Pith. sign in

REVIEW 5 major objections 6 minor 34 references

Large Language Model Evaluated Stand-alone Attention-Assisted Graph Neural Network with Spatial and Structural Information Interaction for Precise Endoscopic Image Segmentation

T0 review · 5 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read FOCUS-Med claims state-of-the-art polyp segmentation on four colonoscopy benchmarks by fusing spatial and shortest-path structural graphs with location-aware self-attention.

desk verdict A plausible architecture, but the empirical claims are unsupported: missing training protocol and inconsistent baselines make the SOTA assertion unverifiable; the LLM evaluation is unvalidated. read the letter →

arxiv 2508.07028 v1 pith:4WWEGNLV submitted 2025-08-09 cs.CV

Juntong Fan , Shuyi Fan , Debesh Jha , Changsheng Fang , Tieyong Zeng , Hengyong Yu , Dayang Wang This is my paper
classification cs.CV
keywords polypsegmentationendoscopicimaginggraphneuralnetworkDual-GCNself-attentionDijkstrashortestpathmedicalimageLLMevaluation
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 argues that precise polyp segmentation in endoscopic images is best served by fusing two complementary graph views of the image—a spatial graph that captures local pixel relationships and a structural graph built from shortest paths that captures long-range topological dependencies—and then blending the result with a location-aware self-attention decoder. The proposed model, FOCUS-Med, adds trainable weighted fusion to combine multi-scale encoder–decoder features, and uses a large language model to provide clinically framed qualitative scoring of segmentation quality. On four public colonoscopy datasets, the paper reports state-of-the-art Dice, IoU, MAE, boundary F, and structure measure scores, and claims the first use of an LLM for evaluating such segmentations. If the results hold under a fair comparison, the architecture offers a practical way to combine graph-based structure reasoning with attention-based context modeling for a clinically important task.

What carries the argument

The load-bearing mechanism is the Dual-GCN feature enhancement block: a spatial graph constructed from pairwise pixel similarity and a structural graph whose edges are weighted by attention coefficients and whose messages follow cheapest paths (retained up to a maximum length). The path-level and cross-length attention updates in this block let distant but semantically related regions influence each other's features. Around this block sit LFSA, which injects relative position into self-attention keys, and WFNF, which adaptively fuses decoder, skip, and bottleneck features.

What would settle it

Retrain every baseline and FOCUS-Med from a single shared codebase with identical data splits, input resolution, optimizer, and epoch count, then compare Dice and IoU on all four datasets. The state-of-the-art claim stands only if the reported margins persist.

Watch

Extended reading notes

Core claim

FOCUS-Med is a ConvNeXt-based encoder-decoder whose bottleneck contains a Dual-GCN block: one graph models fine-grained spatial similarity among pixels, the other builds a structural graph whose edge costs come from averaged attention coefficients and whose node updates aggregate features along shortest paths, allowing each node to attend to high-order semantically relevant nodes. The decoder inserts a location-fused stand-alone self-attention (LFSA) that adds row and column offsets into the attention key, and a weighted fast normalized fusion (WFNF) combines three feature sources with trainable weights. The paper reports that this combination outperforms existing polyp segmentation models o

Load-bearing premise

The evaluation depends on all baseline models being trained and tested under the same protocol as FOCUS-Med, including data split, loss, optimizer, and input size; the paper does not disclose this protocol, and some reported baseline numbers differ widely from their published values.

Editorial extensions

If this is right

  • On the benchmarks tested, FOCUS-Med reports the highest Dice and IoU, and either best or near-best MAE, boundary F, and S-measure, across all four datasets.
  • The five-metric gains are statistically significant on Kvasir-SEG (Wilcoxon p-values all below 0.001), so the improvement is not attributed to a single lucky split.
  • The model runs at about 10.6 FPS with 43.6 million parameters, meaning the accuracy gains do not require prohibitive compute and could plausibly fit in a clinical workflow.
  • The LLM-based evaluation shows more consistent high scores on boundary accuracy and clinical usefulness, suggesting the model avoids low-quality tail cases.
  • Ablation results reported in the appendix attribute the gains to the Dual-GCN block, LFSA, and WFNF, with each component contributing.

