Pith. sign in

REVIEW 4 major objections 6 minor 42 references

Novel Extraction of Discriminative Fine-Grained Feature to Improve Retinal Vessel Segmentation

T0 review · 4 major / 6 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 The paper claims that AttUKAN, a U-shaped Kolmogorov-Arnold network with attention-gated skip connections and a label-guided pixel-wise contrastive loss, extracts more discriminative encoder features and reports the highest F1 and MIoU…

desk verdict A workmanlike combination of known components with broad but unverified ablations; the MIoU formula as printed contradicts every reported table. read the letter →

arxiv 2505.03896 v1 pith:NLLBUCNZ submitted 2025-05-06 cs.CV cs.AI

Shuang Zeng , Chee Hong Lee , Micky C Nnamdi , Wenqi Shi , J Ben Tamo , Lei Zhu , Hangzhou He , Xinliang Zhang
show 4 more authors
Qian Chen May D. Wang Yanye Lu Qiushi Ren
This is my paper · ORCID
classification cs.CVcs.AI
keywords retinalvesselsegmentationfundusimageKolmogorov-ArnoldNetworksattentionmechanismcontrastivelearningfine-grainedfeatureextractionmedical
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

Retinal vessel segmentation is a pixel-level classification problem where the hard part is distinguishing vessel from non-vessel within a single image, so the model needs fine-grained, discriminative features rather than just a decoder that matches labels. The paper proposes AttUKAN, a U-shaped Kolmogorov-Arnold network, together with a Label-guided Pixel-wise Contrastive Loss (LPCL) that pulls same-class pixel representations together and pushes different-class pixels apart. Attention gates on the skip connections filter the encoder features that flow to the decoder. The paper reports that AttUKAN achieves the highest F1 and MIoU among 12 compared methods on DRIVE, STARE, CHASE_DB1, HRF, and a private dataset, attributing the gains to the attention gates and the contrastive loss.

What carries the argument

The machinery is two additions to a U-shaped Kolmogorov-Arnold Network (KAN). First, Attention Gates in the skip connections compute a gating coefficient from the encoder feature and a higher-level gating signal, multiply the feature map elementwise, and suppress irrelevant activations before concatenation with the decoder. Second, the Label-guided Pixel-wise Contrastive Loss (LPCL) samples pixel-level feature vectors from the bottleneck feature map; pairs of pixels with the same ground-truth class are positives and different-class pairs are negatives, and a contrastive objective pulls same-class vectors together and pushes different-class vectors apart in representation space. The KAN blocks themselves replace linear layers with learnable nonlinear activation functions, giving the encoder more nonlinear modeling capacity.

What would settle it

Retrain AttUKAN and the 11 baselines under one fixed public protocol (same splits, same preprocessing, several random seeds) on DRIVE, STARE, and CHASE_DB1, then compare each method's F1 mean and standard deviation; if AttUKAN's lead over the second-best method is smaller than the seed-to-seed spread, the reported advantage would fall within training noise.

Watch

Extended reading notes

Core claim

The central claim is that feature-level supervision is the missing ingredient in retinal vessel segmentation. Existing models train primarily by comparing decoder output with ground-truth labels, but AttUKAN applies a contrastive loss directly to the bottleneck feature map, using the label to define positive pairs (same class) and negative pairs (different classes). Attention gates in skip connections suppress irrelevant activations so finer vessel structure is retained. The paper reports F1 scores of 82.50%, 81.14%, 81.34%, 80.21%, and 80.09% and MIoU scores of 70.24%, 68.64%, 68.59%, 67.21%, and 66.94% across the five datasets, each the highest among the 11 baselines, and ablation studies attribute the gains to both the attention gates and LPCL.

Load-bearing premise

The comparison assumes every one of the 12 methods is trained and evaluated under identical splits, preprocessing, and hyperparameter settings, and that the reported single-run differences of about 0.1 to 1.0 F1 points reflect real model differences rather than training noise.

Editorial extensions

If this is right

  • If AttUKAN's results hold, feature-level contrastive supervision at the bottleneck is an effective complement to decoder-level losses for retinal vessel segmentation.
  • The ablations show LPCL also improves several existing U-Net style baselines when added to them, suggesting the loss transfers beyond the proposed architecture.
  • Attention-gated skip connections on a KAN backbone may generalize to other fine-structure segmentation tasks where thin structures dominate.
  • The reported gains on high-resolution datasets (HRF and the private set) indicate the approach works outside the small-patch regime.
  • The reported F1 and MIoU numbers, if reproducible, establish new quantitative baselines for these five datasets.

