Pith. sign in

REVIEW 5 major objections 3 minor 33 references

MS-UMamba: An Improved Vision Mamba Unet for Fetal Abdominal Medical Image Segmentation

T0 review · 5 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read MS-UMamba, a hybrid CNN-Mamba U-Net, reports higher mIoU, mDice, and mSE than VM-UNet on a private fetal abdominal ultrasound dataset.

desk verdict A plausible recombined Mamba-UNet for fetal ultrasound, but the core empirical claim is unverifiable as presented: private data, no code, no error bars, and a 1.97-point mIoU margin that the paper itself never shows is outside run-to-run noise. read the letter →

arxiv 2506.12441 v1 pith:JGAQJZ3N submitted 2025-06-14 cs.CV cs.AI

Caixu Xu , Junming Wei , Huizhen Chen , Pengchen Liang , Bocheng Liang , Ying Tan , Xintong Wei This is my paper
classification cs.CVcs.AI
keywords fetalultrasoundsegmentationVisionMambastatespacemodelsmulti-scalefeaturefusionmedicalimagehybridCNN-MambaU-Netattentionmechanism
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

MS-UMamba is a semantic segmentation network for six anatomical structures in fetal abdominal ultrasound images. It combines a Mamba state-space branch for long-range context with a CNN branch that uses a Monte Carlo attention block for local detail, and adds an attention-based multi-scale feature fusion module between encoder and decoder. On a private dataset of 696 images the authors report mIoU of 67.62%, mDice of 79.82%, and mSE of 84.78%, all above VM-UNet, the strongest baseline in their comparison. They also state in the conclusion that small structures remain the hardest cases, the dataset is scarce, and the model is parameter-heavy, so the practical payoff depends on later lightweight and semi-supervised improvements.

What carries the argument

The load-bearing object is the SS-MCAT-SSM block, which channel-splits its input into two halves: one goes through a Mamba (selective state-space) branch that models long-range dependencies, and the other through a CNN branch whose BottleNeck block applies Monte Carlo attention, reweighting multi-scale pooled features by association probabilities before 1x1 and 3x3 convolutions, after which the branches are concatenated and channel-shuffled. The second load-bearing piece is the Attention-based Dynamic Feature Fusion (ADFF) module, adapted from the DFF module: it maps encoder and decoder features through 1x1 convolutions, sums them, forms a spatial attention weight with a Sigmoid, concatenates the original features, applies channel attention via a shared MLP on global average and max pooling, and combines both weightings. Together these modules are what the ablation study varies to attribute the performance gain.

What would settle it

Re-run the same comparison on a fixed public fetal ultrasound dataset with patient-disjoint folds and report per-class Dice; if the mIoU gap over VM-UNet drops below the reported 1.97 points, the claimed advantage is an artifact of the private split or labels. Additionally, reporting inter-annotator agreement on the 696 images would show whether label noise explains the small-target errors.

Watch

Extended reading notes

Core claim

The paper's central claim is that adding a CNN branch with Monte Carlo attention to a Mamba-based U-Net, together with an attention-weighted dynamic feature fusion module, yields better segmentation of small, overlapping fetal abdominal organs than Mamba-only or CNN-only alternatives. The two proposed modules are named SS-MCAT-SSM, a visual state-space block split into a CNN branch and a Mamba branch, and ADFF, a multi-scale fusion module with spatial and channel attention. Ablation results show the combination reaches the best numbers, while replacing the fused module with the plain DFF module drops mIoU from 67.62% to about 53.54%, which the authors interpret as evidence that the attention-based fusion is what makes the model work. Per-structure results show the largest scores on the L V class and the lowest on SL and UV, the two small-target classes.

Load-bearing premise

The load-bearing premise is that the private 696-image dataset is split and annotated in a way that does not systematically favor MS-UMamba; the paper does not describe how train, validation, and test sets were built or how consistent the technician labels are.

