Pith. sign in

REVIEW 3 major objections 5 minor 57 references

Neovascularization Segmentation via a Multilateral Interaction-Enhanced Graph Convolutional Network

T0 review · 3 major / 5 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 The paper claims that joint graph-based reasoning over CNV region and vessel morphology yields state-of-the-art segmentation, with Dice scores of 87.21% and 88.12% on the first public CNV dataset.

desk verdict A useful dataset wrapped in a conventional graph-reasoning network; the segmentation claims are promises until the full paper reveals the protocol. read the letter →

arxiv 2508.03197 v1 pith:FTUV2F43 submitted 2025-08-05 cs.CV

Tao Chen , Dan Zhang , Da Chen , Huazhu Fu , Kai Jin , Shanshan Wang , Laurent D. Cohen , Yitian Zhao
show 2 more authors
Quanyong Yi Jiong Zhang
This is my paper · ORCID
classification cs.CV
keywords choroidalneovascularizationsegmentationOCTAgraphconvolutionalnetworkmulti-tasklearningregionandvesseluncertainty-weightedlossCNVSegdatasetwetage-relatedmaculardegeneration
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

This paper is trying to establish that choroidal neovascularization (CNV) in OCTA images can be segmented accurately when region and vessel morphology are modeled together, and that graph-based cross-task reasoning is the mechanism that makes the joint modeling work. To support that claim, the authors assemble and release CNVSeg, described as the first publicly accessible CNV segmentation dataset, and introduce MTG-Net, a multi-task network with two graph reasoning modules. Reported results place MTG-Net above prior methods, with Dice scores of 87.21% for regions and 88.12% for vessels. A sympathetic reader would care because wet age-related macular degeneration is a leading cause of blindness, and reliable automatic measurement of CNV burden is a step toward objective clinical assessment.

What carries the argument

The machinery is a multi-task encoder-decoder that decouples each OCTA image into three task-specific feature maps, plus two graph-based cross-task modules: Multilateral Interaction Graph Reasoning (MIGR) and Multilateral Reinforcement Graph Reasoning (MRGR). These modules build a learned graph over the task-specific representations and iteratively propagate higher-order relationships between region, vessel, and shape cues, so that each task's objective is refined by information from the others. An uncertainty-weighted loss then down-weights loss contributions at pixels where the network is less confident, reducing the influence of projection artifacts, noise, and boundary blur on the final segmentation.

What would settle it

Re-annotate a held-out subset of CNVSeg with a second independent set of expert annotators and recompute MTG-Net's Dice scores against those labels on the same test split; if the scores fall materially below the reported 87.21% and 88.12%, the headline result is partly an artifact of the original annotation protocol.

Watch

Extended reading notes

Core claim

The central claim is that CNV region segmentation and CNV vessel segmentation are mutually reinforcing tasks, and that a network which explicitly reasons about their relationship in a graph domain can outperform single-task and simpler multi-task baselines. On the newly released CNVSeg dataset, MTG-Net achieves a Dice score of 87.21% for region segmentation and 88.12% for vessel segmentation, which the authors report as the best among compared methods. If the claim holds, the discovery is that cross-task graph reasoning over region, vessel, and shape-related feature maps converts the geometric duality between a lesion and its vasculature into concrete segmentation gains, while an uncertainty-weighted loss absorbs the artifacts and noise typical of OCTA imaging.

Load-bearing premise

The whole evaluation rests on CNVSeg's manual annotations being accurate and consistently defined; if the ground-truth labels are noisy, biased, or inconsistent, the reported Dice scores do not measure true segmentation quality and the comparison against baselines is not meaningful.

Editorial extensions

If this is right

  • A single network can output both CNV region masks and vessel masks in one forward pass, so a clinical workflow would obtain both at once.
  • The uncertainty-weighted loss is designed to suppress projection artifacts and noise, which should make segmentation more reliable on lower-quality OCTA scans.
  • With CNVSeg released, future CNV segmentation methods can be compared on the same public data instead of private datasets.
  • If the reported margin over baselines is caused by the graph-based cross-task interaction, similar joint reasoning could improve segmentation of other paired lesion-and-vessel structures.