Reading between the lines

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

  • Because LPCL is applied only at the bottleneck, one could test whether applying it at multiple decoder levels further improves small vessel tips; the paper does not explore this.
  • The method uses label-guided pixel pairs rather than heavy augmentation or large contrastive batches, which may make it easier to adopt when training data are scarce.
  • A direct probe of feature-map cluster purity (vessel vs background) could test whether LPCL's mechanism is indeed better feature separability and whether it correlates with F1 gain.
  • When the private dataset is released, external groups can check whether the advantage persists under an independent evaluation protocol.
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 / 6 minor

Summary. The paper proposes AttUKAN, a U-shaped Kolmogorov-Arnold Network with Attention Gates placed on skip connections, and a Label-guided Pixel-wise Contrastive Loss (LPCL) to improve retinal vessel segmentation. The method is evaluated on DRIVE, STARE, CHASE_DB1, HRF, and a private dataset, with F1 and MIoU reported as the primary metrics. The authors claim state-of-the-art results against 11 baseline networks across all five datasets, and support the claim with ablation studies on the network components, loss terms, the LPCL weighting coefficient, the feature level used for LPCL, and the transferability of LPCL to other architectures.

Significance. If the results are validated, the paper offers a practical and portable improvement: attention-gated KAN blocks plus a supervised contrastive loss that also improves several existing U-shaped networks. The study is unusually broad in its evaluation, covering five datasets and multiple ablations, and the transfer experiment of LPCL across architectures is a genuine strength. However, the central quantitative claim is currently undermined by an internally inconsistent metric definition, a complete absence of statistical uncertainty, and an unclear statement of the spatial resolution at which LPCL operates. The reported F1 gains over the best baselines are small (0.07 to 0.54 percentage points), so without error bars or significance testing the 'highest compared to 11 networks' claim is not yet demonstrated. The paper ships no code and gives no experimental logs, further limiting verification.