Editorial extensions

If this is right

  • If the reported results hold on patient-disjoint data, automated fetal ultrasound screening could measure abdominal structures without manual contouring, reducing the subjectivity and time cost the paper cites as motivation.
  • The SS-MCAT-SSM and ADFF modules are described generally enough to be inserted into other encoder-decoder segmentation backbones, giving evidence that hybrid CNN-Mamba fusion is competitive on ultrasound, not only on MRI or CT.
  • The ablation says plain DFF fusion hurts performance, implying the attention mechanisms in ADFF are necessary; future hybrid models should expect naive multi-scale fusion to backfire.
  • Per-structure results identify small targets (SL, UV) as the bottleneck, so the next natural increment is a loss function or training scheme targeted at small-object recall rather than a larger backbone.

Reading between the lines

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

  • The private dataset has no described patient-disjoint split or inter-annotator agreement; before relying on the 1.97-point mIoU gap over VM-UNet, a reader would want the data released or a public fetal ultrasound benchmark run under identical training settings.
  • The table's precision column is internally inconsistent with the prose claim of outperforming all models, because SegMenter shows a higher mPRE, so the defensible claim is the mIoU, mDice, and mSE gain rather than an all-metric win.
  • A testable extension would be to ablate only the Monte Carlo attention inside SS-MCAT-SSM while keeping ADFF fixed, to see whether the gain comes from the CNN branch itself or from the attention mechanism.
  • If the same architecture transfers to fetal cardiac or head planes, it would support the more general claim that hybrid CNN-Mamba designs handle ultrasound speckle and boundary blur better than either family alone; the paper only tests one plane.
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

5 major / 3 minor

Summary. The paper proposes MS-UMamba, a U-Net-style architecture that combines Vision Mamba state space blocks with a CNN branch (SS-MCAT-SSM) and an attention-based multi-scale feature fusion module (ADFF), for segmenting six fetal abdominal anatomical structures in ultrasound images. The authors report experiments on a private dataset of 696 images, comparing against several existing segmentation models and claiming that MS-UMamba achieves the best performance, with the main result being an mIoU of 67.62% versus 65.65% for VM-UNet. They also provide an ablation study and qualitative visualizations.

Significance. If the reported results are reliable, the proposed hybrid Mamba-CNN architecture and the attention-based fusion module would be a useful contribution to medical image segmentation, particularly for fetal ultrasound where boundary ambiguity and small structures are challenging. The paper includes a reasonable set of baseline comparisons and covers multiple evaluation metrics. However, the significance is currently limited by the use of a private dataset, absence of code and data release, lack of statistical validation (no error bars or significance tests), and incomplete experimental protocol details, all of which make it difficult to verify the claimed architectural advantage.

major comments (5)
  1. [Section 4.1, Table 1] The dataset split is not described in a way that supports the claimed results. The paper states that 696 images were annotated but does not explain how the train/validation/test sets were constructed, whether the split is patient-disjoint, or whether a single image can contribute to multiple per-structure counts in Table 1. Without this information, the effective test-set size and the independence of per-structure metrics are unclear, so the reported performance gap may be an artifact of the split rather than of the architecture.
  2. [Section 4.3, Table 2] No run-to-run variance or statistical significance testing is reported for any of the comparisons. A single training run per model on a private dataset cannot establish that the 1.97-point mIoU advantage over VM-UNet is outside the range of random initialization, augmentation order, and batch composition. In addition, the sentence claiming that MS-UMamba 'outperforms all other models in these metrics' is directly contradicted by the next sentence and by Table 2, where SegMenter reports a higher mPRE (76.54% versus 76.11%).
  3. [Section 3.4, Eq. (9)] The Focal loss equation as printed appears degenerate and consequently the exact loss used for training is not reproducible. In the second term, the factor p^y (1-y) equals 1 for y=0, which removes the focusing term for negative samples, and for y=1 it becomes 0, making the expression inconsistent with the standard focal loss; additionally, the exponent on (1−P) in the first term is ambiguous. The authors need to provide a correct, clearly notated formula that matches their implementation.
  4. [Table 4, ablation experiment] The DFF-only ablation row shows a catastrophic drop in performance, with mIoU falling from 66.54% in the baseline to 53.54% when the basic DFF module is added, while the text gives no explanation for this behavior. This result is unexpected given that the ADFF module—which is claimed to build on DFF—improves performance when added together with MCAT. The authors should explain what causes this degradation, or the comparison must be considered uncontrolled and the role of the DFF module unclear.
  5. [Section 4.3, comparison protocol] The experimental protocol for baseline comparisons is under-specified. The manuscript does not state whether all models were trained with identical training budgets, input resolutions, data augmentation, and hyperparameter tuning procedures. Without this information, the relative improvements over baselines cannot be attributed to the proposed SS-MCAT-SSM and ADFF modules rather than to differences in training configuration.
minor comments (3)
  1. [Section 4.3, Table 2] The last cell of Table 2 reads '99.3176.11' with a missing separator, and the text reports an mPRE of 76.09% while the table shows 76.11%; these values should be reconciled.
  2. [Throughout] There are several typos and grammatical errors, including 'Blcok' in section headings, 'Sgmiod' in Eq. (4), 'Swim-TransFormer' instead of 'Swin-Transformer', and the abstract's incomplete sentence 'which Integrating feature information from different layers enhances the feature representation ability of the model.'
  3. [Section 4.1] Table 1 lists per-structure counts for train/validation/test, but it is unclear whether these counts refer to images, instances, or annotated regions; because each image contains multiple structures, the counts may overlap, and the table should be clarified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmarking study with external baselines and standard metrics, and none of its claims reduce to fitted inputs or self-cited theorems.

