Pith. sign in

REVIEW 3 major objections 6 minor 53 references

Two-Steps Neural Networks for an Automated Cerebrovascular Landmark Detection

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a two-step pipeline—nnDetection for coarse regions of interest followed by a U-Net for heatmap refinement—achieves the highest Circle of Willis bifurcation landmark detection accuracy on its test sets, with mean…

desk verdict A sensible two-step cascade for Circle of Willis landmarks, but the 'highest performance' claim is under-supported without the [40] comparison and error bars. read the letter →

arxiv 2507.02349 v1 pith:UC6D6IGH submitted 2025-07-03 cs.CV cs.AI

Rafic Nader , Vincent L'Allinec , Romain Bourcier , Florent Autrusseau This is my paper · ORCID
classification cs.CVcs.AI
keywords CircleofWillislandmarkdetectiontwo-stepneuralnetworknnheatmapregressionMRAtime-of-flightintracranialaneurysmbifurcation
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

Intracranial aneurysms tend to form at thirteen specific bifurcations of the Circle of Willis, the arterial loop at the base of the brain. This paper claims that a two-step neural network pipeline can find those bifurcation points automatically in MRA time-of-flight scans, without hand-crafted vessel centerlines or pre-segmentation. The first step uses an object detector to propose small regions of interest around candidate bifurcations; the second step uses a modified U-Net to regress heatmaps inside those regions and pick the exact coordinates. On the authors’ In-House test set the pipeline reports a mean radial error of 1.97 mm and only 9 false positives, compared with 2.38 mm and 47 false positives for SCN and 2.73 mm and 47 for FARNet. The method is also designed to handle scans where some bifurcations are absent because of anatomical variation or partial acquisition, which the compared whole-volume heatmap methods do not accommodate.

What carries the argument

The mechanism has two stages. Stage one is nnDetection, a self-configuring 3D object detector built on a Retina U-Net topology; it regresses fixed-size bounding boxes centered on bifurcations, with ten classes that group landmarks that often sit close together and look alike (A/B, C/G, D/H are paired into single ROIs). Stage two is a residual 3D U-Net with deep supervision that takes $32^{3}$ patches cropped from the proposed ROIs and regresses Gaussian heatmaps for all thirteen labels; the argmax of the relevant heatmap channel gives the coordinate, and a peak-intensity threshold decides whether a landmark is present. The heatmaps from multiple decoder levels are merged by multiplication so that only locations agreed upon across scales survive. At inference only the highest-scoring ROI per label is kept, and the second stage can only refine inside those boxes.

What would settle it

Take the In-House test set and inspect the stage-1 nnDetection proposals before the top-1 selection: if any true landmark is absent from the proposed ROIs or assigned the wrong label, the pipeline cannot recover it. Counting such cases across the 48 test volumes, especially for the MCA landmarks E and F, would directly show whether the 3 false negatives and the over-10 mm failures come from stage-1 misses rather than stage-2 localization.

Watch

Extended reading notes

Core claim

On its own terms, the central claim is that the correct architecture for vascular landmark detection is a coarse-to-fine pairing: an object-detection stage that localizes and labels candidate regions, followed by a heatmap-regression stage that refines coordinates inside those regions. The paper reports that this pairing outperforms both compared whole-volume heatmap-regression baselines on two datasets, with the largest advantage in false positives and in the MCA landmarks E and F that are visually similar and close together. It also claims the two-step structure makes the method robust to variable landmark counts, since the detector decides which landmarks exist in a given scan before the regressor tries to localize them. The aneurysm-bearing subset shows detection rates around 73 percent within 3 mm and 97 percent within 10 mm, with failure cases attributed to stage-1 ROI misclassification of MCA branches rather than to aneurysm presence.

Load-bearing premise

The whole pipeline assumes the first-stage detector always proposes a correct region of interest for every landmark that is actually present; only the highest-scoring box per label is kept, so if the right landmark is not the top box, the second stage never sees it and the landmark is lost.