major comments (4)
  1. [Section 4.2, Eq. (20)] The definition of MIoU is internally inconsistent with all reported results. Equation (19) defines F1 = 2TP/(2TP+FP+FN) and Eq. (20) defines MIoU = 2TP/(TP+FP+FN). Since 2TP+FP+FN > TP+FP+FN for any non-degenerate confusion matrix, Eq. (20) forces MIoU > F1 in every row, yet every entry of Tables 2–7 shows the opposite. For example, DRIVE AttUKAN reports F1=82.50% and MIoU=70.24%; combining Eqs. (19) and (20) algebraically gives MIoU = 2F1/(2−F1) = 140.4% for F1=82.50%, not 70.24%. The reported MIoU values are numerically consistent with the standard foreground IoU, TP/(TP+FP+FN): for F1=82.50%, IoU=82.50/(200−82.50)=70.21%, matching 70.24% up to rounding. The formula must be corrected (most plausibly to TP/(TP+FP+FN)) or the reported MIoU values cannot be reproduced. Because the abstract and Section 4.4 base the SOTA claim jointly on F1 and MIoU, this is a load-bearing defect.
  2. [Section 4.4, Tables 2–6] No variance, error bars, or statistical significance tests are reported. The F1 advantages over the best baseline are 0.22 points on DRIVE, 0.15 on STARE, 0.54 on CHASE_DB1, 0.07 on HRF, and 0.11 on the private dataset, while the component ablations in Table 8 improve F1 by 0.10 to 0.75 points. With a single run and no reported seeds, these differences could easily fall within training noise. Please provide mean and standard deviation over at least three random seeds, or an equivalent paired statistical test, for the main comparisons and the ablations, so the claimed superiority can be distinguished from stochastic variation.
  3. [Section 3.4 and Section 4.3] LPCL is computed on the 5th-level bottleneck feature of size S×S. With the stated patch size of 64×64 and the 32× downsampling at the bottleneck shown in Figure 2, S=2, so only four spatial positions per image are available as anchors. This is at odds with the description of a 'pixel-wise' contrastive loss for fine-grained representations. The paper does not specify how the ground-truth label is mapped to the 2×2 feature grid, nor how the positive set Ω+_i is constructed across the S×S positions and across the two random augmentations. Please clarify the actual feature resolution, the label downsampling procedure, and the positive-pair construction; if S=2, justify how such a coarse spatial loss provides fine-grained supervision.
  4. [Section 4.4, baseline comparisons] The paper asserts that 'all experiments across different methods utilize the same dataset settings and partitions,' but it does not state whether the baseline results in Tables 2–6 are re-implemented by the authors or cited from the original papers, and no code is provided. The STARE/CHASE_DB1 split (first 10/14 images for training) is nonstandard, and the private dataset is not publicly available. Without experimental logs or code, the reader cannot verify that the small F1/MIoU gains are not artifacts of differing preprocessing, split, or training protocols. Please specify the provenance of each baseline number and release code or detailed configuration files for all experiments.
minor comments (6)
  1. [Table 11] The 3rd-level-feature row for STARE reports MIoU=88.25% and AUC=88.25%, values that are mathematically incompatible with F1=81.04% under either MIoU definition and that deviate sharply from the other rows; these appear to be data-entry errors and should be corrected.
  2. [Eq. (11)] The LPCL loss is not fully specified: the outer factor appears to be printed as '2N' but is likely intended to be '1/(2N)', and the construction of Ω+_i relative to the two augmentations X~2i and X~2i−1 is not defined explicitly.
  3. [Section 3.4] The sentence 'For every two input images, we can form S² pixel-level contrastive pairs' does not match the loss formula, which sums over anchors s and positives j; the relationship between the number of pairs and the anchor/positive sets should be clarified.
  4. [Figures 2 and 3] The two versions of Figure 3 use inconsistent loss labels ('Dice + BCE Loss' versus 'Dice + Jaccard + BCE Loss'), and Figure 3(c) contains a subscript typo in the activation notation (ReLu(σ1) versus ReLU(σ2)).
  5. [Section 4.2, Eq. (15)] The Hausdorff distance formula has an unbalanced parenthesis in the quantile expressions, making the definition ambiguous.
  6. [Section 4.1, private dataset] The private dataset is a core part of the five-dataset SOTA claim but is not publicly available; please add a supplementary data sheet or an explicit availability statement describing the acquisition, annotation, and ethical approval details.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: AttUKAN and LPCL are trained end-to-end and evaluated on test data; the MIoU formula inconsistency is an internal metric-definition bug, not a circular reduction.

full rationale

The paper's central claims are empirical evaluations of a trained segmentation network. The architecture (Attention Gates plus tokenized KAN blocks) and the Label-guided Pixel-wise Contrastive Loss are defined in Section 3 from external building blocks (KAN, UNet, additive attention) and are optimized end-to-end on labeled fundus images; there is no fitted constant that is subsequently renamed as a prediction. The loss weights lambda_1..lambda_3 are taken from prior work and lambda_4 is chosen by ablation (Tables 8-10); this is standard hyperparameter selection, and although per-dataset selection from an ablation sweep is a possible overfitting risk, it does not make the reported F1/MIoU numbers equivalent to the loss definition by construction. The references to KAN [16] and UKAN [17] are independent prior work, not self-citations, and no uniqueness theorem is invoked. The only notable defect is formal: Eq. (20) defines MIoU = 2TP/(TP+FP+FN), which would always exceed F1 = 2TP/(2TP+FP+FN), while all tables report F1 > MIoU; the reported values match the standard IoU definition. This is an internal metric inconsistency that undermines reproducibility of the MIoU claim, but it is not a circular derivation: the tabulated numbers remain measurements, not consequences of the definition. Therefore no circularity is found.

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

The central claim rests on conventional deep-learning assumptions: fixed data splits, identical training across methods, and standard hyperparameters. No new physical or mathematical entities are introduced. The main unverified assumption is that all 12 methods were evaluated fairly under identical protocols, which is the standard risk in benchmark papers that do not ship code.

