Pith. sign in

REVIEW 4 major objections 6 minor 50 references

ViG3D-UNet: Volumetric Vascular Connectivity-Aware Segmentation via 3D Vision Graph Representation

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read A 3D vision-graph U-Net produces coronary artery segmentations that stay connected while matching or beating four baselines on two CTA datasets.

desk verdict Architecturally plausible and honestly reported, but the connectivity headline is not actually measured: the evaluation uses precision and HD95 as proxies, and the gains over STUNet are within the reported noise. read the letter →

arxiv 2504.13599 v1 pith:HXACCKKT submitted 2025-04-18 eess.IV cs.CV

classification eess.IVcs.CV
keywords coronaryarterysegmentation3DvisiongraphneuralnetworkvascularconnectivitycomputedtomographyangiographyU-Netrepresentationmedicalimagetubularstructure
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes ViG3D-UNet, a U-shaped segmentation network that adds a 3D vision graph neural network branch alongside a convolutional encoder, so that vessels are treated as connected graphs of voxels during training rather than only as textured volumes. The authors aim to show that this graph representation reduces the fragmentation and missing distal branches that plague coronary artery segmentation from CTA while preserving, or even improving, overlap accuracy. On the large ImageCAS dataset they report the best DSC (81.72), HD95 (15.21), IoU (69.40), precision (83.16), and recall (80.79) among four competing baselines; on the smaller ASOCA dataset they report the best HD95 (13.39) and precision (90.39) and second-best DSC (84.23). A sympathetic reader would care because coronary segmentation feeds reconstruction and stenosis quantification, where a few breaks can change the diagnosis.

What carries the argument

The load-bearing object is the ViG3D block, a volumetric graph-aggregation unit. A stem reduces an input patch to $\frac{H}{4} \times \frac{W}{4} \times \frac{D}{4}$, each voxel feature becomes a graph node, and each node is connected to its $K=7$ nearest neighbors; the block then applies max-relative graph convolution, $g(\cdot) = [x_i, \max(\{x_j - x_i \mid j \in \mathcal{N}(x_i)\})]$, followed by an MLP update and a feed-forward network, stacked in a pyramid of 2, 4, 16, and 2 blocks across scales. The graph aggregation is what transmits same-branch information between nearby vascular voxels, while the parallel CNN branch supplies texture; channel attention fuses the two, and the paperclip-shaped offset decoder concatenates early CNN texture features into the last decoder layers and fusion features into the first four layers to avoid redundant 3D computation.

What would settle it

Compute the number of connected components, the fraction of ground-truth branch endpoints covered by the prediction, or the centerline continuity on the ASOCA and ImageCAS test folds; if ViG3D-UNet does not improve these direct measures over the baselines at matched thresholds, the central connectivity claim is falsified.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central discovery is that embedding a 3D Vision GNN module as one branch of a parallel encoder, fusing its topological features with CNN texture features through channel attention, and decoding with a paperclip-shaped offset decoder produces volumetric coronary artery segmentations that maintain connectivity across branches and at endpoints better than standard CNN and transformer baselines. The evidence is the score table: on ImageCAS the method has the best DSC, HD95, IoU, precision, and recall among the four compared methods, with DSC 81.72 against 81.54 for the strongest baseline and HD95 15.21 against 17.60 for the self-configuring U-Net baseline; on ASOCA it has the best HD95 and precision and second-best DSC (84.23, within 0.57 of the strongest baseline). The authors interpret the lower HD95 and higher precision, together with qualitative figures showing fewer breaks, as evidence of superior connectivity.

Load-bearing premise

The connectivity claim rests on treating precision and HD95 as measures of vascular connectivity; if those two metrics do not track true branch continuity, the paper has not actually shown better connectivity.

Editorial extensions

If this is right

  • On ImageCAS, ViG3D-UNet reports the best DSC (81.72), HD95 (15.21), IoU (69.40), precision (83.16), and recall (80.79) among the four compared baselines.
  • On ASOCA, it reports the best HD95 (13.39) and precision (90.39) and second-best DSC (84.23), suggesting the graph prior helps when training data are scarce.
  • Ablation studies attribute the gain to all three components: removing the ViG3D module lowers DSC by 0.61 and raises HD95 by 2.39, removing channel attention lowers DSC by 0.51, and removing the offset decoder lowers DSC by 1.16 and raises HD95 by 3.13.
  • Because the pipeline is end-to-end, it avoids the two-stage graph-construction-then-refinement cascade of earlier graph-based vessel segmentation, which should simplify training.