Reading between the lines

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

  • The reported baseline scores (for example, PraNet at 69.35% Dice on CVC-ClinicDB) are far below the numbers published in the original baseline papers; a reader who re-runs all models under identical training settings may find the margins shrink, which would weaken the state-of-the-art claim.
  • The shortest-path structural graph has a natural interpretation as a message-passing scheme that lets features travel through chains of similar regions, a mechanism that could transfer to other medical segmentation tasks with elongated or branching structures such as vessels or nerves.
  • The LLM scoring protocol treats the language model as a proxy for clinical judgment, but the paper does not calibrate it against human raters; a direct comparison with gastroenterologist scores would test whether the LLM adds signal beyond Dice and IoU.
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 / 6 minor

Summary. FOCUS-Med is a deep-learning architecture for endoscopic polyp segmentation that combines a ConvNeXt encoder with (i) a Dual-GCN block consisting of a spatial similarity graph and a Dijkstra shortest-path structural graph, (ii) a location-fused stand-alone self-attention module (LFSA), and (iii) a weighted fast normalized fusion (WFNF). The paper reports quantitative results on Kvasir-SEG, CVC-ClinicDB, EndoScene, and CVC-ColonDB, claims state-of-the-art performance across five metrics, and introduces a GPT-4o-based qualitative evaluation protocol. An ablation study and related work are said to be in an appendix.

Significance. The paper proposes a reasonably well-motivated combination of graph-based structural reasoning, stand-alone self-attention, and learnable fusion for polyp segmentation, and it includes complexity comparisons and a first attempt at LLM-based segmentation assessment. The Dijkstra-derived shortest-path attention is a potentially interesting extension of graph attention to medical images. However, the empirical evidence for the central SOTA claim is not currently convincing. No code is made available, the training protocol is absent, and the reported baseline numbers are inconsistent with published results. The significance of the method therefore cannot be assessed until the evaluation is made reproducible and fair.

major comments (5)
  1. [§4.1, Table 1] The SOTA claim rests on a comparison that is not credible as reported. PraNet is listed at 69.35% Dice on CVC-ClinicDB, whereas the original PraNet paper reports ~90% for that dataset; U-Net at 66.79% is also far below common published values. The stated test sizes (30 CVC-ClinicDB, 200 CVC-ColonDB, 80 EndoScene, 120 Kvasir) do not match the standard splits of the baseline papers (e.g., 62 test images for ClinicDB in PraNet). Since no training protocol is given, the large margins in Table 1 may simply reflect undertrained baselines or different data usage. This must be fixed by re-running all methods under an identical, explicitly described protocol and reporting standard splits.
  2. [§4.1] The manuscript does not state the loss function, optimizer, learning rate, batch size, number of epochs, input resolution, data augmentation, or train/validation/test split. Without these details, none of the quantitative results, including the Wilcoxon tests in §4.1, can be reproduced. Please add a full experimental setup section and, ideally, release code to make the comparison verifiable.
  3. [Abstract and Table 1] The abstract states SOTA 'across five key metrics', but Table 1 itself contradicts this on Kvasir-SEG: FOCUS-Med has MAE 3.20% vs DCRNet 2.90%, Boundary F 79.43% vs 82.05%, and S-measure 91.43% vs 91.49%. Thus the model is not uniformly best on all five metrics on that dataset. Please qualify the claim per dataset/metric.
  4. [§5] The LLM evaluation is described as 'expert-aligned' but no evidence of alignment with human experts is provided. The prompt, GPT-4o version, temperature, number of test samples, and whether scores were averaged over repeated calls are not reported; no statistical test accompanies Table 2. At best this is an exploratory, LLM-simulated rating; it should not be presented as expert-aligned until calibrated against gastroenterologist judgments.
  5. [§3 and missing Appendix] Key hyperparameters for Algorithm 1 — maximum path length C, sampling ratio r, number of attention heads K — are never specified. Eq. (6) and (7) introduce trainable/weight parameters ω1, ω2, ω4–ω6 and ϵ without initialization or normalization constraints. Moreover, the paper repeatedly refers to an Appendix (related work, ablation tables, GPT-4o responses) that is not present in the submitted text. This prevents full evaluation of the method's design and the ablation claims.
minor comments (6)
  1. [§4.1] 'In summary, the proposed DSFNet...' — the model name should be FOCUS-Med.
  2. [Table 1] FOCUS-Med* uses an asterisk but no footnote defines it.
  3. [Throughout] Dataset/model naming is inconsistent: 'Endoscene'/'EndoScene', 'UNET'/'U-Net', 'ParNet'/'PraNet'.
  4. [References [10] and [25]] These references appear to point to papers on temporal action localization and fashion image retrieval, not the polyp segmentation baselines ACSNet and DCRNet used in Table 1. Verify these citations.
  5. [§4.1 Statistical Analysis] Wilcoxon tests are reported only for Kvasir-SEG; p-values for the other datasets are missing.
  6. [§5] Please specify the GPT-4o model version and API/interface details; the current citation to a general ChatGPT URL is insufficient for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the architectural components are defined independently of the reported metrics, the benchmark evaluation is external, and the GPT-4o overlay is a qualitative check rather than a fitted input.