full rationale

MS-UMamba is an empirical architecture-comparison paper. The central claim is a measured performance advantage over VM-UNet and other baselines on a non-public fetal ultrasound dataset using standard metrics (mIoU, mDC, mSE, mSP, mPRE). No quantity is defined in terms of another quantity in a way that forces the outcome: the SS-MCAT-SSM block and ADFF fusion module are introduced as architectural proposals with explicit equations for their internal operations, but these equations do not encode the evaluation metrics or the reported results. The baseline comparisons use external published models, and the ablation study compares the proposed model against its own variants, which is a controlled experiment rather than a circular derivation. There are no fitted parameters that are later relabeled as predictions, no uniqueness theorem from the authors' prior work, and no load-bearing self-citation: the reference list contains external prior work, and the present authors do not appear to cite their own earlier results to justify the central premise. The paper's stated limitations--poor small-structure segmentation, scarce data, and large model parameters--are explicit weakness disclosures, not circularity. Concerns about the undocumented train/validation/test split and lack of statistical significance testing are external-validity or reproducibility risks, not circular reasoning. The derivation chain is therefore self-contained as an empirical study, and no circular step can be exhibited with specific equations or self-citations.

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

The center of the paper is an empirical segmentation benchmark. It rests on a private annotated dataset, an unreported split, fair baseline training, and a correctly implemented loss. None of these are independently verifiable from the preprint, and the paper introduces no new physical or mathematical entities.

free parameters (4)
  • Focal loss alpha (alpha) = not reported
    Used in Eq. (9) to reweight hard/easy samples; its value is not stated, but the final loss and therefore the results depend on it.
  • Focal loss gamma (gamma) = not reported
    Used in Eq. (9) to control the focusing effect; not reported in the paper.
  • Loss combination weight between Focal and Dice loss = not reported
    The text says the loss combines Focal and Dice terms but no weighting coefficient is given.
  • Number of attention scales n in Eq. (2) = not reported
    The Monte Carlo attention sum runs over i with unspecified n; the paper does not define this value or how it was chosen.
assumptions (4)
  • domain assumption Manual annotations of the six fetal structures are correct and consistent.
    The paper relies on labels drawn by ultrasound technicians (Section 4.1) but reports no inter-observer agreement or quality control.
  • domain assumption The train/validation/test split is unbiased and leakage-free.
    No split procedure or patient-overlap check is described; Table 1 only reports per-structure counts.
  • domain assumption Baseline models were trained fairly and comparably.
    No training code, hyperparameters, or compute budget per baseline is provided, so fair comparison is assumed.
  • ad hoc to paper The loss functions in Eqs. (9)-(10) are implemented as intended.
    The focal loss as printed contains y(1-y) in the second term, which would make the loss degenerate; the authors may have intended a standard focal loss, but as written it is not implementable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MS-UMamba: An Improved Vision Mamba Unet for Fetal Abdominal Medical Image Segmentation." pith.science (2026). https://pith.science/paper/JGAQJZ3N