Reading between the lines

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

  • A direct topological evaluation—connected-component count, fraction of ground-truth branch endpoints recovered, or centerline continuity—would be a sharper test of the connectivity claim, and the paper does not report such metrics.
  • If the connectivity gain is real, the same graph-in-U-Net design should transfer to other sparse tubular structures such as airways, retinal vessels, and cerebral arteries, since the paper tests only coronary arteries.
  • The optimal neighbor count $K=7$ and the stacked ViG3D configuration were tuned on ImageCAS; a testable extension is whether the best $K$ scales with voxel anisotropy or branch thickness across datasets.
  • The lower standard deviation on the small ASOCA dataset hints that graph aggregation regularizes the model under limited training data, which could be checked by training on progressively smaller training subsets.
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 ViG3D-UNet, a 3D U-shaped segmentation network whose encoder combines a 3D CNN branch with a 3D vision graph neural network (ViG3D) branch, fuses the two feature streams with channel attention, and employs a 'paperclip-shaped' offset decoder. The method is evaluated on two public coronary CTA datasets, ASOCA and ImageCAS, against nnU-Net, STUNet, UNETR, and SwinUNETR. The authors claim that ViG3D-UNet achieves state-of-the-art accuracy and, as the central contribution, superior vascular connectivity, supported by precision, HD95, and visual examples. Ablation studies examine the graph neighbor count K, the number of ViG3D blocks per scale, and the contribution of each architectural module.

Significance. If the connectivity claim were established, this would be a useful empirical contribution: it offers an end-to-end graph-based alternative to two-stage graph segmentation pipelines and reports competitive accuracy on two public datasets against strong baselines. The architectural idea is clearly presented, and the use of public datasets plus a fair baseline setup is commendable. However, the paper's central claim—connectivity improvement—is supported only by proxy metrics, the reported accuracy gains over the strongest baseline are small, and no significance testing is provided. With direct topological metrics and proper statistical evaluation, the work could be a solid contribution; in its current form the headline claim is overstated.