full rationale

The paper is an empirical architecture paper rather than a derivation chain. Equations (1)-(7) define the Dual-GCN, LFSA, and WFNF modules as learnable feature transformations; none of these equations takes the reported evaluation metrics (Dice, IoU, MAE, Boundary F, S-measure) as inputs, and no fitted parameter is subsequently renamed as a prediction. The central SOTA claim is supported by Table 1, an external benchmark comparison, not by a mathematical derivation from the model definition. The unusually low baseline numbers (e.g., PraNet at 69.35% Dice on CVC-ClinicDB versus published values near 90%) and the absence of a full training protocol are serious reproducibility and fairness concerns, but they are correctness issues, not circularity: the baseline numbers are inputs to the comparison, not consequences of FOCUS-Med's own equations. The GPT-4o evaluation in Section 5 is a qualitative overlay in which the LLM is given the original image, predicted mask, and ground-truth mask and asked to score five Likert-scale criteria; it is self-referential in that the authors chose the LLM and prompt, and the term 'expert-aligned' overclaims without human calibration, but the LLM scores are not fitted parameters and do not enter the model's training or the primary quantitative derivation. The self-citations [4]-[8] are background examples of deep learning advances and are not load-bearing for FOCUS-Med's design or results. No uniqueness theorem, ansatz, or prior result by the same authors is invoked to force the model choice. The paper even acknowledges failure cases in Section 6 and Fig. 7, which is a limitation statement rather than a circular step. Overall, no claimed 'prediction' or 'first-principles result' reduces by construction to its own inputs.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim depends mainly on an unspecified training protocol and fair baseline comparison, not on a parameter-free derivation. The disclosed free hyperparameters (path length, sampling ratio, attention heads) are not assigned values. No new physical or conceptual entities are introduced.

free parameters (4)
  • C (max path length) = unspecified
    Introduced in Algorithm 1 as the maximum path length for Dijkstra-based graph attention; no value or tuning procedure is given.
  • r (sampling ratio) = unspecified
    Used in Algorithm 1 line 20 to select top-k paths per node; value not provided.
  • K (number of attention heads) = unspecified
    Used in Equation (2) and Algorithm 1; no value stated.
  • epsilon (in LFSA and WFNF) = 0.0001
    Small constant to avoid division by zero in Equations (6) and (7), chosen by hand.
assumptions (3)
  • standard math Dijkstra's algorithm correctly computes shortest paths on a fully connected graph with attention-derived weights.
    Invoked in Section 3.1.1, Algorithm 1. The algorithm is standard, but applying it to a dense image graph with N=H*W nodes assumes computational feasibility that is not analyzed.
  • domain assumption Public polyp segmentation datasets (Kvasir-SEG, CVC-ClinicDB, EndoScene, CVC-ColonDB) and their provided ground truth are reliable and the test splits are standard.
    Used in all experiments in Section 4. The paper does not state which images were used for training/validation/test beyond counts, so the split protocol is an assumption.
  • domain assumption The graph built from pixel feature similarity and shortest paths captures meaningful topological structure of polyps.
    Core motivation of the Dual-GCN block, Section 3.1. This is a modeling premise that is neither derived nor independently verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Language Model Evaluated Stand-alone Attention-Assisted Graph Neural Network with Spatial and Structural Information Interaction for Precise Endoscopic Image Segmentation." pith.science (2026). https://pith.science/paper/4WWEGNLV

@misc{pith2026250807028,
  author       = {Pith},
  title        = {Pith review of: Large Language Model Evaluated Stand-alone Attention-Assisted Graph Neural Network with Spatial and Structural Information Interaction for Precise Endoscopic Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4WWEGNLV}},
  note         = {Machine review of arXiv:2508.07028}
}
read the original abstract