@misc{pith2026250612441,
  author       = {Pith},
  title        = {Pith review of: MS-UMamba: An Improved Vision Mamba Unet for Fetal Abdominal Medical Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JGAQJZ3N}},
  note         = {Machine review of arXiv:2506.12441}
}
read the original abstract

Recently, Mamba-based methods have become popular in medical image segmentation due to their lightweight design and long-range dependency modeling capabilities. However, current segmentation methods frequently encounter challenges in fetal ultrasound images, such as enclosed anatomical structures, blurred boundaries, and small anatomical structures. To address the need for balancing local feature extraction and global context modeling, we propose MS-UMamba, a novel hybrid convolutional-mamba model for fetal ultrasound image segmentation. Specifically, we design a visual state space block integrated with a CNN branch (SS-MCAT-SSM), which leverages Mamba's global modeling strengths and convolutional layers' local representation advantages to enhance feature learning. In addition, we also propose an efficient multi-scale feature fusion module that integrates spatial attention mechanisms, which Integrating feature information from different layers enhances the feature representation ability of the model. Finally, we conduct extensive experiments on a non-public dataset, experimental results demonstrate that MS-UMamba model has excellent performance in segmentation performance.

Figures

Figures reproduced from arXiv: 2506.12441 by the authors.

Figure 1
Figure 1. The illustration of MS-UMamba architecture. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Design Details of the SS-MCAT-SSM Module. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Design Details of the Attention-Based Dynamic Feature Fusion Module. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of segmentation results. The red, yellow, yellow-green, [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 25 canonical work pages

  1. [1]

    Cmaj182(2), 143–151 (2010)

    You, J.J., Alter, D.A., Stukel, T.A., McDonald, S.D., Laupacis, A., Liu, Y., Ray, J.G.: Proliferation of prenatal ultrasonography. Cmaj182(2), 143–151 (2010)

    work page 2010

  2. [2]

    IEEE journal of biomedical and health informatics22(5), 1512–1520 (2017)

    Jang, J., Park, Y., Kim, B., Lee, S.M., Kwon, J.Y., Seo, J.K.: Automatic estima- tion of fetal abdominal circumference from ultrasound images. IEEE journal of biomedical and health informatics22(5), 1512–1520 (2017)

    work page 2017

  3. [3]

    Journal of clinical medicine 12(9), 3298 (2023)

    Xiao, S., Zhang, J., Zhu, Y., Zhang, Z., Cao, H., Xie, M., Zhang, L.: Application and progress of artificial intelligence in fetal ultrasound. Journal of clinical medicine 12(9), 3298 (2023)

    work page 2023

  4. [4]

    IEEE Journal of Biomedical and Health Informatics (2024)

    Pu, B., Li, K., Chen, J., Lu, Y., Zeng, Q., Yang, J., Li, S.: Hfsccd: a hybrid neural network for fetal standard cardiac cycle detection in ultrasound videos. IEEE Journal of Biomedical and Health Informatics (2024)

    work page 2024

  5. [5]

    Computers in Biology and Medicine165, 107,399 (2023)

    Gao, Z., Tian, Z., Pu, B., Li, S., Li, K.: Deep endpoints focusing network under geometric constraints for end-to-end biometric measurement in fetal ultrasound images. Computers in Biology and Medicine165, 107,399 (2023)

    work page 2023

  6. [6]

    Neural Computing and Applications pp

    Wu, X., Tan, G., Pu, B., Duan, M., Cai, W.: Dh-gac: Deep hierarchical context fu- sion network with modified geodesic active contour for multiple neurofibromatosis segmentation. Neural Computing and Applications pp. 1–16 (2022)

    work page 2022

  7. [7]

    Computers in Biology and Medicine169, 107,898 (2024)