Editorial extensions

If this is right

  • The method can be applied directly to raw MRA-TOF volumes; no centerline extraction, graph matching, or pre-segmentation step is required.
  • Scans with missing bifurcations, whether from anatomical aplasia or hypoplasia or from acquisition protocols that cut off the posterior circulation, still get correct per-scan landmark counts instead of forced false positives.
  • The false-positive count drops from 47 to 9 on the In-House test set because the heatmap stage never searches the full volume, only the proposed ROIs.
  • The same two-stage recipe should transfer to other anatomical landmark sets where landmarks are locally similar and variable in number.

Reading between the lines

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

  • The top-1 ROI selection rule makes recall the silent bottleneck: the paper’s own failure cases for E and F are stage-1 misclassifications, which implies that keeping a small set of candidate ROIs per label or adding a second-pass search could recover misses without re-introducing 47 false positives.
  • The paired-ROI trick suggests a general recipe: when two landmarks are close and visually confusable, merge them into one detection class and let the heatmap channels separate them.
  • On the complete-landmark public dataset the margin over SCN is smaller, so the two-step design may pay off most precisely in the variable-anatomy regime; this could be tested by removing the detector and running the same stage-2 U-Net on full-volume heatmaps, which the paper does not do.
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 / 6 minor

Summary. The paper introduces a two-step deep learning pipeline for detecting thirteen Circle of Willis (CoW) bifurcation landmarks in MRA Time-of-Flight images. In the first step, a 3D nnDetection model proposes and classifies regions of interest (ROIs) around landmarks, with some ROIs grouping nearby landmark pairs (A/B, C/G, D/H). In the second step, a modified 3D U-Net with deep supervision regresses Gaussian heatmaps within each ROI to obtain final landmark coordinates. The method is evaluated on an in-house dataset of 145 scans and a public dataset of 104 scans, and is compared against reimplementations of SCN and FARNet. The paper reports a mean radial error of 1.97 mm on the in-house test set and 1.47 mm on the public test set, with fewer false positives than the baselines, and claims the highest level of performance on the bifurcation detection task. An additional experiment examines detection rates on 134 aneurysm-bearing scans.

Significance. If the reported results hold, the two-step design is practically valuable: the ROI stage provides a natural mechanism for handling scans with variable numbers of present landmarks, and the false-positive reduction relative to global heatmap regression is clinically relevant. The paper gives credit to a clear and reproducible pipeline description, a public-dataset evaluation, and explicit reporting of false positives and false negatives. However, the central 'highest performance' claim is not yet supported: the most directly relevant prior method on the public dataset, the multi-task U-Net of [40] that produced the landmark annotations, is never compared, and all metrics are point estimates without error bars or significance tests. The strengths are real but the evidence base is currently insufficient for the strength of the claim.