free parameters (3)
  • lambda_4 (LPCL weight) = 0.3 for DRIVE/STARE/CHASE_DB1; 0.5 for HRF/private (Table 10)
    Chosen by per-dataset ablation to maximize F1/MIoU; this is a fitted hyperparameter, not a prediction.
  • lambda_1, lambda_2, lambda_3 loss weights = 0.8, 0.2, 1.0
    Set 'according to the previous experience' [36]; a hand-chosen hyperparameter combination.
  • patch size, batch size, learning rate, epochs = 64x64, 25, 0.003, 100
    Training hyperparameters chosen without a reported search; they influence results but are standard.
assumptions (3)
  • domain assumption STARE and CHASE_DB1 are split first-10/first-14 for training per DUNet [7]; HRF 30/15; private 90/25.
    The comparison validity depends on these splits being the same across all methods; stated in Section 4.1 but not independently verifiable without code.
  • domain assumption The 11 baseline methods were re-implemented or run with the same preprocessing pipeline.
    Section 4.4 claims same dataset settings and partitions, but the paper does not state whether the baseline numbers come from original papers, author code, or the authors' reimplementation.
  • domain assumption Augmented views for LPCL (tilde X_2i and tilde X_2i-1) both share the same ground truth label map.
    Required for the label-guided contrastive loss to treat same-class pairs as positives; asserted implicitly in Section 3.4 but not tested or discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Novel Extraction of Discriminative Fine-Grained Feature to Improve Retinal Vessel Segmentation." pith.science (2026). https://pith.science/paper/NLLBUCNZ

@misc{pith2026250503896,
  author       = {Pith},
  title        = {Pith review of: Novel Extraction of Discriminative Fine-Grained Feature to Improve Retinal Vessel Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NLLBUCNZ}},
  note         = {Machine review of arXiv:2505.03896}
}
read the original abstract

Retinal vessel segmentation is a vital early detection method for several severe ocular diseases. Despite significant progress in retinal vessel segmentation with the advancement of Neural Networks, there are still challenges to overcome. Specifically, retinal vessel segmentation aims to predict the class label for every pixel within a fundus image, with a primary focus on intra-image discrimination, making it vital for models to extract more discriminative features. Nevertheless, existing methods primarily focus on minimizing the difference between the output from the decoder and the label, but ignore fully using feature-level fine-grained representations from the encoder. To address these issues, we propose a novel Attention U-shaped Kolmogorov-Arnold Network named AttUKAN along with a novel Label-guided Pixel-wise Contrastive Loss for retinal vessel segmentation. Specifically, we implement Attention Gates into Kolmogorov-Arnold Networks to enhance model sensitivity by suppressing irrelevant feature activations and model interpretability by non-linear modeling of KAN blocks. Additionally, we also design a novel Label-guided Pixel-wise Contrastive Loss to supervise our proposed AttUKAN to extract more discriminative features by distinguishing between foreground vessel-pixel pairs and background pairs. Experiments are conducted across four public datasets including DRIVE, STARE, CHASE_DB1, HRF and our private dataset. AttUKAN achieves F1 scores of 82.50%, 81.14%, 81.34%, 80.21% and 80.09%, along with MIoU scores of 70.24%, 68.64%, 68.59%, 67.21% and 66.94% in the above datasets, which are the highest compared to 11 networks for retinal vessel segmentation. Quantitative and qualitative results show that our AttUKAN achieves state-of-the-art performance and outperforms existing retinal vessel segmentation methods. Our code will be available at https://github.com/stevezs315/AttUKAN.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 34 canonical work pages

  1. [1]

    Thomas J Smart, Christopher J Richards, Rhythm Bhat- nagar, Carlos Pavesio, Rupesh Agrawal, Philip H Jones, in Optical trapping and optical micromanipulation XII, SPIE,2015, pp. 342–348

    work page 2015

  2. [2]

    Jie Ding, Khin Lay Wai, Kevin McGeechan, M Kam- ran Ikram, Ryo Kawasaki, Jing Xie, Ronald Klein, Bar- baraBKKlein,MaryFrancesCotch,JieJinWang,etal., Journal of hypertension2014, 32 (2), 207–215

    work page

  3. [3]

    Tong Na, Jianyang Xie, Yitian Zhao, Yifan Zhao, Yue Liu, Yongtian Wang, Jiang Liu,Medical physics2018, 45 (7), 3132–3146

    work page

  4. [4]

    Zhe Han, Yilong Yin, Xianjing Meng, Gongping Yang, XiaoweiYan,in 2014IEEEInternationalConferenceon Data Mining Workshop, IEEE,2014, pp. 960–967

    work page 2014

  5. [5]