    Chen, G., Tan, G., Duan, M., Pu, B., Luo, H., Li, S., Li, K.: Mlmseg: a multi-view learning model for ultrasound thyroid nodule segmentation. Computers in Biology and Medicine169, 107,898 (2024)

    work page 2024

  8. [8]

    In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, Oc- tober 5-9, 2015, proceedings, part III 18, pp

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomed- ical image segmentation. In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, Oc- tober 5-9, 2015, proceedings, part III 18, pp. 234–241. Springer (2015)

    work page 2015

Show all 33 references
  1. [9]

    IEEE transac- tions on medical imaging39(6), 1856–1867 (2019)

    Zhou, Z., Siddiquee, M.M.R., Tajbakhsh, N., Liang, J.: Unet++: Redesigning skip connections to exploit multiscale features in image segmentation. IEEE transac- tions on medical imaging39(6), 1856–1867 (2019)

    2019

  2. [10]

    In: ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp

    Huang, H., Lin, L., Tong, R., Hu, H., Zhang, Q., Iwamoto, Y., Han, X., Chen, Y.W., Wu, J.: Unet 3+: A full-scale connected unet for medical image segmentation. In: ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP), pp. 1055–1059...

    2020

  3. [11]

    arXiv preprint arXiv:1804.03999 (2018) MS-UMamba 13

    Oktay, O., Schlemper, J., Folgoc, L.L., Lee, M., Heinrich, M., Misawa, K., Mori, K., McDonagh, S., Hammerla, N.Y., Kainz, B., et al.: Attention u-net: Learning where to look for the pancreas. arXiv preprint arXiv:1804.03999 (2018) MS-UMamba 13

    2018 arXiv

  4. [12]

    Health Information Science and Systems11(1), 13 (2023)

    Zhang, J., Zhang, Y., Jin, Y., Xu, J., Xu, X.: Mdu-net: Multi-scale densely con- nected u-net for biomedical image segmentation. Health Information Science and Systems11(1), 13 (2023)

    2023

  5. [13]

    In: European conference on computer vision, pp

    Cao, H., Wang, Y., Chen, J., Jiang, D., Zhang, X., Tian, Q., Wang, M.: Swin- unet: Unet-like pure transformer for medical image segmentation. In: European conference on computer vision, pp. 205–218. Springer (2022)

    2022

  6. [14]

    Computers in Biology and Medicine159, 106,960 (2023)

    Zhang, H., Zhong, X., Li, G., Liu, W., Liu, J., Ji, D., Li, X., Wu, J.: Bcu-net: Bridg- ing convnext and u-net for medical image segmentation. Computers in Biology and Medicine159, 106,960 (2023)

    2023

  7. [15]

    IEEE Transactions on Emerging Topics in Computational Intelligence8(1), 55–68 (2023)

    Chen, B., Liu, Y., Zhang, Z., Lu, G., Kong, A.W.K.: Transattunet: Multi-level attention-guided u-net with transformer for medical image segmentation. IEEE Transactions on Emerging Topics in Computational Intelligence8(1), 55–68 (2023)

    2023

  8. [16]

    arXiv preprint arXiv:2312.00752 (2023)

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

    2023 arXiv

  9. [17]

    arXiv preprint arXiv:2401.04722 (2024)

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

    2024 arXiv

  10. [18]

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

    Xing, Z., Ye, T., Yang, Y., Liu, G., Zhu, L.: Segmamba: Long-range sequential modeling mamba for 3d medical image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 578–588. Springer (2024)

    2024

  11. [19]

    arXiv preprint arXiv:2402.02491 (2024)

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

    2024 arXiv

  12. [20]

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

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

    2024

  13. [21]

    arXiv preprint arXiv:2402.05079 (2024)

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

    2024 arXiv

  14. [22]

    arXiv preprint arXiv:2402.07245 (2024)

    Ma, C., Wang, Z.: Semi-mamba-unet: Pixel-level contrastive and pixel-level cross- supervised visual mamba-based unet for semi-supervised medical image segmenta- tion. arXiv preprint arXiv:2402.07245 (2024)