major comments (3)
  1. [§III-D2, Table II] The abstract and Section III claim the 'highest level of performance,' but the most directly relevant prior method on the public dataset, the multi-task U-Net of [40] that produced the 19-landmark annotations, is never included in Tables I or II. Without [40] in the comparison, the state-of-the-art claim is unsupported. Moreover, on the public set FARNet attains SDR-5 = 96.28% versus 95.62% for the proposed method (Table II), so the proposed method is not the best on every reported metric; this discrepancy needs analysis (e.g., which landmarks and thresholds drive FARNet's advantage).
  2. [§III-C, Tables I-II] All comparison metrics are point estimates from a single train/test split, with no error bars, confidence intervals, or significance tests. The reported MRE differences (0.24–0.76 mm) may be within run-to-run variability, especially given that the SCN and FARNet baselines are reimplementations 'with their default settings' and no evidence is given that they reproduce the published methods' performance. Please add repeated-seed runs, confidence intervals, and at least a paired test (e.g., Wilcoxon signed-rank over landmarks or scans), or explicitly restrict the claim to 'comparable to' rather than 'highest.'
  3. [§II-B3, §III-E] The inference procedure keeps only the highest-scoring ROI per label, so any missed or mislabeled first-stage proposal is unrecoverable; the paper acknowledges this for landmarks E and F in the aneurysm experiment. Because the method's large false-positive reduction is achieved by this hard selection, the trade-off between missed landmarks and false positives should be quantified (e.g., by sweeping the ROI retention threshold or reporting recall of the first stage). As written, the reader cannot tell how much of the improvement comes from the ROI mechanism versus the heatmap refinement.
minor comments (6)
  1. [§II-B] There is a typo: 'we lake use of' should read 'we make use of.'
  2. [§III-D2] The text says 'median radial error (MRE) of 1.47 mm,' but MRE is defined as the mean radial error in §III-B; please correct the terminology.
  3. [§III-E] There is a typo: 'nnDetction' should be 'nnDetection.'
  4. [Fig. 9] The y-axis of Fig. 9 is unlabeled; please specify that the distance is in millimeters.
  5. [§II-A2] The text introduces '10 classes' corresponding to 10 ROIs, but three of these classes group two landmarks each; it would be clearer to state explicitly that 10 ROI classes cover the 13 landmark labels.
  6. [§II-A2, §II-B2] The bounding-box size (32 voxels), the grouping of A/B, C/G, D/H, the Gaussian sigma, and the acceptance threshold th are validation-time design choices; a sensitivity analysis, even on the validation set, would strengthen the presentation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark whose two-stage pipeline is evaluated against reimplemented baselines on held-out test sets; no derivation reduces to fitted inputs or self-cited results.

full rationale

This paper is an empirical benchmarking study, not a formal derivation, and its central claims are supported by independent test-set evaluations. The two-step pipeline (nnDetection ROI proposal followed by U-Net heatmap refinement) is a standard architecture combination; the ground-truth bounding boxes are centered on annotated landmark coordinates, and the ground-truth heatmaps are Gaussian functions of the same coordinates. Inference locates each landmark by the argmax of the relevant predicted heatmap channel, and detection thresholds are chosen on validation data and then applied to unseen test data. None of these steps equates a fitted parameter with the reported prediction: the bounding-box size is a design choice, the heatmap sigma is a fixed dispersion parameter, and the validation-derived threshold is not used to construct the ground-truth labels for the comparison methods. The only self-citation is reference [39], which supplies the In-House dataset and a narrative prior method; the paper does not rely on [39] for any equation or for the reported metric values. Comparisons are made against reimplementations of SCN [8] and FARNet [26] with default settings on the same test splits, and the reported MRE/SDR/FP numbers are direct outcomes of those runs. Concerns about the absence of a comparison with [40], the lack of error bars or significance tests, and the possibility that the reimplementations do not reproduce the published baselines are legitimate threats to the strength of the 'highest performance' claim, but they are correctness or external-validity issues, not circularity. No equation in the paper reduces by construction to its inputs, and no load-bearing premise is justified only by a self-citation. The circularity score is therefore 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The empirical claim rests on manual annotations, manually chosen bounding-box sizes, an unreported heatmap sigma, and a validation-derived acceptance threshold. The paper introduces no new physical entities or mathematical objects. The listed constants are design choices that affect the reported numbers, but none is a fitted parameter in a derivation sense; the model weights themselves are learned from data and not itemized here.