Reading between the lines

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

  • A direct test this paper leaves implicit is whether the same dual-task graph mechanism transfers to other vascular lesions, such as diabetic retinopathy neovascularization, where a lesion region and its feeding vessels have similar geometric duality.
  • The architecture bundles several contributions, including multi-task decoding, graph interaction, and uncertainty weighting, so the reported margin over baselines does not by itself show which component carries the gain; component-wise ablations would separate them.
  • If CNVSeg becomes widely adopted, the main lasting effect may be the public benchmark itself, since OCTA CNV data were previously scarce and comparisons had to be made on private datasets.
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

3 major / 5 minor

Summary. The manuscript proposes MTG-Net, a multi-task graph-based convolutional network for segmenting choroidal neovascularization (CNV) regions and vessels in OCTA images, together with a new dataset named CNVSeg. The abstract claims state-of-the-art performance with Dice scores of 87.21% for region segmentation and 88.12% for vessel segmentation. However, the text provided for review contains only the abstract, the opening of the introduction, and the reference list; no method formulation, dataset description, experimental protocol, or results are present. Consequently, the central claims of architectural novelty, dataset contribution, and state-of-the-art performance cannot be assessed from the submitted material.

Significance. If fully substantiated, the work would have practical value: a publicly accessible CNV OCTA dataset would fill a recognized gap, and a graph-based multi-task framework exploiting region-vessel duality is a plausible direction for improving segmentation performance. The idea of coupling uncertainty-weighted loss with graph reasoning is also potentially useful. However, the visible manuscript provides no experimental evidence, no dataset release details, and no quantitative comparison that can be checked. The significance as a contribution therefore rests entirely on material that is absent from the submitted text.

major comments (3)
  1. [Abstract / Section 1] The central performance claim is unverifiable from the submitted text. No experimental section is present: there is no dataset size or train/validation/test split, no preprocessing or augmentation details, no list of compared baselines, no definition of the Dice metric, no indication of the number of runs, and no statistical significance tests. The headline Dice scores of 87.21% and 88.12% are stated without any supporting protocol. The authors must provide a complete experimental section, including standard deviations and significance testing, before the state-of-the-art claim can be evaluated.
  2. [Abstract / Section 1] The claim that CNVSeg is the first publicly accessible CNV dataset is unsupported in the visible text. No download link, number of images, patient demographics, acquisition device, scanning protocol, or annotation procedure is given. There is also no description of how many annotators produced the ground truth or what inter-observer agreement was obtained. Without this information, the validity of the ground-truth labels, and therefore the meaningfulness of all reported Dice scores, cannot be assessed. This information is load-bearing for the dataset-contribution claim and must be supplied.
  3. [Section 1 / Abstract] The technical method is not described in the submitted text. MIGR, MRGR, the multi-task framework, and the uncertainty-weighted loss are only named; no equations, graph construction rules, adjacency definitions, or loss formulations appear anywhere in the visible pages. As a result, the claimed novelty of the graph interaction mechanism cannot be evaluated. The full method section, including all mathematical definitions, must be present in the submitted manuscript for review.
minor comments (5)
  1. [Abstract] There is a typo in the abstract: 'Dice socre' should be 'Dice score'.
  2. [Section 1] The citation style '[4] [5] [6]' should be consolidated as '[4]–[6]' for readability.
  3. [Abstract / Section 1] The phrasing 'challenges existed' should be 'challenges exist', and 'constraints' should be 'constrains' for grammatical correctness.
  4. [Figure 1] Figure 1 is referenced in the text but is not included in the submitted material; if this is a submission issue, the figure must be embedded in the full PDF.
  5. [Section 1] The phrase 'region and vessel morphological information' is vague; the authors should define precisely what the region task and the vessel task predict and how they relate to the clinical annotation of CNV.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; the visible text offers an empirical benchmark claim, not a derivation that reduces to its own inputs.