Accurate endoscopic image segmentation on the polyps is critical for early colorectal cancer detection. However, this task remains challenging due to low contrast with surrounding mucosa, specular highlights, and indistinct boundaries. To address these challenges, we propose FOCUS-Med, which stands for Fusion of spatial and structural graph with attentional context-aware polyp segmentation in endoscopic medical imaging. FOCUS-Med integrates a Dual Graph Convolutional Network (Dual-GCN) module to capture contextual spatial and topological structural dependencies. This graph-based representation enables the model to better distinguish polyps from background tissues by leveraging topological cues and spatial connectivity, which are often obscured in raw image intensities. It enhances the model's ability to preserve boundaries and delineate complex shapes typical of polyps. In addition, a location-fused stand-alone self-attention is employed to strengthen global context integration. To bridge the semantic gap between encoder-decoder layers, we incorporate a trainable weighted fast normalized fusion strategy for efficient multi-scale aggregation. Notably, we are the first to introduce the use of a Large Language Model (LLM) to provide detailed qualitative evaluations of segmentation quality. Extensive experiments on public benchmarks demonstrate that FOCUS-Med achieves state-of-the-art performance across five key metrics, underscoring its effectiveness and clinical potential for AI-assisted colonoscopy.

Figures

Figures reproduced from arXiv: 2508.07028 by the authors.

Figure 1
Figure 1. The clinical acquisition of Colonoscopy image and ap [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall frameworks of FOCUS-Med. Path Feature Shortest [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The Dual-GCN enhancement block with contextual spa [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Representative qualitative results of different models on [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The feature maps of different models on a repsentative [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: The scores evaluated by LLM in terms of the five metrics. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The example of failing cases of FOCUS-Med from [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 27 canonical work pages

  1. [1]

    Colorec- tal cancer: a review

    Juan Jos ´e Granados-Romero, Alan Isaac Valderrama- Trevi˜no, Ericka Hazzel Contreras-Flores, Bal- tazar Barrera-Mera, Miguel Herrera Enr ´ıquez, Karen Uriarte-Ru ´ız, Jes ´us Carlos Ceballos- Villalba, Aranza Guadalupe Estrada-Mata, C Al- varado Rodr´ıguez, and Gerardo Arauz-Pe˜na. Colorec- tal cancer: a review. Int J Res Med Sci , 5(11):4667,

    work page

  2. [2]

    Biomarkers for early detection of colorectal cancer and polyps: systematic review

    Reena Shah, Emma Jones, Victoire Vidart, Pe- ter JK Kuppen, John A Conti, and Nader K Francis. Biomarkers for early detection of colorectal cancer and polyps: systematic review. Cancer Epidemiology, Biomarkers & Prevention, 23(9):1712–1728, 2014. 1

    work page 2014

  3. [3]

    Polyp-pvt: Polyp seg- mentation with pyramid vision transformers

    Bo Dong, Wenhai Wang, Deng-Ping Fan, Jinpeng Li, Huazhu Fu, and Ling Shao. Polyp-pvt: Polyp seg- mentation with pyramid vision transformers. arXiv preprint arXiv:2108.06932, 2021. 1

    arXiv 2021

  4. [4]

    Deep dual- domain united guiding learning with global–local transformer-convolution u-net for ldct reconstruction

    Zhan Wu, Xinyun Zhong, Tianling Lyv, Dayang Wang, Ruifeng Chen, Xu Yan, Gouenou Coatrieux, Xu Ji, Hengyong Yu, Yang Chen, et al. Deep dual- domain united guiding learning with global–local transformer-convolution u-net for ldct reconstruction. IEEE Transactions on Instrumentation and Measure- ment, 72:1–15, 2023. 1

    work page 2023

  5. [5]

    Lo- 8 mae: simple streamlined low-level masked autoen- coders for robust, generalized, and interpretable low- dose ct denoising

    Dayang Wang, Shuo Han, Yongshun Xu, Zhan Wu, Li Zhou, Bahareh Morovati, and Hengyong Yu. Lo- 8 mae: simple streamlined low-level masked autoen- coders for robust, generalized, and interpretable low- dose ct denoising. IEEE Journal of Biomedical and Health Informatics, 2024

    work page 2024

  6. [6]

    Physics-informed score-based diffu- sion model for limited-angle reconstruction of cardiac computed tomography