major comments (4)
  1. [Section IV.A.3 and Section IV.D, Table V] The central claim that ViG3D-UNet 'surpassed competing methods in maintaining vascular segmentation connectivity' is not supported by the reported metrics. Precision (Eq. 11) counts voxel-level false positives, and HD95 (Eq. 9) measures surface distance; neither quantifies topology. A prediction can be highly precise and have small HD95 while still containing broken branches, disconnected components, or missing distal segments. The lower recall of ViG3D-UNet relative to STUNet on both datasets (ASOCA 79.65 vs 82.31; ImageCAS 80.79 vs 83.51) is consistent with omitted branches, which is a direct connectivity failure. I request direct connectivity metrics over the full test sets: number of connected components per coronary tree, fraction of ground-truth branch endpoints covered, centerline overlap or Dice on centerlines, and/or Betti-number error. The anecdotal visual evidence in Figures 6 and 7 should also be replaced or supplemented by these quantitative topology measurements.
  2. [Section IV.D, Table V and Fig. 5] No variance or significance testing is reported for the comparison against baselines. Table II shows the proposed method's standard deviations are large relative to the reported differences (e.g., ImageCAS DSC 81.72 ± 5.09 vs STUNet 81.54; HD95 15.21 ± 18.53 vs STUNet 17.97). Since the improvements over the strongest baseline are 0.57 DSC and 2.39 HD95 on ImageCAS, the reader cannot determine whether these differences are statistically reliable. Please report per-fold or per-case results and apply paired tests (e.g., Wilcoxon signed-rank or paired t-test on per-case DSC and HD95) across the five folds or on the test set, and make the figures show these comparisons explicitly.
  3. [Section IV.A.2 and Section IV.C] The hyperparameters K=7 and the per-scale ViG3D block counts (2, 4, 16, 2) are selected by experiments on the ImageCAS dataset, which is also the dataset used for the final head-to-head comparison. This makes the reported ImageCAS numbers a selection outcome rather than an unbiased evaluation. The authors should either perform nested validation on ImageCAS or demonstrate that the same hyperparameters are optimal on ASOCA before claiming generalizable connectivity improvement. Additionally, the paper's own limitation—'a blind search of feature distances is involved in the current spatial graph representation'—acknowledges that the graph construction is not informed by vessel structure, which weakens the mechanistic claim that the graph module specifically encodes vascular connectivity.
  4. [Section IV.A.1 and Section IV.A.2] The experimental protocol is described inconsistently. Section IV.A.1 states that ASOCA is 'strategically divided into 30 samples for training and 10 samples for testing' and ImageCAS into 800/200, while Section IV.A.2 says 'A five-fold cross-validation strategy was used for experiments on both datasets.' These statements cannot both describe the identical protocol. Please clarify whether the reported numbers are from a fixed split or from an aggregation of five folds, and if five-fold cross-validation was used, describe how the folds were constructed and why the dataset descriptions mention fixed splits.
minor comments (6)
  1. [Table V] The column header 'Recall↓' is incorrect; recall is a higher-is-better metric and should be marked with an upward arrow.
  2. [Section III.A, after Eq. (3)] In the sentence 'For each voxel vi, its neighboring voxels N (vj) were identified using the K-nearest neighbors method,' the notation N(vj) should be N(vi).
  3. [Section IV.C] The text says the impact of different numbers of ViG3D blocks is shown in Table IV, but the ViG3D unit scaling results are in Table III; Table IV contains the structural ablations.
  4. [Section IV.B] The sentence 'Experiments of a five-fold cross-validation strategy on the ASOCA and ImageCAS datasets, respectively' is grammatically incomplete.
  5. [Figures 6 and 7] The figure captions concatenate model names without spaces (e.g., 'nnUNet STU-Net Swin-UNETRUNETR ViG3D-UNetGT'); these should be separated for readability.
  6. [Table V title] The table title contains a stray closing parenthesis: 'WITH THE STATE-OF-THE-ART ALGORITHMS . )'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: all central claims rest on external benchmark evaluations; connectivity-proxy concerns are a validity issue, not a definitional reduction.

full rationale

ViG3D-UNet is an empirical architecture paper rather than a derivation. The central claim—superior vascular connectivity with high segmentation accuracy—is supported by five-fold cross-validated comparisons on two public datasets (ASOCA and ImageCAS) against external baselines (nnU-Net, STUNet, UNETR, SwinUNETR). The network components are built from standard graph-convolution and U-Net operations, and the reported metrics (DSC, IoU, precision, recall, ASSD, HD95) are defined independently of the method's inputs. No equation in the paper defines an input parameter in terms of an output metric, and no fitted parameter is later renamed as a prediction: the K and L ablations are hyperparameter selection, and the final comparison is an external benchmark result. The one overlapping-author citation ([20], used to support the general statement that CNN-based networks have improved segmentation accuracy) is not load-bearing for any of the paper's claims, so it does not constitute circularity under the stated rules. The concern that precision and HD95 are imperfect proxies for true vascular connectivity is a construct-validity or correctness issue, not a circularity reduction: the paper does not define connectivity as precision or HD95, nor define precision/HD95 in terms of connectivity. The paper's own limitations (small training sets; 'a blind search of feature distances' in graph construction) further indicate that the results are contingent empirical findings rather than consequences of the method's definitions. No self-definitional, fitted-input-called-prediction, self-citation-load-bearing, uniqueness-imported, ansatz-smuggled, or renaming step is present.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new physical entities. It depends on standard domain assumptions about graph construction, metric interpretation, and training comparability. The free parameters are architecture hyperparameters tuned on the same public benchmarks used for the final comparison.