free parameters (4)
  • ROI bounding box size (single-landmark) = 32 voxels
    Manually chosen to balance contextual information against the risk of enclosing multiple landmarks (Section II-A.2).
  • ROI bounding box grouping for A/B, C/G, D/H = unified box enclosing two 32-voxel boxes
    Hand-designed to handle close landmark pairs; not derived from data (Section II-A.2).
  • Heatmap Gaussian standard deviation sigma = not reported
    Controls the spatial spread of ground-truth heatmaps; no value or tuning procedure is given (Section II-B.2, Eq. 1).
  • Heatmap acceptance threshold th = 40 (voxel intensity units)
    Selected as the first crossing of validation KDE distributions for positive versus negative heatmap channels, then applied to test data (Section III-D.1, Fig. 5).
assumptions (5)
  • domain assumption The 13 selected bifurcations are the clinically relevant landmarks and the manual annotations are correct.
    Section III-A states annotations were made by a trained operator and validated by a neuroradiologist; the entire evaluation depends on this ground truth.
  • domain assumption For the In-House dataset, a missing artery implies the corresponding landmark is absent and should not be detected.
    Section III-A.1 defines landmark absence by aplasia, hypoplasia, or acquisition truncation; this assumption shapes the variable-label-count evaluation.
  • domain assumption nnDetection's self-configuring defaults are suitable after only augmentation changes.
    Section II-A.1 trusts the rule-based and empirical parameters of nnDetection without modification, except for disabling horizontal flipping.
  • ad hoc to paper Nearby landmark pairs A/B, C/G, and D/H can be grouped into single ROIs without degrading classification.
    Section II-A.2 creates R1, R2, and R3 as unified boxes for close landmark pairs; this assumes the detector can localize one box containing two distinct landmarks and that the U-Net can disambiguate them.
  • ad hoc to paper The test-time threshold derived from validation data transfers to unseen test distributions.
    Section III-D.1 sets th=40 from the first crossing of KDE curves on validation data and applies it to test data without additional calibration or uncertainty quantification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Two-Steps Neural Networks for an Automated Cerebrovascular Landmark Detection." pith.science (2026). https://pith.science/paper/UC6D6IGH

@misc{pith2026250702349,
  author       = {Pith},
  title        = {Pith review of: Two-Steps Neural Networks for an Automated Cerebrovascular Landmark Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UC6D6IGH}},
  note         = {Machine review of arXiv:2507.02349}
}
read the original abstract

Intracranial aneurysms (ICA) commonly occur in specific segments of the Circle of Willis (CoW), primarily, onto thirteen major arterial bifurcations. An accurate detection of these critical landmarks is necessary for a prompt and efficient diagnosis. We introduce a fully automated landmark detection approach for CoW bifurcations using a two-step neural networks process. Initially, an object detection network identifies regions of interest (ROIs) proximal to the landmark locations. Subsequently, a modified U-Net with deep supervision is exploited to accurately locate the bifurcations. This two-step method reduces various problems, such as the missed detections caused by two landmarks being close to each other and having similar visual characteristics, especially when processing the complete MRA Time-of-Flight (TOF). Additionally, it accounts for the anatomical variability of the CoW, which affects the number of detectable landmarks per scan. We assessed the effectiveness of our approach using two cerebral MRA datasets: our In-House dataset which had varying numbers of landmarks, and a public dataset with standardized landmark configuration. Our experimental results demonstrate that our method achieves the highest level of performance on a bifurcation detection task.

Figures

Figures reproduced from arXiv: 2507.02349 by the authors.

Figure 1
Figure 1. Schematic representation of the CoW.The yellow labels [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Schematic representation of the proposed landmark detection pipeline. The pipeline consists of 2 steps: detection of [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Retina U-Net uses a backbone network (stacked convolutions) that extracts features at multiple scales. This backbone [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: UNet variant for heatmap regression. bounding boxes enclosing two separate bifurcations of interest. To do so, we lake use of a second neural network as explained in the following. 1) Neural Network: To find the exact location of each landmark within the ROI patch obta…
Figure 5
Figure 5. Figure 5: Voxel intensity distribution analysis of heatmaps [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Mean radial errors (mm) for 13 landmark labels for the In-House dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Qualitative Comparisons between different methods. Red color refer to GT landmarks or landmarks predicted within [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Mean radial errors(mm) for 11 landmarks labels for the public dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Distance distribution between the predicted bifurcations [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Qualitative results on landmark detection with the presence of aneurysms [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [40]

    Multi-task global optimization-based method for vascular landmark detection,

    Z. Tan, J. Feng et al. , “Multi-task global optimization-based method for vascular landmark detection,” Computerized Medical Imaging and Graphics, vol. 114, p. 102364, 2024

    work page 2024

  2. [39]

    Using deep learning for an automatic detection and classification of the vascular bifurcations along the circle of willis,