full rationale

The visible manuscript contains only the abstract, the first page of the introduction, and the reference list. No method equations, training procedure, dataset split, or evaluation protocol are shown in the provided text. The headline claim that 'MTG-Net outperforms existing methods, achieving a Dice socre of 87.21% for region segmentation and 88.12% for vessel segmentation' is an empirical result on a held-out benchmark, not a quantity derived by construction from the network definition. There is no fitted parameter that is renamed as a prediction, no equation in which an output is defined in terms of the target metric, and no load-bearing self-citation chain: the authors' own prior works appear only as ordinary background references (e.g., [6], [13]) and are not invoked to justify a uniqueness theorem or to define the architecture. The absence of experimental details, baseline descriptions, and annotation protocol is a legitimate reproducibility and correctness concern, but it is not circularity under the criteria of this review. Without visible equations or a self-referential argument, there is no specific reduction to exhibit, so the circularity burden is not met.

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

No free parameters or invented physical entities are identifiable from the abstract and first page. The network's learned weights and hyperparameters are not listed, and the paper introduces no new physical objects.

assumptions (3)
  • domain assumption OCTA en-face images contain sufficient signal to distinguish CNV regions and vessels from background and artifacts.
    The abstract claims that accurate segmentation is possible from OCTA images; if image quality or artifacts prevent reliable labeling, the central result collapses.
  • domain assumption The CNVSeg ground-truth labels are correct and consistent.
    The reported Dice scores are computed against these labels; no annotation protocol or inter-rater reliability is provided in the visible text.
  • domain assumption The train/test split is unbiased and the comparison protocol is standard.
    The performance claims depend on fair evaluation; without the experimental section, this cannot be verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neovascularization Segmentation via a Multilateral Interaction-Enhanced Graph Convolutional Network." pith.science (2026). https://pith.science/paper/FTUV2F43

@misc{pith2026250803197,
  author       = {Pith},
  title        = {Pith review of: Neovascularization Segmentation via a Multilateral Interaction-Enhanced Graph Convolutional Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FTUV2F43}},
  note         = {Machine review of arXiv:2508.03197}
}
read the original abstract

Choroidal neovascularization (CNV), a primary characteristic of wet age-related macular degeneration (wet AMD), represents a leading cause of blindness worldwide. In clinical practice, optical coherence tomography angiography (OCTA) is commonly used for studying CNV-related pathological changes, due to its micron-level resolution and non-invasive nature. Thus, accurate segmentation of CNV regions and vessels in OCTA images is crucial for clinical assessment of wet AMD. However, challenges existed due to irregular CNV shapes and imaging limitations like projection artifacts, noises and boundary blurring. Moreover, the lack of publicly available datasets constraints the CNV analysis. To address these challenges, this paper constructs the first publicly accessible CNV dataset (CNVSeg), and proposes a novel multilateral graph convolutional interaction-enhanced CNV segmentation network (MTG-Net). This network integrates both region and vessel morphological information, exploring semantic and geometric duality constraints within the graph domain. Specifically, MTG-Net consists of a multi-task framework and two graph-based cross-task modules: Multilateral Interaction Graph Reasoning (MIGR) and Multilateral Reinforcement Graph Reasoning (MRGR). The multi-task framework encodes rich geometric features of lesion shapes and surfaces, decoupling the image into three task-specific feature maps. MIGR and MRGR iteratively reason about higher-order relationships across tasks through a graph mechanism, enabling complementary optimization for task-specific objectives. Additionally, an uncertainty-weighted loss is proposed to mitigate the impact of artifacts and noise on segmentation accuracy. Experimental results demonstrate that MTG-Net outperforms existing methods, achieving a Dice socre of 87.21\% for region segmentation and 88.12\% for vessel segmentation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 51 canonical work pages

  1. [2]

    Choroidal neovas- cularization,

    Hans E Grossniklaus and W Richard Green, “Choroidal neovas- cularization,”American journal of ophthalmology, vol. 137, no. 3, pp. 496–503, 2004

    work page 2004

  2. [3]

    Adverse reaction in patients with drug allergy history after simultaneous intravenous fundus fluorescein angiography and indocyanine green angiography,