free parameters (2)
  • K-nearest neighbors K = 7
    Tuned on the ImageCAS validation set; Fig. 4 shows the best DSC and HD95 at K=7. This controls the range of graph aggregation and is central to the ViG3D module.
  • ViG3D block counts per scale = 2, 4, 16, 2
    Selected from the scaling experiment in Table III; the Small to Huge variants differ in block counts, and Large (2,4,16,2) is reported as best on ImageCAS.
assumptions (3)
  • domain assumption Vascular topology can be captured by K-nearest-neighbor graphs built on dense voxel features.
    Section III.A constructs graph edges from KNN over voxels and assumes these edges enhance information transmission along vascular branches.
  • domain assumption Precision and HD95 are meaningful proxies for vascular connectivity.
    Section IV.A.3 defines precision as an indicator of connectivity and the Discussion uses HD95 to conclude superior connectivity, without validating these proxies against direct topology metrics.
  • domain assumption All compared models are trained under equivalent protocols, making the comparisons fair.
    Section IV.A.2 states all models are trained from scratch with the nnU-Net framework; the validity of the state-of-the-art claim depends on this equivalence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ViG3D-UNet: Volumetric Vascular Connectivity-Aware Segmentation via 3D Vision Graph Representation." pith.science (2026). https://pith.science/paper/HXACCKKT

@misc{pith2026250413599,
  author       = {Pith},
  title        = {Pith review of: ViG3D-UNet: Volumetric Vascular Connectivity-Aware Segmentation via 3D Vision Graph Representation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HXACCKKT}},
  note         = {Machine review of arXiv:2504.13599}
}
read the original abstract

Accurate vascular segmentation is essential for coronary visualization and the diagnosis of coronary heart disease. This task involves the extraction of sparse tree-like vascular branches from the volumetric space. However, existing methods have faced significant challenges due to discontinuous vascular segmentation and missing endpoints. To address this issue, a 3D vision graph neural network framework, named ViG3D-UNet, was introduced. This method integrates 3D graph representation and aggregation within a U-shaped architecture to facilitate continuous vascular segmentation. The ViG3D module captures volumetric vascular connectivity and topology, while the convolutional module extracts fine vascular details. These two branches are combined through channel attention to form the encoder feature. Subsequently, a paperclip-shaped offset decoder minimizes redundant computations in the sparse feature space and restores the feature map size to match the original input dimensions. To evaluate the effectiveness of the proposed approach for continuous vascular segmentation, evaluations were performed on two public datasets, ASOCA and ImageCAS. The segmentation results show that the ViG3D-UNet surpassed competing methods in maintaining vascular segmentation connectivity while achieving high segmentation accuracy. Our code will be available soon.

Figures

Figures reproduced from arXiv: 2504.13599 by the authors.

Figure 1
Figure 1. Limitations of general-purpose coronary CTA segmentation: vascular discontinuity and incomplete distal branch detection. patches as input due to the high computational demands of deep learning networks [9], [15] . This patch-based approach frequently neglects the integration of prior knowledge about vascular topology from the trunk to the branches. Moreover, compared to CT medical images of abdominal and chest organ… view at source ↗
Figure 2
Figure 2. Overview of ViG3D-UNet Architecture. The encoder is composed of a 3D vision GNN module and a 3D CNN module.The two modules are combined into a parallelized encoder under the block of channel attention. The fusion feature obtained via channel attention is concatenated with the first three layers of the decoder through skip connections to form the decoder feature. The first two layers of texture features, obtained thr… view at source ↗
Figure 3
Figure 3. Illustration of the 3D Vision GNN Module which identifies the connections between vascular nodes through graph aggregation operations during network training. (a) An overview of the branch of 3D Vision GNN in parallelized encoder. (b) Two stages in 3D Vision GNN Block: 3D graph processing and feed forward network. culminating in the creation of a channel attention map FAM ∈ R 1×1×1×(Ci+Cj ) . This attention map is t… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Performance comparison with varying numbers of neighbor nodes K on the ImagesCAS dataset, with K values ranging from 3 to 32. Optimal model performance is observed when K is set to 7. In addition to the number of neighboring nodes, the number of ViG3D blocks (denoted a…
Figure 5
Figure 5. Figure 5: Box plot of HD95 values for individual cases across different methods. Outliers, or data points that deviate significantly from the HD95 mean, are marked with circles Quantitative comparisons with state-of-the-art methods are shown in Table V. In the ASOCA dataset, the…
Figure 6
Figure 6. Figure 6: In the visual comparison on the small-scale ASOCA dataset, our method yields more complete segmentation results at the terminal branches of the coronary arteries. GT nnUNet STU-Net UNETR Swin-UNETR ViG3D-UNet [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: In the visual comparison on the ImageCAS dataset, our method maintains greater connectivity at the coronary artery endpoints. Even when segmentation breaks occur, it results in fewer disconnected voxels compared to other methods. Additionally, the segmentation results …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 32 canonical work pages