    3431–3440

    Jonathan Long, Evan Shelhamer, Trevor Darrell, inPro- ceedingsoftheIEEEconferenceoncomputervisionand pattern recognition, 2015, pp. 3431–3440

    work page 2015

  6. [6]

    Olaf Ronneberger, Philipp Fischer, Thomas Brox, in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international confer- ence, Munich, Germany, October 5-9, 2015, proceed- ings, part III 18, Springer,2015, pp. 234–241. 14 Shuang ZengET AL

    work page 2015

  7. [7]

    Qiangguo Jin, Zhaopeng Meng, Tuan D Pham, Qi Chen, LeyiWei,RanSu, Knowledge-BasedSystems 2019,178, 149–162

    work page 2019

  8. [8]

    6070– 6079

    Yaolei Qi, Yuting He, Xiaoming Qi, Yuan Zhang, Guanyu Yang, inProceedings of the IEEE/CVF Interna- tional Conference on Computer Vision, 2023, pp. 6070– 6079

    work page 2023

Show all 42 references
  1. [9]

    3656–3665

    Liangzhi Li, Manisha Verma, Yuta Nakashima, Hajime Nagahara, Ryo Kawasaki, in Proceedings of the IEEE/CVF winter conference on applications of computer vision, 2020, pp. 3656–3665

    2020

  2. [10]

    1237– 1241

    KunWang,XiaohongZhang,ShengHuang,QiuliWang, FeiyuChen,in 2020IEEE17thInternationalSymposium on Biomedical Imaging (ISBI), IEEE,2020, pp. 1237– 1241

    2020

  3. [11]

    Reza Azad, Maryam Asadi-Aghbolaghi, Mahmood Fathy, Sergio Escalera, inProceedings of the IEEE/CVF international conference on computer vision workshops, 2019, pp. 0–0

    2019

  4. [12]

    Ozan Oktay, Jo Schlemper, Loic Le Folgoc, Matthew Lee,MattiasHeinrich,KazunariMisawa,KensakuMori, Steven McDonagh, Nils Y Hammerla, Bernhard Kainz, et al.,arXiv preprint arXiv:1804.039992018

    arXiv

  5. [13]

    Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, Jianming Liang, inDeep Learning in Medi- calImageAnalysisandMultimodalLearningforClinical Decision Support: 4th International Workshop, DLMIA 2018, and 8th International Workshop, ML-CDS 2018, Held in Conjunction with...

    2018

  6. [14]

    3819–3827

    Yutong Liu, Haijiang Zhu, Mengting Liu, Huaiyuan Yu, ZihanChen,JieGao,in ProceedingsoftheAAAIConfer- ence on Artificial Intelligence, 2024, pp. 3819–3827

    2024

  7. [15]

    Ziyang Wang, Jian-Qing Zheng, Yichi Zhang, Ge Cui, Lei Li,arXiv preprint arXiv:2402.050792024

    arXiv

  8. [16]

    Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Soljačić, Thomas Y Hou, Max Tegmark,arXiv preprint arXiv:2404.19756 2024

    2024 arXiv

  9. [17]

    Chenxin Li, Xinyu Liu, Wuyang Li, Cheng Wang, Hengyu Liu, Yixuan Yuan, arXiv preprint arXiv:2406.029182024

    arXiv

  10. [18]

    1597–1607

    Ting Chen, Simon Kornblith, Mohammad Norouzi, GeoffreyHinton,in Internationalconferenceonmachine learning, PMLR,2020, pp. 1597–1607

    2020

  11. [19]

    9729–9738

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, Ross Girshick, in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 9729–9738

    2020

  12. [20]

    1166–1170

    Nagendra Pratap Singh, Rajesh Kumar, Rajeev Srivas- tava, inInternational Conference on Computing, Com- munication & Automation, IEEE,2015, pp. 1166–1170

    2015

  13. [21]

    Frederic Zana, J-C Klein,IEEE transactions on image processing2001, 10 (7), 1010–1019

  14. [22]

    Wendeson S Oliveira, Joyce Vitor Teixeira, Tsang Ing Ren, George DC Cavalcanti, Jan Sijbers,PloS one2016, 11 (2), e0149943

  15. [23]

    Huazhu Fu, Yanwu Xu, Stephen Lin, Damon Wing Kee Wong, Jiang Liu, in Medical Image Comput- ing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, October 17-21,2016,Proceedings,PartII19 ,Springer, 2016,pp. 132–139