    Zhaoan Su, Panpan Ye, Yan Teng, Lixia Zhang, and Xiaohong Shu, “Adverse reaction in patients with drug allergy history after simultaneous intravenous fundus fluorescein angiography and indocyanine green angiography,”Journal of ocular pharmacology and therapeutics, vol. 28, no. 4, pp. 410–413, 2012

    work page 2012

  3. [4]

    A review of optical coherence tomography angiography (octa),

    Talisa E De Carlo, Andre Romano, Nadia K Waheed, and Jay S Duker, “A review of optical coherence tomography angiography (octa),”International journal of retina and vitreous, vol. 1, pp. 1–15, 2015

    work page 2015

  4. [5]

    Optical coherence tomography angiography in retinal diseases,

    KV Chalam and Kumar Sambhav, “Optical coherence tomography angiography in retinal diseases,”Journal of ophthalmic & vision research, vol. 11, no. 1, pp. 84, 2016

    work page 2016

  5. [6]

    3d shape modeling and analysis of retinal microvasculature in oct- angiography images,

    Jiong Zhang, Yuchuan Qiao, Mona Sharifi Sarabi, Maziyar M Khansari, Jin K Gahm, Amir H Kashani, and Yonggang Shi, “3d shape modeling and analysis of retinal microvasculature in oct- angiography images,”IEEE Transactions on Medical Imaging, vol. 39, no. 5, pp. 1335–1346, 2019

    work page 2019

  6. [7]

    Rethinking dual-stream super-resolution semantic learning in medical image segmentation,

    Zhongxi Qiu, Yan Hu, Xiaoshan Chen, Dan Zeng, Qingyong Hu, and Jiang Liu, “Rethinking dual-stream super-resolution semantic learning in medical image segmentation,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    work page 2023

  7. [8]

    Querying labeled for unlabeled: Cross-image semantic consistency guided semi-supervised semantic segmenta- tion,

    Linshan Wu, Leyuan Fang, Xingxin He, Min He, Jiayi Ma, and Zhun Zhong, “Querying labeled for unlabeled: Cross-image semantic consistency guided semi-supervised semantic segmenta- tion,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 7, pp. 8827–8844, 2023

    work page 2023

  8. [9]

    Automatic quantification of choroidal neovascular- ization lesion area on oct angiography based on density cell-like p systems with active membranes,

    Jie Xue, Acner Camino, Steven T Bailey, Xiyu Liu, Dengwang Li, and Yali Jia, “Automatic quantification of choroidal neovascular- ization lesion area on oct angiography based on density cell-like p systems with active membranes,”Biomedical Optics Express, vol. 9, no. 7, pp. 3208–3219, 2018

    work page 2018

Show all 57 references
  1. [10]

    Wei Feng, Meihan Duan, Bingjie Wang, Yu Du, Yiran Zhao, Bin Wang, Lin Zhao, Zongyuan Ge, and Yuntao Hu, “Automated segmentation of choroidal neovascularization on optical coherence tomography angiography images of neovascular age-related mac- ular degeneration patients based o...

    2023

  2. [11]

    Automated choroidal neovascularization detec- tion algorithm for optical coherence tomography angiography,

    Li Liu, Simon S Gao, Steven T Bailey, David Huang, Dengwang Li, and Yali Jia, “Automated choroidal neovascularization detec- tion algorithm for optical coherence tomography angiography,” Biomedical Optics Express, vol. 6, no. 9, pp. 3564–3576, 2015

    2015

  3. [12]

    Automated diagnosis and segmentation of choroidal neovascularization in oct angiography using deep learning,

    Jie Wang, Tristan T Hormel, Liqin Gao, Pengxiao Zang, Yukun Guo, Xiaogang Wang, Steven T Bailey, and Yali Jia, “Automated diagnosis and segmentation of choroidal neovascularization in oct angiography using deep learning,”Biomedical Optics Express, vol. 11, no. 2, pp. 927–944, 2020

    2020

  4. [13]

    Rbgnet: Reliable boundary- guided segmentation of choroidal neovascularization,

    Tao Chen, Yitian Zhao, Lei Mou, Dan Zhang, Xiayu Xu, Mengting Liu, Huazhu Fu, and Jiong Zhang, “Rbgnet: Reliable boundary- guided segmentation of choroidal neovascularization,” inInterna- tional Conference on Medical Image Computing and Computer-Assisted Intervention. Springer...