  1. [1]

    It consists of 40 annotated Cardiac CTA images with an initial resolution of 512× 512× N , where N ranges from 168 to 334

    Datasets and Description: The Automated Segmentation of Coronary Arteries (ASOCA) dataset, used in the MICCAI- 2020 Challenge [38], is a publicly accessible dataset tailored TABLE I DETAILED SETTINGS OF PARALLELIZED ENCODER Layer Output Size CNN Module ViG3D Module Enc 1 H 2 × W 2 × D 2 Conv3d C = 32 S = 1 Downsampling Conv3d C = 32 S = 2 × 1 Enc 2 H 4 × ...

  2. [2]

    Fur- thermore, the structure of ViG3D was verified through ablation studies, with specific details provided in the following sections

    Implementation Details: The optimal hyperparameters for the neighbor nodes in the ViG3D module were determined by primarily evaluating the performance of the method while adjusting the number of nodes for graph representation. Fur- thermore, the structure of ViG3D was verified through ablation studies, with specific details provided in the following secti...

  3. [3]

    The DSC metrics reflect how accurately the model captures coronary structures

    Performance Metrics: The Dice Similarity Coefficient (DSC) is employed to measures the accuracy of segmentation results by evaluating the overlap between predicted and actual segmentation maps, defined as: DSC = 2×TP 2×TP +FP +FN , (7) where TP (True Positive) represents the number of pixels correctly identified as part of the coronary class, FP (False Po...

  4. [4]

    3d graph-connectivity constrained network for hepatic vessel segmen- tation,

    R. Li, Y .-J. Huang, H. Chen, X. Liu, Y . Yu, D. Qian, and L. Wang, “3d graph-connectivity constrained network for hepatic vessel segmen- tation,” IEEE Journal of Biomedical and Health Informatics , vol. 26, no. 3, pp. 1251–1262, 2022

  5. [5]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 . Springer, 2015, pp. 234–241

  6. [6]

    A review of deep-learning-based medical image segmentation methods,

    X. Liu, L. Song, S. Liu, and Y . Zhang, “A review of deep-learning-based medical image segmentation methods,” Sustainability, vol. 13, no. 3, p. 1224, 2021

  7. [7]

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

    F. Isensee, P. F. Jaeger, S. A. Kohl, J. Petersen, and K. 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

  8. [9]

    G. A. Roth, D. Abate, K. H. Abate, S. M. Abay, C. Abbafati, N. Abbasi, H. Abbastabar, F. Abd-Allah, J. Abdela, A. Abdelalim et al. , “Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980–2017: a systematic analysis for the global burden of disease study 2017,” The lancet, vol. 392, no. 10...

Show all 50 references
  1. [10]

    indicates that segmentation performance is significantly improved by utilizing the geometric structure of the target object. Recently, the achievements of large language models in medical image segmentation [11] have led to an enhancement in the ability of segmentation models ...

  2. [11]

    Cooper, J

    R. Cooper, J. Cutler, P. Desvigne-Nickens, S. P. Fortmann, L. Friedman, R. Havlik, G. Hogelin, J. Marler, P. McGovern, G. Morosco et al. , “Trends and disparities in coronary heart disease, stroke, and other cardiovascular diseases in the united states: findings of the nationa...

  3. [12]

    These limitations are exacerbated by conventional models’ inability to simul- taneously handle tubular geometry priors and microvascular contrast variations

    , and (3) ill-defined boundaries [13]. These limitations are exacerbated by conventional models’ inability to simul- taneously handle tubular geometry priors and microvascular contrast variations. First, the U-Net framework and its variants struggle to capture all low-level fe...