    R. Nader, R. Bourcier, and F. Autrusseau, “Using deep learning for an automatic detection and classification of the vascular bifurcations along the circle of willis,” Medical Image Analysis , p. 102919, 2023

    work page 2023

  3. [1]

    Completeness of the circle of willis and risk of ischemic stroke in patients without cerebrovascular disease,

    T. van Seeters, J. Hendrikse et al., “Completeness of the circle of willis and risk of ischemic stroke in patients without cerebrovascular disease,” Neuroradiology, vol. 57, pp. 1247–1251, 2015

    work page 2015

  4. [2]

    Relationship of a1 segment hypoplasia to anterior communicating artery aneurysm morphology and risk factors for aneurysm formation,

    L. Rinaldo, B. A. McCutcheon et al. , “Relationship of a1 segment hypoplasia to anterior communicating artery aneurysm morphology and risk factors for aneurysm formation,” J. Neurosurg., vol. 127(1), 2016

    work page 2016

  5. [3]

    Cerebral aneurysms and variations in the circle of willis

    K. Kayembe, M. Sasahara, and F. Hazama, “Cerebral aneurysms and variations in the circle of willis.” Stroke, vol. 15(5), pp. 846–850, 1984

    work page 1984

  6. [4]

    Bifurcation geometry remodelling of vessels in de novo and growing intracranial aneurysms: a multicenter study,

    J. Boucherit, B. Kerleroux et al., “Bifurcation geometry remodelling of vessels in de novo and growing intracranial aneurysms: a multicenter study,” Journal of NeuroInterventional Surgery , 2022

    work page 2022

  7. [5]

    The hemodynamic alterations induced by the vascular angular deformation in stent-assisted coiling of bifurcation aneurysms,

    W. Jeong, M. Han, and K. Rhee, “The hemodynamic alterations induced by the vascular angular deformation in stent-assisted coiling of bifurcation aneurysms,” Computers in Biology and Medicine , vol. 53, pp. 1–8, 2014

    work page 2014

  8. [6]

    A benchmark for comparison of dental radiography analysis algorithms,

    C.-W. Wang, C.-T. Huang et al., “A benchmark for comparison of dental radiography analysis algorithms,” Medical image analysis , vol. 31, pp. 63–76, 2016

    work page 2016

Show all 53 references
  1. [7]

    Automatic landmark estimation for adolescent idiopathic scoliosis assessment using boostnet,

    H. Wu, C. Bailey et al., “Automatic landmark estimation for adolescent idiopathic scoliosis assessment using boostnet,” in MICCAI 2017: 20th Intl Conf., Quebec, Canada, Sept. 11-13, 2017 , 2017, pp. 127–135

    2017

  2. [8]

    Integrating spatial configuration into heatmap regression based cnns for landmark localization,

    C. Payer, D. ˇStern et al., “Integrating spatial configuration into heatmap regression based cnns for landmark localization,” Med. Image Anal. , vol. 54, pp. 207–219, 2019

    2019

  3. [9]

    Vertebrae localization in ct using both local and global symmetry features,

    K. Kim and S. Lee, “Vertebrae localization in ct using both local and global symmetry features,” Computerized Medical Imaging and Graphics, vol. 58, pp. 45–55, 2017

    2017

  4. [10]

    Automatic localization of cephalometric landmarks,

    V . Grau, M. Alcaniz et al. , “Automatic localization of cephalometric landmarks,” J. of Biomedical Informatics, vol. 34(3), pp. 146–156, 2001

    2001

  5. [11]

    3D statistical shape models incorporating landmark-wise random regression forests for omni- directional landmark detection,

    T. Norajitra and K. H. Maier-Hein, “3D statistical shape models incorporating landmark-wise random regression forests for omni- directional landmark detection,” IEEE Trans. on Medical Imaging , vol. 36(1), 2016

    2016

  6. [12]

    Automatic landmark detection in cephalometry using a modified active shape model with sub image matching,

    R. Kafieh, A. Mehri, and S. Sadri, “Automatic landmark detection in cephalometry using a modified active shape model with sub image matching,” in Int. Conf. on Machine Vision . IEEE, 2007, pp. 73–78