    2023

  5. [14]

    Cnv- net: Segmentation, classification and activity score measurement of choroidal neovascularization (cnv) using optical coherence to- mography angiography (octa),

    Mahsa Vali, Behzad Nazari, Saeed Sadri, Elias Khalili Pour, Hamid Riazi-Esfahani, Hooshang Faghihi, Nazanin Ebrahimiadib, Momeneh Azizkhani, Will Innes, David H Steel, et al., “Cnv- net: Segmentation, classification and activity score measurement of choroidal neovascularizatio...

    2023

  6. [15]

    have been developed for CNV segmentation, offering arXiv:2508.03197v1 [cs.CV] 5 Aug 2025 15

    2025 arXiv

  7. [16]

    Deep learning for diagnosing and segmenting choroidal neovascularization in oct angiography in a large real-world data set,

    Jie Wang, Tristan T Hormel, Kotaro Tsuboi, Xiaogang Wang, Xi- aoyan Ding, Xiaoyan Peng, David Huang, Steven T Bailey, and Yali Jia, “Deep learning for diagnosing and segmenting choroidal neovascularization in oct angiography in a large real-world data set,”Translational Vision...

    2023

  8. [17]

    Interactive medical image seg- mentation using deep learning with image-specific fine tuning,

    Guotai Wang, Wenqi Li, Maria A Zuluaga, Rosalind Pratt, Pre- mal A Patel, Michael Aertsen, Tom Doel, Anna L David, Jan Deprest, S´ebastien Ourselin, et al., “Interactive medical image seg- mentation using deep learning with image-specific fine tuning,” IEEE transactions on med...

    2018

  9. [18]

    Encoder-decoder with atrous sep- arable convolution for semantic image segmentation,

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam, “Encoder-decoder with atrous sep- arable convolution for semantic image segmentation,” inProceed- ings of the European conference on computer vision (ECCV), 2018, pp. 801–818

    2018

  10. [19]

    Attention u- net: Learning where to look for the pancreas,

    Ozan Oktay, Jo Schlemper, Loic Le Folgoc, Matthew Lee, Mat- tias Heinrich, Kazunari Misawa, Kensaku Mori, Steven McDon- agh, Nils Y Hammerla, Bernhard Kainz, et al., “Attention u- net: Learning where to look for the pancreas,”arXiv preprint arXiv:1804.03999, 2018

    2018 arXiv

  11. [20]

    Ia-net: informative attention convolu- tional neural network for choroidal neovascularization segmenta- tion in oct images,

    Xiaoming Xi, Xianjing Meng, Zheyun Qin, Xiushan Nie, Yilong Yin, and Xinjian Chen, “Ia-net: informative attention convolu- tional neural network for choroidal neovascularization segmenta- tion in oct images,”Biomedical Optics Express, vol. 11, no. 11, pp. 6122–6136, 2020

    2020

  12. [21]

    Mf-net: Multi-scale information fusion network for cnv segmentation in retinal OCT images,

    Qingquan Meng, Lianyu Wang, Tingting Wang, Meng Wang, Weifang Zhu, Fei Shi, Zhongyue Chen, and Xinjian Chen, “Mf-net: Multi-scale information fusion network for cnv segmentation in retinal OCT images,”Frontiers in Neuroscience, vol. 15, pp. 743769, 2021

    2021

  13. [22]

    Mpb-cnn: a multi-scale parallel branch cnn for choroidal neovascularization segmentation in sd- oct images,

    Yuhan Zhang, Zexuan Ji, Yuexuan Wang, Sijie Niu, Wen Fan, Songtao Yuan, and Qiang Chen, “Mpb-cnn: a multi-scale parallel branch cnn for choroidal neovascularization segmentation in sd- oct images,”OSA Continuum, vol. 2, no. 3, pp. 1011–1027, 2019

    2019

  14. [23]

    Au- tomated segmentation of choroidal neovascularization in optical coherence tomography images using multi-scale convolutional neural networks with structure prior,

    Xiaoming Xi, Xianjing Meng, Lu Yang, Xiushan Nie, Gongping Yang, Haoyu Chen, Xin Fan, Yilong Yin, and Xinjian Chen, “Au- tomated segmentation of choroidal neovascularization in optical coherence tomography images using multi-scale convolutional neural networks with structure p...