  4. [13]

    Computed tomographic angiography in coronary artery disease,

    P. W. Serruys, N. Kotoku, B. L. Nørgaard, S. Garg, K. Nieman, M. R. Dweck, J. J. Bax, J. Knuuti, J. Narula, D. Perera et al. , “Computed tomographic angiography in coronary artery disease,” EuroIntervention, vol. 18, no. 16, p. e1307, 2023

  5. [14]

    Uctransnet: rethinking the skip connections in u-net from a channel-wise perspective with trans- former,

    H. Wang, P. Cao, J. Wang, and O. R. Zaiane, “Uctransnet: rethinking the skip connections in u-net from a channel-wise perspective with trans- former,” in Proceedings of the AAAI conference on artificial intelligence, vol. 36, no. 3, 2022, pp. 2441–2449

  6. [15]

    Unetr: Transformers for 3d medical image segmentation,

    A. Hatamizadeh, Y . Tang, V . Nath, D. Yang, A. Myronenko, B. Land- man, H. R. Roth, and D. Xu, “Unetr: Transformers for 3d medical image segmentation,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , 2022, pp. 574–584

  7. [16]

    In CNN-based networks, deeper architectures

    have become the mainstream method for medical image segmentation. In CNN-based networks, deeper architectures

  8. [17]

    The attention mechanism [19] was introduced for more effective aggregation of contextual information

    and denser connections between layers [18] are employed to improve the performance of segmentation algorithms. The attention mechanism [19] was introduced for more effective aggregation of contextual information. Although CNN-based networks have improved segmentation accuracy ...

  9. [18]

    Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,

    A. Hatamizadeh, V . Nath, Y . Tang, D. Yang, H. R. Roth, and D. Xu, “Swin unetr: Swin transformers for semantic segmentation of brain tumors in mri images,” in International MICCAI Brainlesion Workshop. Springer, 2021, pp. 272–284

  10. [20]

    Segment anything model for medical image segmentation: Current applications and future directions,

    Y . Zhang, Z. Shen, and R. Jiao, “Segment anything model for medical image segmentation: Current applications and future directions,” Com- puters in Biology and Medicine , vol. 171, p. 108238, 2024

  11. [21]

    An energy matching vessel segmentation framework in 3-d medical images,

    P. Liu, G. Huang, J. Jing, S. Bian, L. Cheng, X. Y . Lu, C. Rao, Y . Liu, Y . Hua, Y . Wang, and K. He, “An energy matching vessel segmentation framework in 3-d medical images,” IEEE Transactions on Medical Imaging, vol. 43, no. 4, pp. 1476–1488, 2024

  12. [22]

    Improved topological preservation in 3d axon segmentation and centerline detec- tion using geometric assessment-driven topological smoothing (gats),

    N. I. Shamsi, A. S. Xu, L. A. Gjesteby, and L. J. Brattain, “Improved topological preservation in 3d axon segmentation and centerline detec- tion using geometric assessment-driven topological smoothing (gats),” in Proceedings of the IEEE/CVF Winter Conference on Applications o...

  13. [23]

    Coronary artery centerline extraction in cardiac ct angiography using a cnn-based orientation classifier,

    J. M. Wolterink, R. W. van Hamersvelt, M. A. Viergever, T. Leiner, and I. I ˇsgum, “Coronary artery centerline extraction in cardiac ct angiography using a cnn-based orientation classifier,” Medical image analysis, vol. 51, pp. 46–60, 2019

  14. [24]

    3d u-net: learning dense volumetric segmentation from sparse annotation,

    ¨O. C ¸ ic ¸ek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ron- neberger, “3d u-net: learning dense volumetric segmentation from sparse annotation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, Oc...