    2007

  7. [13]

    Stratified decision forests for accurate anatomical landmark localization in cardiac images,

    O. Oktay, W. Bai et al. , “Stratified decision forests for accurate anatomical landmark localization in cardiac images,” IEEE Trans. on Medical Imaging, vol. 36(1), pp. 332–342, 2016

    2016

  8. [14]

    From local to global random regression forests: exploring anatomical landmark localization,

    D. ˇStern, T. Ebner, and M. Urschler, “From local to global random regression forests: exploring anatomical landmark localization,” in MICCAI : 19th Intl. Conf., Athens, Greece, Oc. 17-21, 2016 , 2016

    2016

  9. [15]

    Integrating geometric configuration and appearance information into a unified framework for anatomical landmark localization,

    M. Urschler, T. Ebner, and D. ˇStern, “Integrating geometric configuration and appearance information into a unified framework for anatomical landmark localization,” Med. Image Anal. , vol. 43, pp. 23–36, 2018

    2018

  10. [16]

    Cephalometric landmark detection in dental x-ray images using convolutional neural networks,

    H. Lee, M. Park, and J. Kim, “Cephalometric landmark detection in dental x-ray images using convolutional neural networks,” in SPIE Medical imaging 2017 , vol. 10134, 2017, pp. 494–499

    2017

  11. [17]

    Detecting anatomical landmarks from limited medical imaging data using two-stage task-oriented deep neural networks,

    J. Zhang, M. Liu, and D. Shen, “Detecting anatomical landmarks from limited medical imaging data using two-stage task-oriented deep neural networks,” IEEE Trans. on Image Proc., vol. 26(10), pp. 4753–64, 2017

    2017

  12. [18]

    Deep learning-based regression and classification for automatic landmark localization in medical images,

    J. M. Noothout, B. D. De V os et al., “Deep learning-based regression and classification for automatic landmark localization in medical images,” IEEE Trans. on medical imaging , vol. 39(12), pp. 4011–4022, 2020

    2020

  13. [19]

    Cascaded convolutional networks for automatic cephalometric landmark detection,

    M. Zeng, Z. Yan et al., “Cascaded convolutional networks for automatic cephalometric landmark detection,” Medical image analysis , vol. 68, 2021

    2021

  14. [20]

    3D deep learning for efficient and robust landmark detection in volumetric data,

    Y . Zheng, D. Liu et al. , “3D deep learning for efficient and robust landmark detection in volumetric data,” in MICCAI : 18th International Conference, Munich, Germany, Oct. 5-9, 2015 , 2015, pp. 565–572

    2015

  15. [21]

    Regressing heatmaps for multiple landmark localization using cnns,

    C. Payer, D. ˇStern et al. , “Regressing heatmaps for multiple landmark localization using cnns,” in Intl Conf. on medical image computing and computer-assisted intervention. Springer, 2016, pp. 230–238

    2016

  16. [22]

    Cephalometric landmark detection by attentive feature pyramid fusion and regression-voting,

    R. Chen, Y . Ma et al., “Cephalometric landmark detection by attentive feature pyramid fusion and regression-voting,” in MICCAI : 22nd Intl. Conf., Shenzhen, China, Oct. 13–17, 2019 , 2019, pp. 873–881

    2019

  17. [23]

    Hip landmark detection with dependency mining in ultrasound image,

    J. Xu, H. Xie et al., “Hip landmark detection with dependency mining in ultrasound image,” IEEE Transactions on Medical Imaging , vol. 40, no. 12, pp. 3762–3774, 2021

    2021

  18. [24]

    Deep anatomical context feature learning for cephalometric landmark detection,

    K. Oh, I.-S. Oh et al. , “Deep anatomical context feature learning for cephalometric landmark detection,” IEEE Journal of Biomedical and Health Informatics, vol. 25, no. 3, pp. 806–817, 2020

    2020

  19. [25]

    Multi-input adaptive neural network for automatic detection of cervical vertebral landmarks on x-rays,