    2024 arXiv

  15. [23]

    arXiv e-prints pp

    Wang, J., Chen, J., Chen, D., Wu, J.: Large window-based mamba unet for medical image segmentation: Beyond convolution and self-attention. arXiv e-prints pp. arXiv–2403 (2024)

    2024

  16. [24]

    IEEE Journal of Biomedical and Health Informatics 26(11), 5540–5550 (2022)

    Pu, B., Lu, Y., Chen, J., Li, S., Zhu, N., Wei, W., Li, K.: Mobileunet-fpn: A semantic segmentation model for fetal ultrasound four-chamber segmentation in edge computing environments. IEEE Journal of Biomedical and Health Informatics 26(11), 5540–5550 (2022)

    2022

  17. [25]

    BMC Medical Informatics and Decision Making24(1), 128 (2024)

    Liang, B., Peng, F., Luo, D., Zeng, Q., Wen, H., Zheng, B., Zou, Z., An, L., Wen, H., Wen, X., et al.: Automatic segmentation of 15 critical anatomical labels and measurements of cardiac axis and cardiothoracic ratio in fetal four chambers using nnu-netv2. BMC Medical Informat...

    2024

  18. [26]

    IEEE/ACM Transactions on Computational Biology and Bioinformatics (2022)

    Lu, Y., Li, K., Pu, B., Tan, Y., Zhu, N.: A yolox-based deep instance segmenta- tion neural network for cardiac anatomical structures in fetal ultrasound images. IEEE/ACM Transactions on Computational Biology and Bioinformatics (2022)

    2022

  19. [27]

    IEEE Journal of Biomedical and Health Informatics28(1), 285–296 (2023) 14 CaiXu Xu et al

    Zhao, L., Tan, G., Pu, B., Wu, Q., Ren, H., Li, K.: Transfsm: Fetal anatomy segmentation and biometric measurement in ultrasound images using a hybrid transformer. IEEE Journal of Biomedical and Health Informatics28(1), 285–296 (2023) 14 CaiXu Xu et al

    2023

  20. [28]

    IEEE transactions on cybernetics47(5), 1336–1349 (2017)

    Wu, L., Cheng, J.Z., Li, S., Lei, B., Wang, T., Ni, D.: Fuiqa: fetal ultrasound image quality assessment with deep convolutional networks. IEEE transactions on cybernetics47(5), 1336–1349 (2017)

    2017

  21. [29]

    Physica Medica88, 127–137 (2021)

    Oghli, M.G., Shabanzadeh, A., Moradi, S., Sirjani, N., Gerami, R., Ghaderi, P., Taheri, M.S., Shiri, I., Arabi, H., Zaidi, H.: Automatic fetal biometry prediction using a novel deep convolutional network architecture. Physica Medica88, 127–137 (2021)

    2021

  22. [30]

    Multimedia Tools and Applications83(9), 27,283–27,304 (2024)

    Liu, L., Tang, D., Li, X., Ouyang, Y.: Automatic fetal ultrasound image segmenta- tion of first trimester for measuring biometric parameters based on deep learning. Multimedia Tools and Applications83(9), 27,283–27,304 (2024)

    2024

  23. [31]

    IEEE Open Journal of Engineering in Medicine and Biology (2024)

    Alzubaidi, M., Shah, U., Agus, M., Househ, M.: Fetsam: Advanced segmentation techniques for fetal head biometrics in ultrasound imagery. IEEE Open Journal of Engineering in Medicine and Biology (2024)

    2024

  24. [32]

    arXiv preprint arXiv:2403.03849 (2024)

    Yue, Y., Li, Z.: Medmamba: Vision mamba for medical image classification. arXiv preprint arXiv:2403.03849 (2024)

    2024 arXiv

  25. [33]

    arXiv preprint arXiv:2403.10674 (2024)

    Yang, J., Qiu, P., Zhang, Y., Marcus, D.S., Sotiras, A.: D-net: Dynamic large kernel with dynamic feature fusion for volumetric medical image segmentation. arXiv preprint arXiv:2403.10674 (2024)

    2024 arXiv

Pith tools

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