  15. [25]

    Stu-net: Scalable and transferable medical image segmentation models empowered by large-scale supervised pre-training,

    Z. Huang, H. Wang, Z. Deng, J. Ye, Y . Su, H. Sun, J. He, Y . Gu, L. Gu, S. Zhang et al., “Stu-net: Scalable and transferable medical image segmentation models empowered by large-scale supervised pre-training,” arXiv preprint arXiv:2304.06716 , 2023

  16. [26]

    Unet++: Redesigning skip connections to exploit multiscale features in image segmentation,

    Z. Zhou, M. M. R. Siddiquee, N. Tajbakhsh, and J. Liang, “Unet++: Redesigning skip connections to exploit multiscale features in image segmentation,” IEEE transactions on medical imaging , vol. 39, no. 6, pp. 1856–1867, 2019

  17. [27]

    Adaptive context selection for polyp segmentation,

    R. Zhang, G. Li, Z. Li, S. Cui, D. Qian, and Y . Yu, “Adaptive context selection for polyp segmentation,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part VI 23 . Springer...

  18. [28]

    An improved u-shape neural network for soft exudate segmentation,

    H. Zhang, K. Lin, Y . Guan, Z. Gan, and C. Ouyang, “An improved u-shape neural network for soft exudate segmentation,” in Fourteenth International Conference on Digital Image Processing (ICDIP 2022) , vol. 12342. SPIE, 2022, pp. 395–401

  19. [29]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, and S. Gelly, “An image is worth 16x16 words: Transformers for image recognition at scale,” in International Conference on Learning Representations , 2021

  20. [30]

    Medical transformer: Gated axial-attention for medical image segmentation,

    J. M. J. Valanarasu, P. Oza, I. Hacihaliloglu, and V . M. Patel, “Medical transformer: Gated axial-attention for medical image segmentation,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September...

  21. [31]

    Exfuse: Enhancing feature fusion for semantic segmentation,

    Z. Zhang, X. Zhang, C. Peng, X. Xue, and J. Sun, “Exfuse: Enhancing feature fusion for semantic segmentation,” Springer, Cham, 2018

  22. [32]

    Ori- net: Orientation-guided neural network for automated coronary arteries segmentation,

    W. Jiang, Y . Li, Y . Jia, Y . Feng, Z. Yi, M. Chen, and J. Wang, “Ori- net: Orientation-guided neural network for automated coronary arteries segmentation,” Expert Systems with Applications , vol. 238, p. 121905, 2024

  23. [33]

    Deep distance transform for tubular structure segmentation in ct scans,

    Y . Wang, X. Wei, F. Liu, J. Chen, Y . Zhou, W. Shen, E. K. Fishman, and A. L. Yuille, “Deep distance transform for tubular structure segmentation in ct scans,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 3833–3842. 11

  24. [34]

    Cs2-net: Deep learning segmentation of curvilinear structures in medical imaging,

    L. Mou, Y . Zhao, H. Fu, Y . Liu, J. Cheng, Y . Zheng, P. Su, J. Yang, L. Chen, A. F. Frangi et al. , “Cs2-net: Deep learning segmentation of curvilinear structures in medical imaging,” Medical image analysis , vol. 67, p. 101874, 2021

  25. [35]

    Vision gnn: An image is worth graph of nodes,

    K. Han, Y . Wang, J. Guo, Y . Tang, and E. Wu, “Vision gnn: An image is worth graph of nodes,” Advances in neural information processing systems, vol. 35, pp. 8291–8303, 2022

  26. [36]

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

    Y . Meng, H. Zhang, Y . Zhao, X. Yang, Y . Qiao, I. J. MacCormick, X. Huang, and Y . Zheng, “Graph-based region and boundary aggregation for biomedical image segmentation,” IEEE transactions on medical imaging, vol. 41, no. 3, pp. 690–701, 2021

  27. [37]

    A joint 3d unet-graph neural network-based method for airway segmen- tation from chest cts,

    A. Garcia-Uceda Juarez, R. Selvan, Z. Saghir, and M. de Bruijne, “A joint 3d unet-graph neural network-based method for airway segmen- tation from chest cts,” in Machine Learning in Medical Imaging: 10th International Workshop, MLMI 2019, Held in Conjunction with MICCAI 2019, ...