    2019

  15. [24]

    A comprehensive survey on graph neural networks,

    Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S Yu Philip, “A comprehensive survey on graph neural networks,”IEEE transactions on neural networks and learning systems, vol. 32, no. 1, pp. 4–24, 2020

    2020

  16. [25]

    Semi-supervised classi- fication with graph convolutional networks,

    Thomas N Kipf and Max Welling, “Semi-supervised classi- fication with graph convolutional networks,”arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  17. [26]

    Cnn- enhanced graph convolutional network with pixel-and superpixel- level feature fusion for hyperspectral image classification,

    Qichao Liu, Liang Xiao, Jingxiang Yang, and Zhihui Wei, “Cnn- enhanced graph convolutional network with pixel-and superpixel- level feature fusion for hyperspectral image classification,”IEEE Transactions on Geoscience and Remote Sensing, vol. 59, no. 10, pp. 8657–8671, 2020

    2020

  18. [27]

    Class-wise dynamic graph convolution for semantic seg- mentation,

    Hanzhe Hu, Deyi Ji, Weihao Gan, Shuai Bai, Wei Wu, and Junjie Yan, “Class-wise dynamic graph convolution for semantic seg- mentation,” inEuropean Conference on Computer Vision. Springer, 2020, pp. 1–17

    2020

  19. [28]

    Uncertainty-based graph convolutional networks for organ seg- mentation refinement,

    Roger D Soberanis-Mukul, Nassir Navab, and Shadi Albarqouni, “Uncertainty-based graph convolutional networks for organ seg- mentation refinement,” inMedical Imaging with Deep Learning. PMLR, 2020, pp. 755–769

    2020

  20. [29]

    Attention graph convolution network for image segmentation in big sar imagery data,

    Fei Ma, Fei Gao, Jinping Sun, Huiyu Zhou, and Amir Hussain, “Attention graph convolution network for image segmentation in big sar imagery data,”Remote Sensing, vol. 11, no. 21, pp. 2586, 2019

    2019

  21. [30]

    Scale-aware graph neural network for few-shot semantic segmentation,

    Guo-Sen Xie, Jie Liu, Huan Xiong, and Ling Shao, “Scale-aware graph neural network for few-shot semantic segmentation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 5475–5484

    2021

  22. [31]

    Beyond grids: Learning graph repre- sentations for visual recognition,

    Yin Li and Abhinav Gupta, “Beyond grids: Learning graph repre- sentations for visual recognition,”Advances in Neural Information Processing Systems, vol. 31, 2018

    2018

  23. [32]

    Graph-based global reasoning networks,

    Yunpeng Chen, Marcus Rohrbach, Zhicheng Yan, Yan Shuicheng, Jiashi Feng, and Yannis Kalantidis, “Graph-based global reasoning networks,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 433–442

    2019

  24. [33]

    Referring image segmentation via cross-modal progressive comprehension,

    Shaofei Huang, Tianrui Hui, Si Liu, Guanbin Li, Yunchao Wei, Jizhong Han, Luoqi Liu, and Bo Li, “Referring image segmentation via cross-modal progressive comprehension,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 10488–10497

    2020

  25. [34]

    Dual graph convolutional network for semantic segmentation,

    Li Zhang, Xiangtai Li, Anurag Arnab, Kuiyuan Yang, Yunhai Tong, and Philip HS Torr, “Dual graph convolutional network for semantic segmentation,”arXiv preprint arXiv:1909.06121, 2019

    1909 arXiv

  26. [35]

    Deep vessel segmentation by learning graphical connectiv- ity,

    Seung Yeon Shin, Soochahn Lee, Il Dong Yun, and Kyoung Mu Lee, “Deep vessel segmentation by learning graphical connectiv- ity,”Medical image analysis, vol. 58, pp. 101556, 2019