    Y . Wang, L. Huang et al. , “Multi-input adaptive neural network for automatic detection of cervical vertebral landmarks on x-rays,” Computers in Biology and Medicine , vol. 146, 2022

    2022

  20. [26]

    Feature aggregation and refinement network for 2d anatomical landmark detection,

    Y . Ao and H. Wu, “Feature aggregation and refinement network for 2d anatomical landmark detection,” Journal of Digital Imaging , vol. 36, no. 2, pp. 547–561, 2023

    2023

  21. [27]

    An artificial agent for anatomical landmark detection in medical images,

    F. C. Ghesu, B. Georgescu et al. , “An artificial agent for anatomical landmark detection in medical images,” in MICCAI : 19th Intl. Conf., Athens, Greece, Oct. 17-21, 2016 , 2016, pp. 229–237

    2016

  22. [28]

    Evaluating reinforcement learning agents for anatomical landmark detection,

    A. Alansary, O. Oktay et al., “Evaluating reinforcement learning agents for anatomical landmark detection,” Medical image analysis , vol. 53, pp. 156–164, 2019

    2019

  23. [29]

    Cephanet: An improved faster r-cnn for cephalometric landmark detection,

    J. Qian, M. Cheng et al. , “Cephanet: An improved faster r-cnn for cephalometric landmark detection,” in IEEE 16th Intl. symposium on biomedical imaging (ISBI) , 2019, pp. 868–871

    2019

  24. [30]

    Fast and accurate craniomaxillofacial landmark detection via 3d faster r-cnn,

    X. Chen, C. Lian et al., “Fast and accurate craniomaxillofacial landmark detection via 3d faster r-cnn,” IEEE Transactions on Medical Imaging , vol. 40, no. 12, pp. 3867–3878, 2021

    2021

  25. [31]

    Craniomaxillofacial bone segmentation and landmark detection using semantic segmentation networks and an unbiased heatmap,

    R. Zhang, B. Jie et al. , “Craniomaxillofacial bone segmentation and landmark detection using semantic segmentation networks and an unbiased heatmap,” IEEE Journal of Biomedical and Health Informatics, vol. 28, no. 1, pp. 427–437, 2024

    2024

  26. [32]

    Anatomical labeling of the circle of willis using maximum a posteriori probability estimation,

    H. Bogunovic, J. M. Pozo et al. , “Anatomical labeling of the circle of willis using maximum a posteriori probability estimation,” IEEE Transactions on Medical Imaging , vol. 32, no. 9, pp. 1587–1599, 2013

    2013

  27. [33]

    Simultaneous segmentation and anatomical labeling of the cerebral vasculature,

    D. Robben, E. T ¨uretken et al. , “Simultaneous segmentation and anatomical labeling of the cerebral vasculature,” Medical Image Analysis, vol. 32, pp. 201–215, 2016

    2016

  28. [34]

    Automated anatomical labeling of the cerebral arteries using belief propagation,

    M. Bilgel, S. Roy et al., “Automated anatomical labeling of the cerebral arteries using belief propagation,” in Medical Imaging 2013: Image Processing, vol. 8669. SPIE, 2013, pp. 340–345

    2013

  29. [35]

    Gurobi optimizer reference manual,

    I. Gurobi et al. , “Gurobi optimizer reference manual,” Optimization, 2015. 11

    2015

  30. [36]

    A stereotactic probabilistic atlas for the major cerebral arteries,

    T. Dun ˚as, A. W ˚ahlin et al. , “A stereotactic probabilistic atlas for the major cerebral arteries,” Neuroinformatics, vol. 15(1), pp. 101–10, 2017

    2017

  31. [37]

    Automatic labeling of vascular structures with topological constraints via HMM,

    X. Wang, Y . Liu et al., “Automatic labeling of vascular structures with topological constraints via HMM,” in MICCAI, 2017, pp. 208–215

    2017

  32. [38]

    3D vascular skeleton extraction and decomposition,

    A. Chowriappa, Y . Seo et al. , “3D vascular skeleton extraction and decomposition,” IEEE J. of Biomedical and Health Informatics , vol. 18(1), pp. 139–147, 2014