    Shuo Han, Yongshun Xu, Dayang Wang, Bahareh Mo- rovati, Li Zhou, Jonathan S Maltz, Ge Wang, and Hengyong Yu. Physics-informed score-based diffu- sion model for limited-angle reconstruction of cardiac computed tomography. IEEE Transactions on Medi- cal Imaging, 2024

    work page 2024

  7. [7]

    Enhancing pathogen identification in cheese with high background microflora using an artificial neural network-enabled paper chromogenic array sen- sor approach

    Zhen Jia, Zhuangsheng Lin, Yaguang Luo, Zachary A Cardoso, Dayang Wang, Genevieve H Flock, Kather- ine A Thompson-Witrick, Hengyong Yu, and Boce Zhang. Enhancing pathogen identification in cheese with high background microflora using an artificial neural network-enabled paper chromogenic array sen- sor approach. Sensors and Actuators B: Chemical , 410:135...

    work page 2024

  8. [8]

    Patch- based dual-domain photon-counting ct data correction with residual-based wgan-vit

    Bahareh Morovati, Mengzhou Li, Shuo Han, Li Zhou, Dayang Wang, Ge Wang, and Hengyong Yu. Patch- based dual-domain photon-counting ct data correction with residual-based wgan-vit. Physics in Medicine & Biology, 70(4):045008, 2025. 1

    work page 2025

Show all 34 references
  1. [9]

    Pranet: Parallel reverse attention network for polyp segmenta- tion

    Deng-Ping Fan, Ge-Peng Ji, Tao Zhou, Geng Chen, Huazhu Fu, Jianbing Shen, and Ling Shao. Pranet: Parallel reverse attention network for polyp segmenta- tion. In Anne L. Martel, Purang Abolmaesumi, Danail Stoyanov, Diana Mateus, Maria A. Zuluaga, S. Kevin Zhou, Daniel Racoceanu...

    2020

  2. [10]

    Acsnet: Action-context separation network for weakly super- vised temporal action localization

    Ziyi Liu, Le Wang, Qilin Zhang, Wei Tang, Jun- song Yuan, Nanning Zheng, and Gang Hua. Acsnet: Action-context separation network for weakly super- vised temporal action localization. 2021. 2, 5

    2021

  3. [11]

    An embedding-unleashing video polyp segmentation framework via region linking and scale alignment

    Zhixue Fang, Xinrong Guo, Jingyin Lin, Huisi Wu, and Jing Qin. An embedding-unleashing video polyp segmentation framework via region linking and scale alignment. In Proceedings of the AAAI conference on artificial intelligence, volume 38, pages 1744–1752,

  4. [12]

    A convnet for the 2020s

    Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. A convnet for the 2020s. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11976–11986, 2022. 2

    2022

  5. [13]

    Graph attention networks

    Petar Veli ˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks. arXiv preprint arXiv:1710.10903, 2017. 3

    2017 arXiv

  6. [14]

    Heterogeneous graph attention network

    Xiao Wang, Houye Ji, Chuan Shi, Bai Wang, Yanfang Ye, Peng Cui, and Philip S Yu. Heterogeneous graph attention network. In The world wide web conference, pages 2022–2032, 2019

    2022

  7. [15]

    Spagan: Shortest path graph attention network

    Yiding Yang, Xinchao Wang, Mingli Song, Junsong Yuan, and Dacheng Tao. Spagan: Shortest path graph attention network. arXiv preprint arXiv:2101.03464, 2021

    2021 arXiv

  8. [16]

    Graph attention networks: a com- prehensive review of methods and applications

    Aristidis G Vrahatis, Konstantinos Lazaros, and Sotiris Kotsiantis. Graph attention networks: a com- prehensive review of methods and applications. Fu- ture Internet, 16(9):318, 2024. 3

    2024

  9. [17]

    A note on two problems in connexion with graphs

    Edsger W Dijkstra. A note on two problems in connexion with graphs. Numerische mathematik , 1(1):269–271, 1959. 3

    1959

  10. [18]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NIPS, 2017. 4

    2017

  11. [19]

    Kvasir-seg: A segmented polyp dataset

    Debesh Jha, Pia H Smedsrud, Michael A Riegler, P˚al Halvorsen, Thomas de Lange, Dag Johansen, and H˚avard D Johansen. Kvasir-seg: A segmented polyp dataset. In MultiMedia Modeling: 26th International Conference, 2020. 5

    2020

  12. [20]

    Wm-dova maps for accurate polyp highlighting in colonoscopy: Validation vs

    Jorge Bernal, F Javier S ´anchez, Gloria Fern ´andez- Esparrach, Debora Gil, Cristina Rodr ´ıguez, and Fer- nando Vilari ˜no. Wm-dova maps for accurate polyp highlighting in colonoscopy: Validation vs. saliency maps from physicians. Computerized medical imag- ing and graphics,...