    2019

  27. [36]

    Ginet: Graph interaction network for scene parsing,

    Tianyi Wu, Yu Lu, Yu Zhu, Chuang Zhang, Ming Wu, Zhanyu Ma, and Guodong Guo, “Ginet: Graph interaction network for scene parsing,” inComputer Vision–ECCV 2020: 16th European 16 Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XVII

    2020

  28. [37]

    Springer, 2020, pp. 34–51

    2020

  29. [38]

    Multiple-human parsing in the wild,

    Jianshu Li, Jian Zhao, Yunchao Wei, Congyan Lang, Yidong Li, Terence Sim, Shuicheng Yan, and Jiashi Feng, “Multiple-human parsing in the wild,”arXiv preprint arXiv:1705.07206, 2017

    2017 arXiv

  30. [39]

    Pedestrian at- tribute recognition by joint visual-semantic reasoning and knowl- edge distillation.,

    Qiaozhe Li, Xin Zhao, Ran He, and Kaiqi Huang, “Pedestrian at- tribute recognition by joint visual-semantic reasoning and knowl- edge distillation.,” inIJCAI, 2019, pp. 833–839

    2019

  31. [40]

    Graph-bas3net: Boundary-aware semi-supervised segmentation network with bilateral graph convolution,

    Huimin Huang, Lanfen Lin, Yue Zhang, Yingying Xu, Jing Zheng, XiongWei Mao, Xiaohan Qian, Zhiyi Peng, Jianying Zhou, Yen-Wei Chen, et al., “Graph-bas3net: Boundary-aware semi-supervised segmentation network with bilateral graph convolution,” inPro- ceedings of the IEEE/CVF Int...

    2021

  32. [41]

    U2-net: Going deeper with nested u-structure for salient object detection,

    Xuebin Qin, Zichen Zhang, Chenyang Huang, Masood Dehghan, Osmar R Zaiane, and Martin Jagersand, “U2-net: Going deeper with nested u-structure for salient object detection,”Pattern recognition, vol. 106, pp. 107404, 2020

    2020

  33. [42]

    A computational approach to edge detection,

    John Canny, “A computational approach to edge detection,”IEEE Transactions on Pattern Analysis and Machine Intelligence, , no. 6, pp. 679–698, 1986

    1986

  34. [43]

    Deepsdf: Learning continuous signed distance functions for shape representation,

    Jeong Joon Park, Peter Florence, Julian Straub, Richard New- combe, and Steven Lovegrove, “Deepsdf: Learning continuous signed distance functions for shape representation,” inProc. CVPR, 2019, pp. 165–174

    2019

  35. [44]

    Fronts propagating with curvature- dependent speed: algorithms based on Hamilton-Jacobi formula- tions,

    S. Osher and J. A. Sethian, “Fronts propagating with curvature- dependent speed: algorithms based on Hamilton-Jacobi formula- tions,”Journal of Computational Physics, vol. 79, no. 1, pp. 12–49, 1988

    1988

  36. [45]

    Shape-aware organ segmentation by predicting signed distance maps,

    Yuan Xue, Hui Tang, Zhi Qiao, Guanzhong Gong, Yong Yin, Zhen Qian, Chao Huang, Wei Fan, and Xiaolei Huang, “Shape-aware organ segmentation by predicting signed distance maps,” inProc. AAAI, 2020, vol. 34, pp. 12565–12572

    2020

  37. [46]

    Bayesian convolutional neu- ral networks with bernoulli approximate variational inference,

    Yarin Gal and Zoubin Ghahramani, “Bayesian convolutional neu- ral networks with bernoulli approximate variational inference,” arXiv preprint arXiv:1506.02158, 2015

    2015 arXiv

  38. [47]

    U-net: Convolutional networks for biomedical image segmentation,

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical image computing and computer-assisted intervention–Miccai 2015: 18th international conference, munich, germany, october 5-9, 2015, proceedings, Par...