    2014

  33. [41]

    The role of circle of willis anomalies in cerebral aneurysm rupture,

    M. A. Lazzaro, B. Ouyang, and M. Chen, “The role of circle of willis anomalies in cerebral aneurysm rupture,” Journal of neurointerventional surgery, vol. 4, no. 1, pp. 22–26, 2012

    2012

  34. [42]

    Unruptured intracranial aneurysms: epidemiology, natural history, management options, and familial screening,

    R. D. Brown and J. P. Broderick, “Unruptured intracranial aneurysms: epidemiology, natural history, management options, and familial screening,” The Lancet Neurology , vol. 13, no. 4, pp. 393–404, 2014

    2014

  35. [43]

    An overview of intracranial aneurysms,

    A. Keedy, “An overview of intracranial aneurysms,” McGill Journal of Medicine, vol. 9, no. 2, pp. 141–146, 2006

    2006

  36. [44]

    A comprehensive study of the anatomical variations of the circle of willis in adult human brains,

    S. Iqbal, “A comprehensive study of the anatomical variations of the circle of willis in adult human brains,” Journal of clinical and diagnostic research: JCDR, vol. 7, no. 11, p. 2423, 2013

    2013

  37. [45]

    Anatomical variations of the circle of willis: evaluation with ct angiography,

    H. S ¸ahin and Y . Pekc ¸evik, “Anatomical variations of the circle of willis: evaluation with ct angiography,” Anatomy, vol. 12(1), pp. 20–26, 2018

    2018

  38. [46]

    nnDetection: a self-configuring method for medical object detection,

    M. Baumgartner, P. F. J ¨ager et al. , “nnDetection: a self-configuring method for medical object detection,” in MICCAI : 24th Intl. Conf., Strasbourg, France, Sept. 27–October 1, 2021 , 2021, pp. 530–539

    2021

  39. [47]

    Retina u-net: Embarrassingly simple exploitation of segmentation supervision for medical object detection,

    P. F. Jaeger, S. A. Kohl et al. , “Retina u-net: Embarrassingly simple exploitation of segmentation supervision for medical object detection,” in Machine Learning for Health Workshop. PMLR, 2020, pp. 171–183

    2020

  40. [48]

    U-Net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in MICCAI: 18th Intl. Conf., Munich, Germany, Oct.5-9, 2015 , vol. 9351, 2015, pp. 234–241

    2015

  41. [49]

    Deep residual learning for image recognition,

    K. He, X. Zhang et al., “Deep residual learning for image recognition,” in IEEE Conf. Computer Vision and Pattern Recog., 2016, pp. 770–778

    2016

  42. [50]

    A multiscale patch based convolutional network for brain tumor segmentation,

    J. Stawiaski, “A multiscale patch based convolutional network for brain tumor segmentation,” arXiv preprint arXiv:1710.02316 , 2017

    2017 arXiv

  43. [51]

    Dilated deeply supervised networks for hippocampus segmentation in mri,

    L. Folle, S. Vesal et al. , “Dilated deeply supervised networks for hippocampus segmentation in mri,” in Bildverarbeitung f ¨ur die Medizin 2019: Algorithmen–Systeme–Anwendungen. Proceedings des Workshops vom 17. bis 19. M ¨arz 2019 in L ¨ubeck. Springer, 2019, pp. 68–73

    2019

  44. [52]

    Brave-net: Fully automated arterial brain vessel segmentation in patients with cerebrovascular disease,

    A. Hilbert, V . I. Madai et al., “Brave-net: Fully automated arterial brain vessel segmentation in patients with cerebrovascular disease,” Frontiers in Artificial Intelligence , vol. 3, 2020

    2020

  45. [53]

    Benchmarking the cow with the topcow challenge: Topology-aware anatomical segmentation of the circle of willis for CTA and MRA,

    K. Yang, F. Musio et al. , “Benchmarking the cow with the topcow challenge: Topology-aware anatomical segmentation of the circle of willis for CTA and MRA,” ArXiv, 2023

    2023

Pith tools

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