    2015

  13. [21]

    A benchmark for endoluminal scene segmentation of colonoscopy images

    David V ´azquez, Jorge Bernal, F Javier S ´anchez, Glo- ria Fern ´andez-Esparrach, Antonio M L ´opez, Adri- ana Romero, Michal Drozdzal, and Aaron Courville. A benchmark for endoluminal scene segmentation of colonoscopy images. J HEALTHC ENG, 2017. 5

    2017

  14. [22]

    Automated polyp detection in colonoscopy videos using shape and context information

    Nima Tajbakhsh, Suryakanth R Gurudu, and Jianming Liang. Automated polyp detection in colonoscopy videos using shape and context information. IEEE transactions on medical imaging , 35(2):630–644,

  15. [23]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Nassir Navab, Joachim Hornegger, William M. Wells, and Alejandro F. Frangi, editors, MICCAI, 2015. 5

    2015

  16. [24]

    Unet++: A nested u-net architecture for medical image segmen- tation

    Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, and Jianming Liang. Unet++: A nested u-net architecture for medical image segmen- tation. In DLMIA, 2018. 5

    2018

  17. [25]

    Dcr-net: Dilated convolutional residual network for fashion image retrieval

    Jia Chen, Haidongqing Yuan, Yi Zhang, Ruhan He, and Jinxing Liang. Dcr-net: Dilated convolutional residual network for fashion image retrieval. COM- PUT ANIMAT VIRT W, 2022. 5

    2022

  18. [26]

    U-kan makes strong backbone for medical im- age segmentation and generation

    Chenxin Li, Xinyu Liu, Wuyang Li, Cheng Wang, Hengyu Liu, Yifan Liu, Zhen Chen, and Yixuan Yuan. U-kan makes strong backbone for medical im- age segmentation and generation. In Proceedings of 9 the AAAI Conference on Artificial Intelligence , vol- ume 39, pages 4652–4660, 2025. 5

    2025

  19. [27]

    Structure-measure: A new way to eval- uate foreground maps

    Deng-Ping Fan, Ming-Ming Cheng, Yun Liu, Tao Li, and Ali Borji. Structure-measure: A new way to eval- uate foreground maps. In Proceedings of the IEEE international conference on computer vision , pages 4548–4557, 2017. 5

    2017

  20. [28]

    Individual comparisons by ranking methods

    Frank Wilcoxon. Individual comparisons by ranking methods. Biometrics Bulletin, 1(6):80–83, 1945. 6

    1945

  21. [29]

    Training-free seman- tic segmentation via llm-supervision

    Wenfang Sun, Yingjun Du, Gaowen Liu, Ramana Kompella, and Cees GM Snoek. Training-free seman- tic segmentation via llm-supervision. arXiv preprint arXiv:2404.00701, 2024. 7

    2024 arXiv

  22. [30]

    Llm-seg: Bridging image segmentation and large language model reasoning

    Junchi Wang and Lei Ke. Llm-seg: Bridging image segmentation and large language model reasoning. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 1765– 1774, 2024

    2024

  23. [31]

    Pre-trained llm is a semantic-aware and generalizable segmenta- tion booster

    Fenghe Tang, Wenxin Ma, Zhiyang He, Xiaodong Tao, Zihang Jiang, and S Kevin Zhou. Pre-trained llm is a semantic-aware and generalizable segmenta- tion booster. arXiv preprint arXiv:2506.18034, 2025. 7

    2025 arXiv

  24. [32]

    Mamba: Linear-time sequence modeling with selective state spaces

    Albert Gu and Tri Dao. Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752, 2023. 8

    2023 arXiv

  25. [33]

    Kan: Kolmogorov-arnold networks

    Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Solja ˇci´c, Thomas Y Hou, and Max Tegmark. Kan: Kolmogorov-arnold networks. arXiv preprint arXiv:2404.19756, 2024. 8

    2024 arXiv

  26. [34]

    Diffusion models: A comprehen- sive survey of methods and applications

    Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming-Hsuan Yang. Diffusion models: A comprehen- sive survey of methods and applications. ACM com- puting surveys, 56(4):1–39, 2023. 8 10

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.