- hensive survey of state space models in medical image analysis,”arXiv preprint arXiv:2406.03430, 2024

    2024 arXiv

  15. [25]

    Rm-unet: Unet-like mamba with rotational ssm module for medical image segmen- tation,

    H. Tang, G. Huang, L. Cheng, X. Yuan, Q. Tao, X. Chen, G. Zhong, and X. Yang, “Rm-unet: Unet-like mamba with rotational ssm module for medical image segmen- tation,”Signal, Image and Video Processing, vol. 18, no. 11, pp. 8427–8443, 2024

    2024

  16. [26]

    Efficient spineunetx for x-ray: A spine segmentation network based on convnext and unet,

    S. Deng, Y . Yang, J. Wang, A. Li, and Z. Li, “Efficient spineunetx for x-ray: A spine segmentation network based on convnext and unet,”Journal of Visual Commu- nication and Image Representation, vol. 103, p. 104245, 2024

    2024

  17. [27]

    Embracing imperfect datasets: A review of deep learning solutions for medical image segmentation,

    N. Tajbakhsh, L. Jeyaseelan, Q. Li, J. N. Chiang, Z. Wu, and X. Ding, “Embracing imperfect datasets: A review of deep learning solutions for medical image segmentation,” Medical image analysis, vol. 63, p. 101693, 2020

    2020

  18. [28]

    A cnn-transformer network combining cbam for change detection in high- resolution remote sensing images,

    M. Yin, Z. Chen, and C. Zhang, “A cnn-transformer network combining cbam for change detection in high- resolution remote sensing images,”Remote Sensing, vol. 15, no. 9, p. 2406, 2023

    2023

  19. [29]

    Center-boundary dual attention for oriented object detection in remote sensing images,

    S. Liu, L. Zhang, H. Lu, and Y . He, “Center-boundary dual attention for oriented object detection in remote sensing images,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–14, 2021

    2021

  20. [30]

    G.-Z. Yang, D. J. Hawkes, D. Rueckert, A. Noble, and C. Taylor,Medical Image Computing and Computer- Assisted Intervention–MICCAI 2009: 12th International Conference, London, UK, September 20-24, 2009, Pro- ceedings. Springer Science & Business Media, 2009, vol. 1

    2009

  21. [31]

    Jiang, N

    T. Jiang, N. Navab, J. P. Pluim, and M. A. Viergever, Medical Image Computing and Computer-Assisted Intervention–MICCAI 2010: 13th International Confer- ence, Beijing, China, September 20-24, 2010, Proceed- ings, Part III. Springer, 2010, vol. 6363

    2010

  22. [32]

    Kvasir- seg: A segmented polyp dataset,

    D. Jha, P. H. Smedsrud, M. A. Riegler, P. Halvorsen, T. De Lange, D. Johansen, and H. D. Johansen, “Kvasir- seg: A segmented polyp dataset,” inMultiMedia model- ing: 26th international conference, MMM 2020, Daejeon, South Korea, January 5–8, 2020, proceedings, part II 26. Spri...

    2020

  23. [33]

    Bus-set: A benchmark for quantitative evaluation of breast ultrasound segmentation networks with public datasets,

    C. Thomas, M. Byra, R. Marti, M. H. Yap, and R. Zwiggelaar, “Bus-set: A benchmark for quantitative evaluation of breast ultrasound segmentation networks with public datasets,”Medical Physics, vol. 50, no. 5, pp. 3223–3243, 2023

    2023

  24. [34]

    Deep neural architectures for medical image semantic segmentation,

    M. Z. Khan, M. K. Gajendran, Y . Lee, and M. A. Khan, “Deep neural architectures for medical image semantic segmentation,”IEEE Access, vol. 9, pp. 83 002–83 024, 2021

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.