    2015

  39. [48]

    Ce- net: Context encoder network for 2d medical image segmenta- tion,

    Zaiwang Gu, Jun Cheng, Huazhu Fu, Kang Zhou, Huaying Hao, Yitian Zhao, Tianyang Zhang, Shenghua Gao, and Jiang Liu, “Ce- net: Context encoder network for 2d medical image segmenta- tion,”IEEE Transactions on Medical Imaging, vol. 38, no. 10, pp. 2281–2292, 2019

    2019

  40. [49]

    Cs- net: channel and spatial attention network for curvilinear structure segmentation,

    Lei Mou, Yitian Zhao, Li Chen, Jun Cheng, Zaiwang Gu, Huaying Hao, Hong Qi, Yalin Zheng, Alejandro Frangi, and Jiang Liu, “Cs- net: channel and spatial attention network for curvilinear structure segmentation,” inMedical image computing and computer-assisted intervention–Micca...

    2019

  41. [50]

    Transunet: Transformers make strong encoders for medical image segmenta- tion,

    Jieneng Chen, Yongyi Lu, Qihang Yu, Xiangde Luo, Ehsan Adeli, Yan Wang, Le Lu, Alan L Yuille, and Yuyin Zhou, “Transunet: Transformers make strong encoders for medical image segmenta- tion,”ArXiv preprint arXiv:2102.04306, 2021

    2021 arXiv

  42. [51]

    Swin-unet: Unet-like pure transformer for medical image segmentation,

    Hu Cao, Yueyue Wang, Joy Chen, Dongsheng Jiang, Xiaopeng Zhang, Qi Tian, and Manning Wang, “Swin-unet: Unet-like pure transformer for medical image segmentation,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 205–218

    2022

  43. [52]

    H2former: An efficient hierarchical hybrid trans- former for medical image segmentation,

    Along He, Kai Wang, Tao Li, Chengkun Du, Shuang Xia, and Huazhu Fu, “H2former: An efficient hierarchical hybrid trans- former for medical image segmentation,”IEEE Transactions on Medical Imaging, vol. 42, no. 9, pp. 2763–2775, 2023

    2023

  44. [53]

    Medformer: A multi-granularity patching transformer for medi- cal time-series classification,

    Yihe Wang, Nan Huang, Taida Li, Yujun Yan, and Xiang Zhang, “Medformer: A multi-granularity patching transformer for medi- cal time-series classification,”arXiv preprint arXiv:2405.19363, 2024

    2024 arXiv

  45. [54]

    Graph-based region and boundary aggregation for biomedical image segmentation,

    Yanda Meng, Hongrun Zhang, Yitian Zhao, Xiaoyun Yang, Yihong Qiao, Ian JC MacCormick, Xiaowei Huang, and Yalin Zheng, “Graph-based region and boundary aggregation for biomedical image segmentation,”IEEE Transactions on Medical Imaging, vol. 41, no. 3, pp. 690–701, 2021

    2021

  46. [55]

    nnu-net: a self-configuring method for deep learning-based biomedical image segmentation,

    Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Petersen, and Klaus H Maier-Hein, “nnu-net: a self-configuring method for deep learning-based biomedical image segmentation,”Nature Methods, vol. 18, no. 2, pp. 203–211, 2021

    2021

  47. [56]

    Diversified and personalized multi-rater medical image segmentation,

    Yicheng Wu, Xiangde Luo, Zhe Xu, Xiaoqing Guo, Lie Ju, Zongyuan Ge, Wenjun Liao, and Jianfei Cai, “Diversified and personalized multi-rater medical image segmentation,” inPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 11470–11479

    2024

  48. [57]

    Multi-view aggregation network for dichotomous image seg- mentation,

    Qian Yu, Xiaoqi Zhao, Youwei Pang, Lihe Zhang, and Huchuan Lu, “Multi-view aggregation network for dichotomous image seg- mentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 3921–3930

    2024

  49. [58]

    Towards au- tomatic polyp detection with a polyp appearance model,

    Jorge Bernal, Javier S ´anchez, and Fernando Vilarino, “Towards au- tomatic polyp detection with a polyp appearance model,”Pattern Recognition, vol. 45, no. 9, pp. 3166–3182, 2012

    2012

Pith tools

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