    2016

  16. [24]

    Yanfei Guo, Yanjun Peng,BMC medical imaging2020, 20, 1–22

  17. [25]

    Seung Yeon Shin, Soochahn Lee, Il Dong Yun, Kyoung Mu Lee, Medical image analysis 2019, 58, 101556

    2019

  18. [26]

    Yinda Chen, Haoyuan Shi, Xiaoyu Liu, Te Shi, Ruob- ing Zhang, Dong Liu, Zhiwei Xiong, Feng Wu,arXiv preprint arXiv:2405.168472024

    arXiv

  19. [27]

    AlexeyDosovitskiy,LucasBeyer,AlexanderKolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al.,arXiv preprint arXiv:2010.11929 2020

    2010 arXiv

  20. [28]

    Jinming Cao, Yangyan Li, Mingchao Sun, Ying Chen, Dani Lischinski, Daniel Cohen-Or, Baoquan Chen, Changhe Tu, IEEE Transactions on Image Processing 2022, 31, 3726–3736

    2022

  21. [29]

    Jimmy Lei Ba, Jamie Ryan Kiros, Geoffrey E Hinton, arXiv preprint arXiv:1607.064502016

    arXiv

  22. [30]

    Dzmitry Bahdanau, Kyunghyun Cho, Yoshua Bengio, arXiv preprint arXiv:1409.04732014

  23. [31]

    Joes Staal, Michael D Abràmoff, Meindert Niemeijer, Max A Viergever, Bram Van Ginneken,IEEE transac- tions on medical imaging2004, 23 (4), 501–509

  24. [32]

    Shuang ZengET AL 15

    ADHoover,ValentinaKouznetsova,MichaelGoldbaum, IEEE Transactions on Medical imaging2000, 19 (3), 203–210. Shuang ZengET AL 15

  25. [33]

    ChristopherGOwen,AlicjaRRudnicka,RobertMullen, SarahABarman,DorothyMonekosso,PeterHWhincup, JeffreyNg,CarlPaterson, Investigativeophthalmology& visual science2009, 50 (5), 2004–2010

    2004

  26. [34]

    Jan Odstrcilik, Radim Kolar, Attila Budai, Joachim Hornegger, Jiri Jan, Jiri Gazarek, Tomas Kubena, Pavel Cernosek, Ondrej Svoboda, Elli Angelopoulou, IET Image Processing2013, 7 (4), 373–383

  27. [35]

    Manuel Emilio Gegúndez-Arias, Arturo Aquino, José Manuel Bravo, Diego Marín,IEEE transactions on medical imaging2011, 31 (2), 231–239

  28. [36]

    Bo Wang, Shuang Qiu, Huiguang He, in Medical Image Computing and Computer Assisted Intervention– MICCAI 2019: 22nd International Conference, Shen- zhen, China, October 13–17, 2019, Proceedings, Part I 22, Springer,2019, pp. 84–92

    2019

  29. [37]

    6077–6086

    Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney,MarkJohnson,StephenGould,LeiZhang,in Pro- ceedingsoftheIEEEconferenceoncomputervisionand pattern recognition, 2018, pp. 6077–6086

    2018

  30. [38]

    Saumya Jetley, Nicholas A Lord, Namhoon Lee, Philip HS Torr,arXiv preprint arXiv:1804.023912018

    arXiv

  31. [39]

    7132–7141

    Jie Hu, Li Shen, Gang Sun, inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132–7141

    2018

  32. [40]

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, In So Kweon, inProceedings of the European conference on computer vision (ECCV), 2018, pp. 3–19

    2018

  33. [41]

    João VB Soares, Jorge JG Leandro, Roberto M Cesar, HerbertFJelinek,MichaelJCree, IEEETransactionson medical Imaging2006, 25 (9), 1214–1222

  34. [42]

    Stephen M Pizer, E Philip Amburn, John D Austin, Robert Cromartie, Ari Geselowitz, Trey Greer, Bart ter Haar Romeny, John B Zimmerman, Karel Zuiderveld, Computer vision, graphics, and image processing1987, 39 (3), 355–368. 16 Shuang ZengET AL RollingUNetUNet UNet++ UKAN DSCNet...

Pith tools

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