  28. [38]

    Graph convolution based cross-network multiscale feature fusion for deep vessel segmentation,

    G. Zhao, K. Liang, C. Pan, F. Zhang, X. Wu, X. Hu, and Y . Yu, “Graph convolution based cross-network multiscale feature fusion for deep vessel segmentation,” IEEE Transactions on Medical Imaging , vol. 42, no. 1, pp. 183–195, 2022

  29. [39]

    Imagecas: A large-scale dataset and benchmark for coronary artery segmentation based on computed tomography angiography images,

    A. Zeng, C. Wu, G. Lin, W. Xie, J. Hong, M. Huang, J. Zhuang, S. Bi, D. Pan, N. Ullah et al. , “Imagecas: A large-scale dataset and benchmark for coronary artery segmentation based on computed tomography angiography images,” Computerized Medical Imaging and Graphics, vol. 109,...

  30. [40]

    was applied to the last five decoders, with the loss weights successively halved from Decoder 6 to Decoder 2

  31. [41]

    Feature fusion for segmentation and classification of skin lesions,

    Y . Zhang, Z. Chen, H. Yu, X. Yao, and H. Li, “Feature fusion for segmentation and classification of skin lesions,” in 2022 IEEE 19th international symposium on biomedical imaging (ISBI) . IEEE, 2022, pp. 1–5

  32. [42]

    Scs-net: A scale and context sensitive network for retinal vessel segmentation,

    H. Wu, W. Wang, J. Zhong, B. Lei, Z. Wen, and J. Qin, “Scs-net: A scale and context sensitive network for retinal vessel segmentation,” Medical Image Analysis , vol. 70, p. 102025, 2021. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1361841521000712

  33. [43]

    Vigu: Vi- sion gnn u-net for fast mri,

    J. Huang, A. I. Aviles-Rivero, C.-B. Sch ¨onlieb, and G. Yang, “Vigu: Vi- sion gnn u-net for fast mri,” in 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI) . IEEE, 2023, pp. 1–5

  34. [44]

    Deepgcns: Can gcns go as deep as cnns?

    G. Li, M. Muller, A. Thabet, and B. Ghanem, “Deepgcns: Can gcns go as deep as cnns?” in Proceedings of the IEEE/CVF international conference on computer vision , 2019, pp. 9267–9276

  35. [45]

    Gaussian error linear units (gelus),

    D. Hendrycks and K. Gimpel, “Gaussian error linear units (gelus),” arXiv preprint arXiv:1606.08415, 2016

  36. [46]

    Instance normalization: The missing ingredient for fast stylization,

    D. Ulyanov, A. Vedaldi, and V . Lempitsky, “Instance normalization: The missing ingredient for fast stylization,”arXiv preprint arXiv:1607.08022, 2016

  37. [47]

    Automated segmen- tation of normal and diseased coronary arteries–the asoca challenge,

    R. Gharleghi, D. Adikari, K. Ellenberger, S.-Y . Ooi, C. Ellis, C.-M. Chen, R. Gao, Y . He, R. Hussain, C.-Y . Leeet al., “Automated segmen- tation of normal and diseased coronary arteries–the asoca challenge,” Computerized Medical Imaging and Graphics, vol. 97, p. 102049, 2022

  38. [48]

    Deeply-supervised nets,

    C.-Y . Lee, S. Xie, P. Gallagher, Z. Zhang, and Z. Tu, “Deeply-supervised nets,” in Artificial intelligence and statistics . Pmlr, 2015, pp. 562–570

  39. [49]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

  40. [50]

    Innovations in cardiac computed tomography: Imaging in coronary artery disease,

    V . S. Manubolu, K. Ichikawa, and M. J. Budoff, “Innovations in cardiac computed tomography: Imaging in coronary artery disease,” Progress in Cardiovascular Diseases, 2024

  41. [51]

    Non-invasive imaging of the coronary arteries,

    P. D. Adamson and D. E. Newby, “Non-invasive imaging of the coronary arteries,” European Heart Journal , vol. 40, no. 29, pp. 2444–2454, 11 2018

  42. [52]

    Fractal analysis of the vascular tree in the human retina

    Masters and R. Barry, “Fractal analysis of the vascular tree in the human retina.” Annual Review of Biomedical Engineering , vol. 6, no. 1, pp. 427–452, 2004

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.